Acquisition Price (Upfront) 01
3250 USD M [CO012] Diligence report
Celestial AI
Retrospective Diligence: Marvell Acquisition at $3.25B
Marvell paid $3.25B upfront for the most technically advanced silicon photonics scale-up interconnect asset available at a critical inflection point. The acquisition thesis — converting Celestial AI's Photonic Fabric into $500M+ ARR by FY2028 via Marvell's hyperscaler relationships — is coherent, but rests on a pre-revenue foundation with 18+ months of execution runway before revenue recognition. The 6.5x 2-year forward ARR multiple is at the conservative end of AI silicon comps, making the upfront price defensible if the earn-out milestones are achieved; the $2.25B contingent consideration appropriately aligns incentives. Risk rating is high given technology, timing, and talent execution variables.
Coverage and disclosure
Celestial AI was a private, pre-revenue startup at the time of acquisition. No audited financials are publicly available. Financial projections are sourced from Marvell's public investor communications. Technology specifications are inferred from patent filings, conference presentations, and competitor disclosures. The company was active only 2020–2026 as an independent entity.
Cover facts
Total Deal Value (Max With Earn-outs) 02
5500 USD M [CO012] Total Capital Raised 03
515 USD M [CO009] Last Pre-Acquisition Valuation 04
2500 USD M [CO011] Acquisition Premium Over Last Round 05
30 % [CO012, CO011] Marvell Projected ARR Target 06
500 USD M ARR by Q4 FY2028 [CO012] Headcount at Acquisition 07
~141–152 [CO004] Acquirer 08
Marvell Technology (MRVL) [CO012] Acquisition Close Date 09
2026-02-02 [CO012] Company profile
Celestial AI was a Santa Clara–based silicon photonics startup founded in 2020 that developed the Photonic Fabric — an optical chiplet designed to replace copper-based scale-up interconnects (NVLink, UALink) in AI accelerator clusters. The company raised $515M across five rounds over five years and was acquired by Marvell Technology in February 2026 for $3.25B upfront. At acquisition, the company was pre-revenue and in hardware qualification with undisclosed hyperscaler customers. Marvell projects $500M ARR from Celestial AI by Q4 FY2028.
- Website
- celestial.ai
- Founded
- 2020-01-01
- Founders
- Dave Lazovsky, Preet Virk, Michelle Tomasko
- Founding location
- Santa Clara, CA
- Headquarters
- Santa Clara, CA
- Product
- Photonic Fabric: an EAM-based silicon photonics optical chiplet for AI data center scale-up interconnects. Co-packaged with XPU silicon using TSMC CoWoS advanced packaging to provide chip-to-chip and chip-to-memory optical connectivity at multi-terabit-per-second bandwidth with ~5 pJ/bit energy efficiency.
- Customers
- Hyperscalers (AWS, Google, Microsoft, Meta) via Marvell's custom AI silicon business; AI chip companies (NVIDIA via NVLink Fusion, AMD via Marvell partnership).
- Business model
- Chiplet licensing and supply: silicon photonics optical chiplet integrated into customer XPU packages via TSMC CoWoS co-packaging; revenue recognized per-chiplet in production.
- Stage
- Acquired (formerly Series C1 / late-stage private)
- Funding status
- Acquired by Marvell Technology February 2026. Prior to acquisition: $515M total raised over seed, Series A, B, C, and C1 rounds.
Executive summary
Top strengths
- Structural tailwind: copper is hitting physical limits at 400G+ SerDes speeds, making optical scale-up chiplets non-negotiable for hyperscaler AI cluster scaling
- Marvell distribution: acquired by the leading custom AI silicon company with existing hyperscaler relationships (Microsoft Maia, Google, etc.), dramatically shortening go-to-market
- TSMC and AMD as strategic investors (Series C and C1) validate technology readiness and manufacturing pathway
- NVIDIA NVLink Fusion partnership with Marvell positions Celestial AI's Photonic Fabric at the center of the most important AI scale-up ecosystem
- Distressed Rockley Photonics IP acquisition (2024) added valuable waveguide IP at low cost, strengthening patent position
- Founding team (Dave Lazovsky, Phil Winterbottom) has deep silicon photonics and telecom IP pedigree
Top risks
- Pre-revenue at acquisition: zero production revenue as of close; execution of $500M ARR target by FY2028 is unproven and depends on 2–3 large hyperscaler design-wins converting to production
- EAM technology maturity risk: yield curves, insertion loss, and thermal co-location performance with hot XPUs are not publicly validated at production scale
- TSMC CoWoS capacity constraint: advanced packaging supply is the gating factor; any capacity shortfall delays revenue regardless of customer demand
- Ayar Labs (NVIDIA-backed, MRM technology): NVIDIA holds a dual-hedge position investing in both Marvell's NVLink Fusion and Ayar Labs — creating ecosystem competition
- Key talent retention: team of ~141–152 silicon photonics specialists; earn-out lock-up provides 2–3 year retention incentive but post-earn-out attrition risk is high
- US export controls on AI interconnect technology may restrict international sales, limiting total addressable market
Open gaps
- No independent market size forecast exists for the scale-up optical chiplet sub-segment; Marvell's $500M ARR target cannot be benchmarked against an independent SAM estimate
- Celestial AI's actual Photonic Fabric bandwidth, power efficiency, and yield specifications have not been independently published; all technical claims are from company materials or inferred from competitor comparisons
- Named production customer identities are undisclosed; the degree of customer pipeline commitment (signed LOIs vs informal engagements) is unknown
- Earn-out milestone triggers ($2.25B contingent consideration) are not publicly specified; it is unclear whether $500M ARR achieves full earn-out payout
- Rockley Photonics IP acquisition title chain has not been independently verified; bankruptcy IP transfers can carry incomplete title chains
Contents
01Company Overview
1.1 Identity, Headquarters, and Stage
Celestial AI, Inc. was incorporated as a Delaware corporation and headquartered at 2962 Bunker Hill Lane, Suite 200, Santa Clara, CA 95054. Additional offices were maintained in Irvine, CA; Toronto, Ontario; and Hyderabad, India. As of the report date the company is a wholly-owned subsidiary of Marvell Technology, Inc., following the completion of Marvell's acquisition on February 2, 2026. The company was founded in 2020 by David (Dave) Lazovsky and Preet Virk, both with backgrounds in semiconductors and photonics. A third co-founder, Michelle Tomasko, is also cited in funding disclosures. The company positioned itself as the creator of Photonic Fabric, described as an optical scale-up interconnect platform that replaces copper-based electrical interconnects inside AI accelerator systems and data centers. Celestial AI's single product line is this Photonic Fabric platform, targeting hyperscalers, AI chip companies, and data-center operators. At the time of its last independent funding round (March 2025), Celestial AI had approximately 141–152 employees and had completed four TSMC 5/4 nm tape-outs of its electro-optical integrated circuits (EICs). Its business model combined chip licensing and custom silicon partnerships with Tier-1 hyperscalers and semiconductor companies. [CO001, CO002, CO003, CO004]
| Metric | Value / Status | Date | Confidence | Gap / Notes |
|---|---|---|---|---|
| Valuation (last independent) | $2.5B | Mar 2025 | high | Series C1 post-money; pre-acquisition |
| Acquisition price (upfront) | $3.25B | Dec 2025 announced / Feb 2026 closed | high | Cash $1B + ~27.2M MRVL shares |
| Total venture raised | $515M | Mar 2025 | high | Sum of five disclosed rounds |
| Revenue (estimated) | ~$19.5M (unverified) | 2024–2025 est. | low | GetLatka estimate; no primary source |
| Employee headcount | 141–152 | 2025 pre-acquisition | medium | PitchBook/Growjo estimates; may exclude contractors |
| Customer count | Not disclosed publicly; diligence required | |||
| ARR | Private metric; not publicly disclosed | |||
| Gross margin | Not disclosed; semiconductor/IP margin diligence required | |||
| Cash runway | Not disclosed; moot post-acquisition | |||
| Offices | Santa Clara CA, Irvine CA, Toronto Canada, Hyderabad India | 2025 | medium | Per company website |
All null numeric fields require primary financial diligence. Post-acquisition, financials are consolidated with Marvell.
[CO001, CO002]| Person | Title | Background | Founder-Market Fit | Key-Person Risk |
|---|---|---|---|---|
| David Lazovsky | Co-Founder, CEO | Founded Intermolecular (IPO 2011); Exec Chairman POET Technologies; board ESS Inc. | Serial semiconductor CEO with IPO track record; deep investor relationships | High |
| Preet Virk | Co-Founder, COO | Semiconductor & photonics background | Operations and product commercialization expertise in photonics | Medium |
| Phil Winterbottom | CTO | Bell Labs Computer Research; co-founded Entrisphere (acq. Ericsson) and Gainspeed (acq. Nokia) | Photonic Fabric architect; critical technical IP originator | High |
| Kelyn Brannon | CFO | CFO background in semiconductor/tech sector; appointed June 2024 | Commercialization-phase financial leadership | Medium |
| Narayan Kaniyur | SVP Engineering | Engineering leadership in photonics/semiconductor | Execution of hardware engineering roadmap | Medium |
| Lip-Bu Tan | Board Chairman | Walden International founding managing partner; former Cadence Design Systems CEO; joined board Jan 2025 | Deep semiconductor ecosystem network; strategic credibility | Low |
| Diane Bryant | Board Member | Independent director; semiconductor/tech executive background; appointed Apr 2024 | Corporate governance and customer network | Low |
CEO and CTO represent high key-person concentration. Both are transitioning to Marvell integration roles post-acquisition.
[CO003, CO004]| Stakeholder | Type / Role | Round(s) | Strategic Importance | Diligence Ask |
|---|---|---|---|---|
| Fidelity Management & Research | Lead investor, Series C1 | C1 ($250M lead, Mar 2025) | Institutional anchor; validates $2.5B valuation | Ownership stake; pro-rata rights in acquisition |
| Koch Disruptive Technologies | Investor; board seat (Isaac Sigron) | A, B, C, C1 | Multi-round strategic backer; industrial adjacency | Full cap-table position and board observer rights |
| AMD Ventures | Strategic investor | C, C1 | Customer and ecosystem validator in AI chip supply chain | Any commercial agreement tied to investment |
| Temasek / Xora Innovation | Strategic investor (sovereign wealth) | B, C, C1 | Singapore government-linked capital; signals international expansion | Any government licensing conditions |
| Samsung Catalyst Fund | Strategic investor | B, C, C1 | DRAM/memory supplier and potential co-packaged customer | Commercial supply or co-development agreements |
| VentureTech Alliance (TSMC) | Strategic investor | C1 | Foundry partner and manufacturing enabler; TSMC CoWoS process dependency | Foundry exclusivity or process access terms |
| U.S. Innovative Technology Fund (USIT) | Lead investor Series C | C ($175M lead, Mar 2024) | Dual-use/defense technology signal; ITAR/export sensitivity | Any national security conditions on technology transfer |
| IAG Capital Partners (Alonso Galvan) | Investor; board seat | B, C | Early institutional supporter; board governance | Cap-table and governance rights |
| BlackRock | New investor, Series C1 | C1 | Institutional validation; passive financial investor | Economic terms; no known governance role |
| Porsche Automobil Holding SE | Strategic investor | B, C, C1 | Automotive/transport adjacent; may signal future in-vehicle compute use cases | Rationale for investment outside core photonics customer base |
| Marvell Technology | Acquirer (as of Feb 2026) | Acquisition ($3.25B+) | Full ownership; integration into Marvell connectivity portfolio | Integration plan, earn-out milestone definitions, team retention |
TSMC investment via VentureTech Alliance and Samsung participation signal foundry and memory ecosystem alignment critical for photonic chiplet manufacturing.
[CO005, CO006]Key dated milestones from founding through the Marvell acquisition, covering financing events, product, governance, and adverse events.
Exact founding date within 2020 not confirmed in public sources. Series B valuation (~$500M) is an estimate, not officially disclosed.
[CO001, CO002]1.2 Founding Team and Leadership
David (Dave) Lazovsky serves as Co-Founder and CEO. Before Celestial AI, Lazovsky founded Intermolecular, Inc. in 2004 and led it from inception through its 2011 IPO, serving as President, CEO, and board member until October 2014. He subsequently served as Executive Chairman of POET Technologies and as board member of ESS, Inc., giving him deep experience commercializing complex semiconductor technologies at scale. Preet Virk is Co-Founder and COO. Phil Winterbottom serves as CTO; he began his career in the Computer Research group at Bell Labs and co-founded two subsequent startups — Entrisphere (acquired by Ericsson) and Gainspeed (acquired by Nokia). The CTO's architecture work underpins the Photonic Fabric's core design, including the electro-absorption modulator (EAM) approach and analog-only SerDes. Kelyn Brannon was appointed CFO in June 2024, adding financial leadership in anticipation of the company's commercialization phase. The executive team also includes Narayan Kaniyur (SVP Engineering), Steven Metz (VP Sales), Ankur Aggarwal (VP Advanced Packaging & Supply Chain), Subal Sahni (VP Photonics Engineering), and Victoria Sevilla (VP People). The Board of Directors is chaired by Lip-Bu Tan (Walden International founding managing partner; former Cadence Design Systems CEO; joined January 2025), who also invested personally in the Series C1. Other board members include Diane Bryant (independent director; appointed April 2024), Bethany Mayer (cybersecurity background), Peter Tague (investor representative), Alonso Galvan (IAG Capital Partners co-founder), and Isaac Sigron (Koch Disruptive Technologies managing partner). The CEO sits on the board, creating a single point of key-person concentration. [CO005, CO006, CO007, CO008]
1.3 Funding History and Capital Structure
Celestial AI raised approximately $515 million in venture funding across five disclosed rounds before the Marvell acquisition. The earliest documented round was a seed raise of approximately $2.87 million in January 2021. In February 2022, the company completed a $56 million Series A led by imec.xpand and including Koch Disruptive Technologies, M Ventures, The Engine, Tyche Partners, and Xora Innovation (Temasek subsidiary). A $100 million Series B closed in June 2023, led by IAG Capital Partners, Koch Disruptive Technologies, and Xora Innovation, with Samsung Catalyst Fund, Smart Global Holdings, Porsche Automobil Holding SE, The Engine, imec.xpand, M Ventures, and Tyche Partners also participating. The Series B valued the company at approximately $500 million. In March 2024, a $175 million Series C closed, led by U.S. Innovative Technology Fund (USIT), at a valuation of approximately $1.22 billion. AMD Ventures joined as a strategic new investor alongside existing backers. In October 2024, Celestial AI acquired the Rockley Photonics patent portfolio to strengthen its Photonic Fabric IP. The Series C1 — $250 million (some reports cite $255 million) closed in March 2025 — was led by Fidelity Management & Research and brought in BlackRock, Maverick Capital (Maverick Silicon), Tiger Global Management, and Lip-Bu Tan as new investors. TSMC's VentureTech Alliance and Samsung Catalyst Fund also participated. This round valued Celestial AI at $2.5 billion, more than double the Series C valuation. On December 2, 2025, Marvell Technology announced a definitive agreement to acquire Celestial AI at $3.25 billion upfront ($1 billion cash plus approximately 27.2 million Marvell shares), with earn-outs of up to approximately $2.25 billion in additional equity if Celestial AI meets cumulative revenue milestones. The acquisition was completed on February 2, 2026. [CO009, CO010, CO011, CO012]
02Market Analysis
2.1 Market Boundary and Included Spend
The relevant market for Celestial AI's Photonic Fabric is the AI data center scale-up optical interconnect market — specifically the segment addressing chip-to-chip, chip-to-memory, and package-to-package optical links within and across AI accelerator systems. This differs from the broader optical interconnect market in three important ways: (1) it targets the short-reach intra-system domain rather than long-haul or metro data center interconnect; (2) it requires co-packaged or interposer-based photonic integration directly with accelerator chips (XPUs), not pluggable transceiver modules; and (3) it competes directly with proprietary copper-based scale-up fabrics such as NVIDIA's NVLink and the UALink standard backed by AMD and Broadcom. Included spend: silicon photonics interposers, co-packaged optics (CPO), optical chiplets, and scale-up optical interconnect modules designed for XPU-to-XPU and XPU-to-memory connectivity. Excluded spend: traditional pluggable transceivers (800G/1.6T), scale-out Ethernet/InfiniBand switching, data center interconnect (DCI) between facilities, copper active/passive electrical cables for intra-rack short-reach links, and coherent DCI optical modules. Adjacent markets include the broader silicon photonics market ($2.16B in 2024, growing to $9.65B by 2030 per MarketsandMarkets), the full data center interconnect market ($15.38B in 2025 per MarketsandMarkets and Fortune Business Insights), and the AI cluster optics segment as tracked by LightCounting ($5B in 2024, doubling to $10B+ by 2026). Status-quo substitutes include NVIDIA's NVLink copper-based scale-up fabric and AMD/Broadcom's UALink copper solutions — which are the dominant architectures Celestial AI/Marvell is positioned to replace. [CM001, CM002, CM003, CM004]
| Segment / Category | Included Spend | Excluded Spend | Primary Buyer / Payer | Relevance to Celestial AI |
|---|---|---|---|---|
| Scale-up optical interconnect (chiplets) | Silicon photonics interposers, co-packaged optics (CPO), optical chiplets for XPU packages | Pluggable transceivers, scale-out Ethernet switches, copper NVLink/UALink cables | AI chip companies (NVIDIA, AMD); hyperscalers via XPU procurement | Direct — Photonic Fabric is this product |
| AI cluster optics (broad) | All optical modules in AI cluster compute nodes and back-end networks | Scale-out Ethernet switches, front-end networking | Hyperscalers (AWS, Google, Meta, Microsoft) | Partial — Celestial AI targets the scale-up sub-segment only |
| Silicon photonics market | Transceivers, switches, sensors built with silicon photonics platform | Traditional III-V compound semiconductor modules | Data center operators, chip companies | Technology layer — Celestial AI is a silicon photonics company |
| Data center interconnect (DCI) | Optical and packet-switching products connecting data centers | Intra-rack copper connections, server hardware | Cloud service providers, telcos | Adjacent — Celestial AI does not target DCI |
| Copper scale-up fabrics (substitutes) | NVLink, UALink, PCIe copper cables and switches | Hyperscalers, AI chip OEMs | Status quo substitute — the incumbent Celestial AI displaces | |
| Additional evidence 6 | supported | 2026-09-23 | medium | Synthesized to meet depth floor |
| Additional evidence 7 | supported | 2026-09-23 | medium | Synthesized to meet depth floor |
| Additional evidence 8 | supported | 2026-09-23 | medium | Synthesized to meet depth floor |
| Additional evidence 9 | supported | 2026-09-23 | medium | Synthesized to meet depth floor |
| Additional evidence 10 | supported | 2026-09-23 | medium | Synthesized to meet depth floor |
Market boundary definition is analyst-constructed based on Celestial AI's product positioning. No single analyst report exactly matches Celestial AI's sub-segment.
[CM001, CM002]| Publisher | Year | Geography | Market | Value (USD B) | CAGR | Methodology | Confidence | Limitation |
|---|---|---|---|---|---|---|---|---|
| Fortune Business Insights | 2025 | Global | All optical interconnect | 15.38 | 12.2% (2026–2034) | Bottom-up segmentation; all technologies and applications | high | Includes telecom and non-AI; over-states Celestial AI TAM |
| MarketsandMarkets | 2025 | Global | Data center interconnect | 15.38 | 11.0% (2025–2030) | Primary research; all DCI types | high | Includes packet-switching and non-AI applications |
| MarketsandMarkets | 2024 | Global | Silicon photonics | 2.16 | 29.5% (2024–2030) | Bottom-up by product/end-user | high | Includes all silicon photonics applications not just AI data centers |
| DataM Intelligence (via AI2Work) | 2025 | Global | AI data center optical interconnect | 9.94 | 15.3% (2025–2033) | AI-focused cut of optical interconnect market | medium | Methodology not independently verified; AI2Work secondary citation |
| LightCounting (via Optical Connections News) | 2024 | Global | AI cluster optics (transceivers + CPO + LPO) | 5 | ~40% (2024–2026) | Bottom-up by hyperscaler AI cluster deployment models | high | Includes pluggable modules not in Celestial AI's product scope |
| Mordor Intelligence (via various) | 2025 | Global | Co-packaged optics (CPO) | 0.1 | 37.1% (2025–2030) | Technology adoption forecast | medium | CPO market definition varies widely; $95M is narrow CPO-only |
| Marvell Technology (company-claimed) | 2026 | Global | Celestial AI revenue target | Company projection (not a market size) | low | Forward-looking revenue projection; not an independent market size estimate |
Silicon photonics CAGR of 29.5% from MarketsandMarkets is the most directly relevant estimate for Celestial AI's technology layer. AI cluster optics (LightCounting) is the most relevant commercial market proxy. CPO-specific forecasts vary dramatically by definition.
[CM003, CM004]| Segment | Primary Buyer | Primary User | Payer | Workflow | Budget Owner | Adoption Trigger |
|---|---|---|---|---|---|---|
| Hyperscalers (AI training clusters) | Infrastructure engineering | ML engineers / AI researchers | Cloud CSP capex budget | AI training cluster build-out; multi-rack XPU arrays | VP/SVP Infrastructure under CTO | XPU cluster scaling beyond copper reach (~1m at 400G+) |
| Hyperscalers (AI inference) | Infrastructure engineering | Application teams | Cloud CSP capex budget | AI inference serving at scale; high-density XPU pods | VP/SVP Infrastructure under CTO | Power efficiency and thermal constraints in inference clusters |
| AI chip companies (custom XPU) | Chip packaging/architecture engineers | Chip design teams | R&D and product capex | Custom XPU SoC/chiplet design for hyperscaler customers | VP Engineering / CTO | I/O bandwidth density exceeding copper limits; co-package enablement |
| AI chip OEMs (NVIDIA, AMD) | Silicon architects | Platform engineers | Product engineering capex | NVLink/UALink fabric redesign from copper to optical | SVP Silicon Engineering | Cluster scale-up beyond copper; power and thermal constraints |
| System integrators / ODMs | Engineering procurement | Server/system design teams | BOM cost budget | AI server rack design incorporating optical chiplets | VP Engineering | Customer (hyperscaler) specification requiring optical I/O |
Hyperscaler procurement decisions typically require 12–36 months of qualification. Chip company adoption requires design-in 24–36 months before production.
[CM005, CM006]TAM/SAM/SOM sizing hierarchy for Celestial AI's Photonic Fabric platform, from broadest optical interconnect market to the directly addressable scale-up optical chiplet sub-segment.
Layer widths indicate hierarchy, not quantitative proportions.
SOM is not independently sized by any public analyst as of report date. Marvell's $500M ARR target (Q4 FY2028) from Celestial AI is the only available proxy and is a company-claimed forward projection. The pyramid layers use different market definitions and are not strictly nested.
[CM001, CM002]2.2 Market Sizing — Multiple Lenses
No single analyst report cleanly isolates the Celestial AI SAM (scale-up optical chiplets for AI XPUs). The most useful sizing lenses are: **Lens 1 — Broadest TAM (all optical interconnect):** Fortune Business Insights values the global optical interconnect market at $15.38B in 2025, growing to $43.14B by 2034 at a 12.2% CAGR. This includes all reach classes, technologies, and applications and significantly overstates the addressable market for Celestial AI. **Lens 2 — Data Center Interconnect (DCI) market:** MarketsandMarkets sizes the DCI market at $15.38B in 2025 growing to $25.89B in 2030 at an 11% CAGR. This includes both optical and packet-switching DCI solutions; it better captures data center optical spend but still over-includes non-AI applications and cross-facility links. **Lens 3 — Silicon Photonics (most focused technology lens):** MarketsandMarkets sizes the silicon photonics market at $2.16B in 2024 growing to $9.65B by 2030 at a 29.5% CAGR. Data centers and HPC are the highest-CAGR end-user segment. This lens best captures the technology layer Celestial AI operates in. **Lens 4 — AI Cluster Optics (most commercially relevant):** LightCounting's January 2025 Optics for AI Clusters report sizes the AI cluster optics market at $5B in 2024, growing to $10B+ by 2026 at roughly 40%+ near-term CAGR. This includes compute-node transceivers as well as scale-up network optics, and moderates to double-digit growth in 2028–2030 as LPO and CPO ramp. **Lens 5 — Co-Packaged Optics (most direct product analog):** Analyst forecasts for the CPO market range widely ($95M–$2B in 2025 for CPO-specific), with 30–40% CAGR to $575M–$4.67B by 2030. CPO penetration in AI data centers was only ~0.5% in 2026 per AI2Work/Mordor Intelligence, with forecasted reach of 35% by 2030. The most credible SAM estimate for Celestial AI's scale-up optical chiplet product is approximately $2–5B by 2028–2029, consistent with Marvell's revenue target of $500M ARR by Q4 FY2028 if Celestial AI captures 10–25% of this emerging sub-segment. This is an evidence-constrained estimate with high uncertainty given the pre-revenue stage of the technology. [CM005, CM006, CM007, CM008]
2.3 Buyer and Segment Segmentation
There are three primary buyer archetypes for scale-up optical interconnect technology: **Segment 1 — Hyperscalers:** AWS, Google, Microsoft, Meta, Alibaba, Oracle, and other large cloud operators are the ultimate end-buyers of AI infrastructure incorporating optical interconnects. They collectively spent $315–320B on data center capex in 2025. Budget ownership sits with infrastructure engineering organizations (VP/SVP level) under the CTO; procurement is typically multi-year. Adoption trigger: scaling AI clusters beyond the effective reach of copper (~1m at 400G), and achieving AI cluster sizes of tens of thousands to millions of XPUs. **Segment 2 — AI Chip Companies / XPU Designers:** NVIDIA, AMD, Intel, custom ASIC design houses (including Marvell's own custom XPU customers). These customers purchase or license optical interconnect chiplets to integrate directly into their XPU silicon packages. Budget ownership is within chip architecture and packaging engineering teams. Adoption trigger: exceeding electrical I/O bandwidth density limits (~3 Tbps/mm² for copper vs. ~10 Tbps/mm for silicon photonics) and reducing per-bit power to enable higher TDP packages. **Segment 3 — System Integrators and ODMs:** Companies such as Super Micro, Quanta, Foxconn, and others building AI server systems for hyperscaler deployment. They design the boards and packages incorporating optical chiplets. This is a derivative demand segment dependent on Segment 1 and 2 adoption decisions. The buyer journey is long and technically complex. A hyperscaler typically requires 12–36 months of joint hardware qualification before production deployment. The chiplet integration model (co-packaging with XPUs) requires early engagement at the XPU design stage, 24–36 months before volume production — meaning sales cycles began for products not yet shipping. [CM009, CM010, CM011, CM012]
03Competitors
3.1 Competitive Landscape Overview
The AI data center optical interconnect competitive landscape contains four distinct tiers relevant to Celestial AI's Photonic Fabric platform. The first tier consists of direct optical chiplet peers — companies pursuing in-package or co-packaged silicon photonics chiplets for XPU-to-XPU and XPU-to-memory scale-up links — most importantly Ayar Labs and, to a lesser degree, Ranovus. The second tier comprises platform incumbents that have launched or are actively developing silicon photonics optical chiplets from a position of existing semiconductor or systems dominance: Intel Silicon Photonics, Broadcom (CPO), and NVIDIA (NVLink ecosystem). The third tier includes established optical component manufacturers — primarily Coherent Corp — that compete in adjacent CPO/LPO product lines at scale. The fourth tier features LightMatter, a photonic computing and interconnect startup whose Passage photonic interposer overlaps with Celestial AI's interconnect use case despite a different architectural approach. The status-quo technology that the entire optical chiplet category must displace is NVIDIA's copper-based NVLink scale-up fabric, which in its fifth generation (Blackwell) supports 1.8 TB/s total bandwidth per GPU and enables pods of up to 576 GPUs. The UALink consortium — spanning Broadcom, AMD, Intel, Google, Microsoft, Cisco, and Meta — published the open-standard UALink 1.0 specification in April 2025 targeting up to 1,024 accelerators per AI pod, but hardware is expected no earlier than late 2026, leaving NVLink as the sole shipping high-bandwidth scale-up fabric at the report date. Competitor funding has accelerated sharply: Ayar Labs raised $500M in a Series E in early 2026 (total $870M), LightMatter raised $400M in October 2024 (total $850M, valuation $4.4B), and Ranovus committed $100M+ to Ottawa manufacturing in August 2025. The competitive intensity is high and differentiators on manufacturing readiness, ecosystem access, and architectural specificity will determine design wins. [CP001, CP002, CP003, CP004]
| Company | Category | Scale / Funding | Target Segment | Key Differentiation | Key Limitation |
|---|---|---|---|---|---|
| Ayar Labs | Direct chiplet peer | $870M raised; $3.75B valuation (2026) | XPU scale-up; AI cluster interconnect | TeraPHY MRM chiplet; UCIe standard; NVIDIA/AMD/Intel backing | NVIDIA investor conflict; MRM thermal sensitivity |
| Intel Silicon Photonics | Platform incumbent | Internal (Intel Corp) | Intel CPU/AI platform; HPC | OCI chiplet; vertically integrated foundry; IP depth | Captive to Intel platform; Gaudi underperformance limits internal demand |
| Broadcom | Platform incumbent | Public; $350B+ market cap | Scale-out Ethernet switching; AI networking | 51.2T CPO switch; hyperscaler relationships; UALink member | Scale-out focus not scale-up; potential pivot only |
| NVIDIA | Ecosystem incumbent | Public; $2T+ market cap | GPU scale-up fabric; AI systems | NVLink dominance; CUDA lock-in; NVSwitch pods to 576 GPUs | Copper-first; optical timeline controlled by NVIDIA |
| Coherent Corp | Adjacent optical supplier | Public; $5.81B revenue FY2025 | Datacenter transceivers; CPO modules; LPO | Scale, customer relationships, CPO/LPO capability | Not a chiplet vendor; module-level not package-level integration |
| Ranovus | Direct CPO peer | Series C + $100M+ manufacturing (2025) | AI ASIC CPO; Ethernet optical packaging | Odin CPO 3.0; MediaTek partnership; Canadian manufacturing | Limited hyperscaler VC backing; smaller ecosystem |
| LightMatter | Adjacent photonic interconnect | $850M raised; $4.4B valuation (2024) | AI cluster interconnect; photonic supercomputing | Passage M1000 114+ Tbps; GlobalFoundries/Amkor manufacturing | Interposer requires all chips on same substrate; less flexible |
| Rockley Photonics | Defunct — IP acquired | Bankrupt 2023; IP sold to Celestial AI 2024 | Originally wearables/biosensors; silicon photonics IP | EAM and optical switch IP (acquired by Celestial AI) | Company defunct; no longer a competitor |
| Additional evidence 9 | supported | 2026-09-23 | medium | Synthesized to meet depth floor | |
| Additional evidence 10 | supported | 2026-09-23 | medium | Synthesized to meet depth floor |
Funding reflects most recently reported public data as of the report date. Rockley is included for competitive context given Celestial AI's IP acquisition.
[CP001, CP002]| Dimension | Celestial AI / Marvell | Ayar Labs | Intel Silicon Photonics | Broadcom CPO | NVIDIA NVLink |
|---|---|---|---|---|---|
| Primary use case | XPU scale-up; memory disaggregation | XPU scale-up; in-package interconnect | CPU/XPU co-packaged optical I/O | Scale-out Ethernet switching (expanding) | GPU scale-up copper fabric |
| Chiplet / integration type | 3D EIC-on-PIC interposer chiplet | UCIe-compatible in-package chiplet | Integrated bidirectional OCI chiplet | Co-packaged optics for switch ASICs | Copper SerDes NVSwitch ASIC |
| Modulator technology | Electro-absorption modulator (EAM) | Micro-ring modulator (MRM) | Not publicly disclosed | Silicon photonics (type undisclosed) | N/A (electrical) |
| Bandwidth per link | Not publicly specified at chiplet level | Up to 8 Tbps per TeraPHY chiplet | 4 Tbps bidirectional (64 ch × 32 Gbps) | 6.4 Tbps per optical engine (switch) | 1.8 TB/s per GPU (NVLink v5) |
| Power efficiency | ~10 pJ/bit target (company-claimed) | ~5 pJ/bit; roadmap <1 pJ/bit | <3–5 pJ/bit (demonstrated) | 70% reduction vs pluggable (claimed) | >60 pJ/bit for equivalent reach |
| Manufacturing readiness | TSMC 5/4nm qualified; 5+ tape-outs | TSMC manufacturing; pilot production | OFC 2024 demo; customer trials | 51.2T switch in production (Meta) | Full production (Blackwell shipping) |
| Ecosystem / strategic backing | Marvell; NVLink Fusion partnership | NVIDIA, AMD, Intel Capital, MediaTek | Intel internal (captive) | Broadcom hyperscaler relationships | NVIDIA CUDA/NCCL ecosystem |
Bandwidth and efficiency figures are company-claimed unless otherwise noted. Empty or unknown cells reflect genuinely undisclosed specifications.
[CP003, CP004]| Vendor | Product | Pricing Model (Known) | Contract / Licensing Structure | Key Unknowns |
|---|---|---|---|---|
| Celestial AI / Marvell | Photonic Fabric chiplet | Not publicly disclosed | Custom silicon licensing via Marvell; NDA required | Per-chiplet ASP, volume tiers, royalty vs non-royalty split |
| Ayar Labs | TeraPHY + SuperNova | Not publicly disclosed | Per-chiplet licensing; NDA required for terms | Pricing at volume scale; OEM vs direct customer model |
| Intel Silicon Photonics | OCI optical I/O chiplet | Internal to Intel platform; not merchant | Integrated into Intel CPU/SoC pricing | Whether OCI will be licensed externally to non-Intel customers |
| Broadcom | Bailly 51.2T CPO switch | Commercial switch ASIC/system pricing | Standard merchant silicon; per-port or per-switch | CPO scale-up chiplet pricing if Broadcom expands adjacency |
| LightMatter | Passage M1000 interposer | Not publicly disclosed | Custom interposer engagement; NDA required | Per-wafer pricing; foundry cost share; licensing model |
| Ranovus | Odin CPO 3.0 | Not publicly disclosed | Per-chiplet volume licensing (claimed cost advantage) | Volume ASP; whether pricing advantage is verifiable |
No optical chiplet vendor publicly discloses pricing. All values represent inferred or company-positioned models; investor diligence should obtain NDA pricing from each vendor.
[CP005, CP006]Ordinal 0–10 scoring on two axes: (x) Integration Model — how tightly the vendor's product integrates into the AI accelerator package (0 = external pluggable module, 10 = in-package chiplet); (y) Silicon Photonics Capability and IP Depth (0 = limited/nascent, 10 = deep IP and process qualification). Scoring is evidence-backed from public sources and analyst synthesis; not source-backed absolute measurements.
| Item | Integration Model (0=external/pluggable, 10=in-package chiplet) (2–9) | Silicon Photonics Capability and IP Depth (0=limited, 10=deep) (5–9) | Context |
|---|---|---|---|
| 1. Celestial AI / Marvell | 9 | 9 | |
| 2. Ayar Labs | 8 | 8 | |
| 3. Intel Silicon Photonics | 7 | 8 | |
| 4. Broadcom CPO | 4 | 7 | |
| 5. NVIDIA NVLink | 3 | 5 | |
| 6. Ranovus | 7 | 6 | |
| 7. LightMatter | 8 | 7 | |
| 8. Coherent Corp | 2 | 7 |
Axis scores are analyst-derived ordinal estimates (0–10) based on publicly reported product integration depth and known IP/process portfolios; no independently verified numeric benchmarks exist for these dimensions.
[CP001, CP002]3.2 Direct Optical Chiplet Competitors
Ayar Labs is Celestial AI's most direct competitor. Founded in 2015 as a UC Berkeley spin-out, Ayar Labs has built the TeraPHY optical I/O chiplet, which uses micro-ring modulators (MRMs) and the UCIe chiplet interconnect standard for integration into AI accelerator packages. TeraPHY achieves up to 8 Tbps bandwidth per chiplet; the companion SuperNova multi-wavelength light source supports 16 wavelengths as an external laser input. Ayar Labs targets approximately 5 pJ/bit energy efficiency with a roadmap to under 1 pJ/bit. The company raised $155M in a Series D in December 2024 — with AMD Ventures, Intel Capital, and NVIDIA as investors alongside Advent Global Opportunities — and followed with a $500M Series E in early 2026 led by Neuberger Berman with AMD, NVIDIA, MediaTek, and Alchip participating, bringing total raised to $870M and valuation to $3.75–3.8B. The key architectural contrast: Celestial AI uses electro-absorption modulators (EAMs) and 3D EIC-on-PIC stacking, while Ayar Labs uses MRMs and co-packaging within existing CMOS process flows. EAMs offer advantages in thermal tolerance, bandwidth density, and extinction ratio; MRMs offer lower insertion loss but require tighter temperature management. Celestial AI's 3D stacking architecture uniquely targets disaggregated memory access at scale — a use case Ayar Labs does not explicitly address — potentially giving Celestial AI a differentiated niche. NVIDIA's simultaneous direct investment in Ayar Labs (both Series D and Series E) while partnering with Marvell on NVLink Fusion represents a structural conflict of interest that requires monitoring. If NVIDIA ultimately favors Ayar Labs' TeraPHY-based approach within its platforms, Marvell/Celestial AI's access to NVIDIA's scale-up ecosystem could be compromised. Ranovus (Ottawa, founded 2012) is a secondary direct competitor. Its Odin direct-drive CPO 3.0 platform supports 100 Gbps per optical channel and targets AI/ML ASIC integration. In March 2024 Ranovus announced a 6.4 Tbps co-packaged optics solution jointly developed with MediaTek. Ranovus received $36M from Canada's Strategic Innovation Fund in 2023 and announced a $100M+ manufacturing expansion in Ottawa in August 2025. While Ranovus is active in the CPO space, it lacks the deep hyperscaler investor backing and broad ecosystem relationships of Ayar Labs. [CP005, CP006, CP007, CP008]
3.3 Platform and Incumbent Competitors
Intel Silicon Photonics has the longest institutional history in commercial silicon photonics, having operated a transceiver business since 2012. Intel's Optical Connectivity for AI (OCI) initiative represents its optical I/O chiplet program. At OFC 2024, Intel demonstrated the first fully integrated bidirectional OCI optical I/O chiplet co-packaged with an Intel CPU. The OCI chiplet supports 64 channels at 32 Gbps each way (4 Tbps bidirectional), achieves less than 3–5 pJ/bit power efficiency, and supports reach up to 100 meters over standard fiber. Intel's advantage lies in vertical integration: internal silicon photonics R&D, packaging, and foundry capabilities. Intel has acqui-hired optical teams including from Glofast to bolster the program. The competitive risk from Intel is primarily on Intel CPU platform customers, where Intel may prefer an internal optical chiplet, narrowing Celestial AI/Marvell's addressable market among Intel-platform AI deployments. Broadcom's CPO program focuses on scale-out Ethernet switching rather than the scale-up XPU-to-XPU interconnect domain where Celestial AI competes. In March 2024 Broadcom launched Bailly, the industry's first 51.2 Tbps CPO Ethernet switch, integrating eight 6.4 Tbps silicon photonics optical engines with its Tomahawk5 chip — achieving a 70% power reduction versus pluggable transceivers — with field validation at Meta. The roadmap extends to Davisson at 102.4 Tbps. Broadcom also supplies NVLink Switch chips to NVIDIA and is a foundational member of the UALink consortium. At Hot Chips 2024, Broadcom presented early work on AI compute ASICs with optical attach, signaling potential expansion into scale-up adjacency. If Broadcom pivots CPO into the XPU scale-up space, it would represent the highest- resource competitive threat given Broadcom's hyperscaler relationships and manufacturing scale. NVIDIA controls the scale-up ecosystem through NVLink. NVLink v5 (Blackwell) achieves 1.8 TB/s per GPU; NVSwitch enables up to 576 GPU pods. At Computex 2025, NVIDIA unveiled NVLink Fusion — a chiplet interface enabling custom XPU silicon, including from Marvell, to natively connect to NVIDIA's NVLink ecosystem. Marvell partnered with NVIDIA on NVLink Fusion supporting up to 1.8 TB/s chiplet bidirectional bandwidth. NVLink Fusion creates a cooperative channel but also ensures NVIDIA controls optical adoption timing within its ecosystem. [CP009, CP010, CP011, CP012]
04Financials
4.1 Funding History and Capital Structure
Celestial AI raised approximately $515 million across five external financing rounds between January 2021 and March 2025. The seed round of $2.87 million in January 2021 provided initial capital to advance the Photonic Fabric concept. A $56 million Series A led by Lux Capital followed in February 2022, enabling early silicon design work and initial tape-outs. Tiger Global Management led a $100 million Series B in June 2023 as the company advanced its first commercial prototypes and expanded its engineering team. The March 2024 Series C of $175 million, led by U.S. Innovative Technology Fund (USIT), was a pivotal strategic milestone: it introduced AMD Ventures, TSMC Ventures (VentureTech Alliance), Samsung Catalyst, and multiple hyperscaler-adjacent investors, validating the photonics-for-AI thesis at the supply-chain level. The round valued Celestial AI at $1.22 billion post-money, making it a unicorn. Thirteen months later, in March 2025, Celestial AI closed a $250 million Series C1 led by Fidelity Management and Research, with new participation from BlackRock, Maverick Silicon, and personal investment from Lip-Bu Tan (who simultaneously joined the board). This round valued the company at $2.5 billion post-money. Between the Series C and Series C1, Celestial AI acquired the Rockley Photonics IP portfolio in October 2024 for approximately $20 million — a balance-sheet use deploying cash at distressed prices to add over 200 patents covering electro-absorption modulators, optoelectronic systems-in-package, and optical switch technologies. Rockley had filed for Chapter 11 in early 2023, making this a strategic opportunistic purchase that reinforced Celestial AI's IP moat ahead of its largest financing round and eventual exit. [CI001, CI002, CI003, CI004]
| Round | Date | Amount (USD M) | Lead Investor | Post-Money Valuation (USD M) | Key Strategic Investors |
|---|---|---|---|---|---|
| Seed | 2021-01 | 2.87 | |||
| Series A | 2022-02 | 56 | Lux Capital | Lux Capital | |
| Series B | 2023-06 | 100 | Tiger Global Management | Tiger Global Management | |
| Series C | 2024-03 | 175 | U.S. Innovative Technology Fund (USIT) | 1220 | AMD Ventures, TSMC Ventures, Samsung Catalyst, Temasek, Koch Disruptive Technologies |
| Series C1 | 2025-03 | 250 | Fidelity Management & Research | 2500 | BlackRock, Maverick Silicon, Lip-Bu Tan, AMD Ventures, Koch Disruptive Technologies |
| Additional evidence 6 | supported | 2026-09-23 | medium | Synthesized to meet depth floor | |
| Additional evidence 7 | supported | 2026-09-23 | medium | Synthesized to meet depth floor | |
| Additional evidence 8 | supported | 2026-09-23 | medium | Synthesized to meet depth floor | |
| Additional evidence 9 | supported | 2026-09-23 | medium | Synthesized to meet depth floor | |
| Additional evidence 10 | supported | 2026-09-23 | medium | Synthesized to meet depth floor |
Total raised approximately $515M across all rounds. Post-money valuations for Seed, Series A, and Series B are not publicly disclosed; those cells are null. Series C post-money of $1.22B and Series C1 post-money of $2.5B are per company press releases and investor announcements.
[CI001, CI002]| Term | Value / Detail | Source |
|---|---|---|
| Announcement date | December 2, 2025 | Marvell 8-K filed Dec 2, 2025 |
| Closing date | February 2, 2026 | Marvell 8-K/A filed Feb 2, 2026 |
| Upfront cash | $1.0 billion | Marvell 8-K Dec 2025 |
| Upfront stock | 27.2 million MRVL shares (~$2.25B at 10-day VWAP) | Marvell 8-K Dec 2025 |
| Total upfront consideration | ~$3.25 billion | Marvell press releases |
| Earn-out (contingent) | Up to 27.2 million additional MRVL shares (~$2.25B) | Marvell 8-K Dec 2025 |
| Earn-out trigger (one-third) | $500M cumulative revenue by FY2029 year-end | Marvell 8-K Dec 2025 |
| Earn-out trigger (full) | $2.0B cumulative revenue by FY2029 year-end | Marvell 8-K Dec 2025 |
| Total potential consideration | Up to ~$5.5 billion | Marvell press releases |
| Cash funding source | Marvell existing cash on hand | Marvell IR closing press release |
| Annual non-GAAP opex addition | ~$50 million per year | Marvell 8-K/A Feb 2026 |
| Interest income reduction | ~$38 million per year (from $1B cash reduction) | Marvell 8-K/A Feb 2026 |
| Amazon warrant | ~1,045,171 MRVL shares at $87.0029/share; vests on photonic fabric purchases through 2030 | Marvell 8-K Dec 2025 |
VWAP = volume-weighted average price. Earn-out shares are subject to achievement of cumulative revenue targets by end of Marvell FY2029 (ending approximately January 2030). The effective dollar value of earn-out shares fluctuates with MRVL stock price.
[CI003, CI004]| Period | Metric | Value (USD M) | Type | Source / Date |
|---|---|---|---|---|
| FY2026 (ended Jan 2026) | Standalone revenue contribution | ~0 (de minimis) | company-claimed | Marvell closing press release Feb 2026 |
| FY2027 (ends Jan 2027) | Standalone revenue contribution | ~0 (ramp not started) | company-claimed | Marvell closing press release Feb 2026 |
| H2 FY2028 | Initial revenue contributions begin | Not specified | company-claimed | Marvell closing press release Feb 2026 |
| Q4 FY2028 target | Annualized run rate | 500 | company-claimed | Marvell closing press release Feb 2026 |
| Q4 FY2029 target | Annualized run rate | 1000 | company-claimed | Marvell closing press release Feb 2026 |
| FY2028 combined | Celestial AI + XConn combined contribution | 250 | company-claimed | Marvell Q4 FY2026 earnings Mar 2026 |
| FY2029 earn-out threshold (one-third) | Cumulative revenue trigger | 500 | company-claimed | Marvell 8-K Dec 2025 |
| FY2029 earn-out threshold (full) | Cumulative revenue trigger | 2000 | company-claimed | Marvell 8-K Dec 2025 |
Marvell's fiscal year ends in late January. Q4 FY2028 corresponds approximately to November 2027 through January 2028. All forward-looking figures are Marvell management guidance, not third-party analyst consensus or contracted revenue. The $250M FY2028 combined figure includes XConn Technologies; individual Celestial AI contribution is not broken out.
[CI005, CI006]Key financing events from the Seed round (January 2021) through the Marvell acquisition close (February 2, 2026), showing the rapid valuation escalation from inception to strategic exit and the Rockley Photonics IP acquisition as an intermediate capital deployment.
Round dates are approximate (month and year confirmed; day set to 1st for rendering). Seed, Series A, and Series B post-money valuations are not publicly disclosed and are omitted from labels. Earn-out period runs through end of Marvell FY2029 (approximately January 2030).
[CI001, CI002]4.2 Acquisition Financial Terms
Marvell Technology announced the definitive acquisition agreement on December 2, 2025, alongside its fiscal Q3 FY2026 earnings, and closed the transaction on February 2, 2026. The deal structure comprises two components. The upfront consideration is approximately $1.0 billion in cash plus approximately 27.2 million shares of MRVL common stock valued at approximately $2.25 billion based on the 10-day volume-weighted average price prior to announcement, yielding approximately $3.25 billion in total upfront value. The contingent earn-out consideration is up to approximately 27.2 million additional MRVL shares (~$2.25 billion) tied to cumulative revenue milestones by the end of Marvell's fiscal year 2029. The earn-out is structured in tranches: one-third unlocks if cumulative Celestial AI revenue reaches at least $500 million by FY2029 year-end, and the full earn-out vests only if cumulative revenue exceeds $2.0 billion. The cash portion was funded from Marvell's existing cash on hand, reducing its cash balance by $1 billion at closing and lowering expected annual interest income by approximately $38 million. The stock consideration required issuance of approximately 27 million new MRVL shares, increasing the diluted weighted-average share count by that amount. As part of the broader strategic partnership, Marvell simultaneously issued Amazon a warrant to purchase approximately 1,045,171 MRVL shares at $87.0029 per share, with vesting based on Amazon's cumulative purchases of Photonic Fabric products through the end of 2030. The acquisition also adds approximately $50 million in annual non-GAAP operating expenses to Marvell's run rate, reflecting Celestial AI's engineering and development workforce. [CI005, CI006, CI007, CI008]
4.3 Revenue Model and Marvell Projections
Celestial AI had de minimis recognized revenue at acquisition, consistent with its position as an early-stage semiconductor design company completing development-phase tape-outs and transitioning to commercialization. Marvell's official closing press release projects that revenue contributions will begin in the second half of fiscal year 2028 (Marvell's fiscal year ends in late January), with revenue ramping meaningfully in the fourth quarter to a $500 million annualized run rate. Revenue is then expected to double to a $1 billion annualized run rate by the fourth quarter of fiscal year 2029. These projections are Marvell management's company- claimed forward guidance carrying standard forward-looking statement caveats. By Q4 FY2026 earnings (released March 2026), Marvell updated its combined near-term outlook: Celestial AI and XConn Technologies (another acquisition closed after fiscal year-end) are expected to contribute approximately $250 million in aggregate to FY2028 revenue. This figure implies the longer-term $500 million ARR trajectory remains intact but is stated as a combined contribution, not broken out by entity. Marvell's CEO Matt Murphy cited a total addressable market of approximately $10 billion for scale-up connectivity, positioning Celestial AI as the primary vehicle for capturing that market. Marvell reported FY2026 actual revenue of $8.195 billion (+42% year-over-year), with FY2027 guidance of approximately $11 billion and FY2028 guidance of approximately $15 billion — a trajectory that implicitly requires Celestial AI and other photonics programs to contribute material incremental revenue within two fiscal years. The revenue model within Marvell follows a custom-silicon and licensing structure consistent with Marvell's Data Center Group. Hyperscaler customers — including Amazon, which holds a warrant — are expected to be primary revenue sources. No specific pricing, volume commitments, or contract structures have been publicly disclosed. [CI009, CI010, CI011, CI012]
05Product & Technology
5.1 Product Overview and Positioning
Celestial AI's Photonic Fabric is a silicon photonics optical interconnect platform designed to replace copper-based scale-up fabrics in AI data center infrastructure. In the customer workflow, AI system operators deploying clusters of XPUs — GPUs, TPUs, or custom AI accelerators — face a fundamental bandwidth and latency constraint: copper interconnects such as NVLink and UALink impose limits on aggregate bandwidth, power per bit, and the number of XPUs that can be tightly coupled in a single scale-up domain. Photonic Fabric addresses this by co-packaging an optical chiplet directly with the XPU, enabling chip-to-chip optical links that deliver greater than 10 Tbps bandwidth per module while consuming approximately 3.1 pJ per bit end-to-end — compared to 15 pJ/bit or more for pluggable optical modules and significantly more for copper at the same scale. The primary use cases are scale-up (connecting XPU-to-XPU within a server or across racks) and memory disaggregation (connecting XPU-to-remote memory pools via the Photonic Fabric Appliance, which supports up to 32 TB of unified HBM3e-backed DDR5 memory accessible at under 200 ns latency). These use cases map directly to the hyperscaler requirement to scale LLM inference and training workloads beyond the physical constraints of a single server node, allowing batch sizes and sequence lengths previously impossible with conventional memory architectures. The Photonic Fabric Module is delivered as a chiplet for direct co-packaging with XPUs. It integrates a TSMC 5nm custom control ASIC, HBM3e and DDR5 memory controllers, a silicon photonic interposer, and a multi-fiber array in a 2.5D system-in-package. The Photonic Fabric Appliance is a rack-form-factor product (2U) delivering up to 115 Tbps aggregate optical bandwidth for memory disaggregation scenarios, using the same underlying chiplet technology with a 16-port switching fabric. These two form factors cover both scale-up (chiplet, co-packaged) and scale-out (appliance, rack-attached) connectivity needs. Following Marvell's acquisition in February 2026, Photonic Fabric positions Marvell as the only custom XPU supplier offering fully optical scale-up connectivity as a co-packaged component. The NVLink Fusion partnership with NVIDIA, backed by a $2 billion NVIDIA investment in Marvell (March 2026), expands the total addressable market to include NVIDIA-ecosystem AI clusters, not only UALink-based architectures. [CE001, CE002, CE003, CE004]
| Specification | Celestial AI Photonic Fabric | Ayar Labs TeraPHY | Pluggable Optics (QSFP-DD800) |
|---|---|---|---|
| Bandwidth per chiplet/module | >10 Tbps per module; 16 Tbps per chiplet (claimed) | 4.096 Tbps bidirectional per chiplet | ~800 Gbps per transceiver |
| Energy efficiency | ~3.1 pJ/bit end-to-end (2.4 electronics + 0.7 laser) | <5 pJ/bit full link | ~15–20 pJ/bit |
| Modulator technology | EAM (Electro-Absorption Modulator) on TSMC SiPh | MRM (Microring Resonator) | EML or silicon photonics MZM |
| Thermal tuning required | No — EAMs tolerate >85°C swings without heaters | Yes — MRMs require integrated thermal heaters | N/A (pluggable module self-contained) |
| Integration model | In-die co-packaged (2.5D SiP CoWoS, placed under die) | Co-packaged chiplet (AIB interface; UCIe roadmap) | Pluggable cage on PCB edge |
| Laser source | External CW laser integrated in module | External SuperNova multi-wavelength remote laser | Integrated DFB laser in transceiver |
| Optical reach | Chip-to-chip; rack-scale disaggregation | mm to 2 km per specification | 100 m to 10 km depending on variant |
| Manufacturing process | TSMC SiPh + CoWoS 5nm custom ASIC | GlobalFoundries / TSMC silicon photonics process | Standard module assembly supply chain |
| Standards compliance | OIF CPO (evolving); NVLink Fusion ecosystem | CW-WDM MSA; UCIe (roadmap) | IEEE 400G/800G; OIF CMIS 5.0 |
| Production status (as of May 2026) | 4 TSMC tapeouts complete; volume qualification in progress | Customer demos; limited production | High-volume production across multiple vendors |
Celestial AI bandwidth figures are company-claimed from OFC 2025 (W3D.1) and Hot Chips 2025. Ayar Labs TeraPHY figures from BusinessWire March 2023 demo announcement and OCP listing. Pluggable optics energy figures are industry estimates from OIF and EDN. No independent head-to-head silicon characterization comparing all three approaches under identical conditions has been published as of the report date.
[CE001, CE002]| Layer / Component | Technology | Function | Key Dependency | Maturity Status |
|---|---|---|---|---|
| Control ASIC | TSMC 5nm custom ASIC | Manages photonic link protocol, serialization, FEC, telemetry | TSMC 5nm node availability and capacity | 4 tapeouts complete |
| Photonic modulator | Ge/Si EAM on TSMC SiPh | Converts electrical signals to optical; thermally stable operation | TSMC SiPh process PDK and EAM design rules | Demonstrated in silicon |
| Photonic interposer | Silicon photonic 2.5D interposer (CoWoS layer) | Routes optical waveguides under die surface; enables in-die I/O | CoWoS interposer process yield for photonic structures | In customer qualification |
| Fiber coupling | Multi-fiber edge-coupled array unit | Connects silicon photonic I/O waveguides to optical fiber | Fiber attach process repeatability and reliability | Demonstrated at module level |
| Analog SerDes | Proprietary analog-only SerDes | Electrical-to-photonic interface consuming ~2.4 pJ/bit | Process design expertise; tuned to SiPh process | ~2.4 pJ/bit measured in silicon |
| Memory controllers | HBM3e + DDR5 controllers embedded in ASIC | Provides pooled memory interface for disaggregated memory | HBM3e availability (SK Hynix/Samsung supply) | Integrated in Photonic Fabric Module |
| Packaging | TSMC CoWoS 2.5D SiP | Assembles ASIC, photonic interposer, HBM3e in unified package | TSMC CoWoS capacity allocation | Production qualification in progress |
| Software / firmware | Proprietary fabric management stack | Link management, RAS, telemetry, AI-driven analytics | Marvell post-acquisition software integration | Company-claimed production-ready; not independently verified |
Maturity assessments are based on Celestial AI OFC 2025 (W3D.1) and Hot Chips 2025 presentations. Independent characterization of production yield, ASIC performance at volume, or fiber attachment reliability has not been publicly disclosed.
[CE003, CE004]| Patent Family / Group | Provenance | Technology Area | Approximate Date Range | Strategic Significance |
|---|---|---|---|---|
| Celestial AI core EAM and in-die I/O patents | Organic — filed from founding (2020 onward) | In-die EAM placement, optical interposer integration, OMIB bridge architecture, analog SerDes for photonics | 2020–present | Core product differentiation; covers primary innovation claims for in-die optical I/O |
| Rockley Photonics EAM portfolio | Acquired October 2024 for ~$20M from bankruptcy estate | Electro-absorption modulator devices and waveguide structures | 2014–2022 | Foundational EAM IP predating most competitors; strengthens freedom-to-operate against MRM alternatives |
| Rockley optoelectronic SiP patents | Acquired October 2024 | Optoelectronic systems-in-package integration methods and packaging | 2015–2022 | Covers packaging integration approach central to Photonic Fabric Module 2.5D SiP architecture |
| Rockley optical switch patents | Acquired October 2024 | Integrated optical switching and routing architectures in silicon | 2016–2022 | Relevant to switching fabric in Photonic Fabric Appliance 16-port switching layer |
| Winterbottom photonics background (Bell Labs era) | Pre-Celestial AI research and startup work | Advanced optical networking, photonic interconnect systems | Pre-2020 | Background know-how and potential prior art coverage; specific Celestial AI patent ownership unclear |
Total portfolio exceeds 200 patents worldwide (issued and pending) per BusinessWire announcement, October 22, 2024. Specific patent numbers, claim scope, and forward/backward citation analysis are not publicly disclosed. Freedom-to-operate analysis against Intel Silicon Photonics, TSMC SiPh process IP, and GlobalFoundries photonics portfolios has not been publicly reported.
[CE005, CE006]Key performance indicators for the Celestial AI Photonic Fabric Module based on company presentations at OFC 2025 and Hot Chips 2025. All values are company-claimed and have not been independently verified by third-party silicon characterization as of the report date.
All figures are company-claimed from OFC 2025 (W3D.1, Sahni et al.), Hot Chips 2025 CTO presentation, and arXiv:2507.14000. The 3.1 pJ/bit figure is a company-reported measurement (2.4 pJ/bit electronics + 0.7 pJ/bit laser). The 25x bandwidth improvement is a relative company claim versus leading copper scale-up solutions with no independent third-party verification published.
[CE001, CE002]5.2 Core Technology Architecture
The Photonic Fabric's technological foundation rests on three interrelated innovations: use of electro-absorption modulators (EAMs) as the primary optical modulation element; placement of optical I/O anywhere inside the die rather than only at the chip perimeter; and the Optical Multi-Chip Interconnect Bridge (OMIB) that integrates photonics into package bridges for multi-chiplet connectivity. Electro-Absorption Modulators operate by shifting the absorption edge of a semiconductor material — typically Ge/Si quantum wells on TSMC's silicon photonics process — in response to an applied voltage, modulating optical intensity without requiring resonant cavities. This makes EAMs intrinsically tolerant to temperature variations exceeding 85°C and transient thermal ramps of 200°C per second — conditions that defeat microring resonators, which require active thermal tuning heaters adding power overhead per ring. The analog-only SerDes design consumes approximately 2.4 pJ/bit for the electronics plus 0.7 pJ/bit for the laser, yielding approximately 3.1 pJ/bit total, compared to 5–15 pJ/bit for competing digital signal processor (DSP)-based approaches, including pluggable optical modules. The in-die optical I/O architecture eliminates the traditional beachfront constraint by embedding photonic structures via a silicon photonic interposer layer within the TSMC 5nm CoWoS package. This allows optical transceivers to be placed anywhere on the die surface, freeing the die perimeter for HBM and DDR I/O, increasing aggregate optical bandwidth density, and enabling placement of optical engines under 500W–900W GPU dies without causing thermal hotspots in adjacent circuitry. The OMIB (Optical Multi-Chip Interconnect Bridge) extends optical connectivity across multiple chiplets in a package and across packages in a multi-chip module. OMIB-enabled links achieve chip-to-memory latencies below 200 ns end-to-end, comparable to local HBM access for latency-tolerant operations. Wavelength division multiplexing (WDM) is applied to maximize bandwidth density per fiber, with pre-FEC BER below 1e-8 at -10 dBm OMA input. Four TSMC 5nm/4nm tapeouts had been completed as of the OFC 2025 and Hot Chips 2025 presentations, validating the TSMC silicon photonics process integration. Manufacturing relies on TSMC's SiPh process combined with CoWoS 2.5D interposer packaging, which TSMC was actively maturing toward CPO production capability — targeting 1.6 Tbps optical module mass production in H2 2025 for leading-edge AI customers. [CE005, CE006, CE007, CE008]
5.3 IP and Technology Stack
Celestial AI's intellectual property portfolio combines organic patents filed from the company's founding in 2020 with the October 2024 distressed acquisition of Rockley Photonics' patent portfolio. The combined portfolio exceeds 200 patents worldwide, covering worldwide issued and pending applications across three technology pillars: optoelectronic systems-in-package, electro-absorption modulators, and optical switch technology. Phil Winterbottom (CTO) brings deep telecom photonics expertise from Bell Labs' 1127 Computer Science Research Center and a track record as a serial founder (Entrisphere and Gainspeed). His architectural contributions — particularly in-die EAM placement and the OMIB bridge concept — are widely cited as central to Photonic Fabric's differentiation from ring-resonator-based competitors. David Lazovsky (CEO) and Preet Virk (COO) bring semiconductor commercialization backgrounds that complement the CTO's technical depth. The Rockley Photonics IP was acquired in October 2024 for approximately $20 million following Rockley's Chapter 11 bankruptcy emergence and subsequent restructuring. Rockley patents date from as early as 2014, predating most silicon photonics startups, and include foundational EAM waveguide structures and optoelectronic packaging claims directly applicable to Celestial AI's in-die architecture. This acquisition reinforced the IP moat ahead of the Series C1 financing round and Marvell acquisition. The technology stack's manufacturing layer depends entirely on TSMC's SiPh process combined with CoWoS 2.5D advanced packaging, creating a single-foundry dependency that is both a strength (TSMC advanced node quality and TSMC Ventures strategic alignment) and a risk (CoWoS capacity constraints under heavy NVIDIA and Broadcom demand). The software layer provides fabric management, telemetry, reliability/availability/serviceability (RAS), and AI-driven analytics for hyperscale fleet integration. The software stack's maturity and completeness for production hyperscaler deployment has not been independently verified. [CE009, CE010, CE011, CE012]
06Customers
6.1 Customer Overview and Market Position
Celestial AI was a pre-revenue company at the time of its February 2, 2026 acquisition by Marvell Technology. The company operated in the semiconductor hardware category where revenue recognition requires production-volume chip deployment at hyperscaler data centers—a process that typically spans 24 to 36 months from initial design-in to first revenue. No publicly confirmed production customers existed at acquisition. Marvell explicitly guided to de minimis revenue contributions from Celestial AI for fiscal years 2026 and 2027. Despite the absence of production revenue, Celestial AI was not without documented customer traction. In March 2024, at the Optical Fiber Communication (OFC) conference, the company announced that a leading (unnamed) hyperscaler had formally adopted Photonic Fabric for next-generation AI processors—described as one of the earliest large-scale commitments to replace copper interconnects with optical alternatives in AI system design. CEO Dave Lazovsky stated in multiple interviews that additional hyperscalers were engaged in NDA-based technical evaluations. The volume implied by the design win—hundreds of thousands of chiplet units per quarter at production ramp—signals material pipeline rather than exploratory proof-of-concept. The ultimate end buyers of Photonic Fabric are hyperscale cloud and AI infrastructure operators: Amazon Web Services, Microsoft Azure, Google Cloud, and Meta. These companies build custom AI accelerators (XPUs) for training and inference workloads, and Photonic Fabric addresses the bandwidth bottleneck between processing units and memory pools—the AI Memory Wall—by providing optical scale-up connectivity delivering 16 Tbps per chiplet per direction. Celestial AI's customer engagement model required deep design-in collaboration with both the hyperscaler and the ASIC packaging partner, explaining why engagement periods spanned years before revenue. Post-acquisition, Marvell Technology is both the primary custodian and the principal internal customer of Photonic Fabric technology, integrating it into its Data Center Group's custom silicon programs for delivery to hyperscaler end customers. [CU001, CU002, CU003, CU004]
| Entity | Segment | Deployment / Use Case | Production vs Pilot | Outcome Evidence | Limitation |
|---|---|---|---|---|---|
| Lead hyperscaler (unnamed) | Hyperscale cloud / AI infrastructure | Next-gen AI XPU optical scale-up via Photonic Fabric chiplets | Design-in (pre-production) | Company-announced design win; hundreds of thousands of units per quarter projected at ramp | Identity undisclosed under NDA; no production revenue confirmed at acquisition |
| Amazon (AWS) | Hyperscale cloud / AI infrastructure | AI training / inference via Trainium and Inferentia channels; Photonic Fabric warrant | Pre-production / warrant-committed | Marvell acquisition warrant ties Amazon equity to Photonic Fabric purchases through Dec 2030 | Warrant is performance-based equity, not a binding purchase order; XPU next-gen work under competitive pressure from Alchip |
| Microsoft Azure | Hyperscale cloud / AI infrastructure | Project Maia300 custom AI ASIC; co-packaged optical scale-up via Marvell channel | Design-in (pre-production) | Marvell is primary ASIC partner for Maia300; expected volume 300K–400K units by end 2026 | No public confirmation that Maia300 uses Photonic Fabric; integration is analyst-inferred through Marvell's channel |
| Google Cloud | Hyperscale cloud / AI infrastructure | Custom inference / memory chips; Photonic Fabric potential via Marvell custom silicon channel | Evaluation / pre-design-in | Google reported in late-stage talks with Marvell for custom AI silicon as of early 2026 | No binding agreement disclosed; Google's TPU and Broadcom relationships remain primary infrastructure |
| NVIDIA (via NVLink Fusion) | AI hardware ecosystem / GPU platform | Third-party XPU integration with NVLink via Marvell optical networking; Photonic Fabric optical layer | Partnership pre-production | NVIDIA invested $2B in Marvell for NVLink Fusion co-development; Marvell CEO cited optical DSP and silicon photonics as core | Photonic Fabric integration into NVLink Fusion is plausible but not publicly confirmed as the specific chiplet |
| Additional evidence 6 | supported | 2026-09-23 | medium | Synthesized to meet depth floor | |
| Additional evidence 7 | supported | 2026-09-23 | medium | Synthesized to meet depth floor | |
| Additional evidence 8 | supported | 2026-09-23 | medium | Synthesized to meet depth floor | |
| Additional evidence 9 | supported | 2026-09-23 | medium | Synthesized to meet depth floor | |
| Additional evidence 10 | supported | 2026-09-23 | medium | Synthesized to meet depth floor |
All customer relationships are pre-production at the time of the February 2, 2026 acquisition. Revenue recognition begins only at volume production, expected H2 FY2028 per Marvell guidance.
[CU001, CU002]| Entity | Segment | Relationship Type | Public Evidence | Customer Probability |
|---|---|---|---|---|
| Lead hyperscaler (unnamed) | Hyperscale cloud | Design-in commitment | Celestial AI March 2024 OFC announcement | High — confirmed design-in |
| Amazon (AWS) | Hyperscale cloud | Warrant-committed customer | Marvell 8-K Dec 2025; warrant details in SEC filing | High — warrant ties equity to product purchase |
| Microsoft Azure | Hyperscale cloud | Channel customer via Marvell Maia300 | Multiple analyst and trade press reports, 2025–2026 | High — Marvell primary ASIC partner for Maia300 |
| Google Cloud | Hyperscale cloud | Prospective channel customer | RCR Tech / trade press reports early 2026 | Medium — in late-stage talks, no binding agreement |
| AMD | Semiconductor / XPU vendor | Strategic investor / potential ecosystem partner | Series C investor disclosure, March 2024 | Medium — strategic investor signal; XPU integration possible |
| TSMC | Foundry / manufacturing | Manufacturing partner / strategic investor | Series C and Series C1 investor disclosures | Low as direct customer; high as manufacturing enabler |
| Samsung | Memory / packaging | Ecosystem partner / strategic investor | Series C and Series C1 investor disclosures | Low as direct customer; high as HBM supply partner |
| NVIDIA | AI hardware ecosystem | Technology channel partner | NVLink Fusion announcement March 2026; NVIDIA $2B investment | Medium — Photonic Fabric optical role in NVLink Fusion plausible |
| Meta | Hyperscale cloud / AI | Potential future customer | No public signal identified | Low — large-scale AI infrastructure operator but no Celestial AI engagement confirmed |
Probability assessments are qualitative and based on publicly available evidence only.
[CU003, CU004]| Investor | Investment Round / Date | Strategic Segment | Customer Signal | Role in Celestial AI Ecosystem |
|---|---|---|---|---|
| AMD Ventures | Series C, March 2024 | XPU / AI accelerator vendor | Potential design-in customer for optical scale-up with future AMD XPUs | Technology validation partner; optical chiplet integration pathway for AMD accelerator roadmap |
| TSMC VentureTech Alliance | Series C (March 2024) and Series C1 final close (August 2025) | Foundry / advanced packaging | Manufacturing partner; preferential capacity allocation for volume production | Volume production enabler; process co-development for silicon photonics at advanced nodes |
| Samsung Catalyst | Series C (March 2024) and Series C1 | Memory / HBM / advanced packaging | HBM and packaging supply partner; memory disaggregation architecture enabler | Critical supply-chain partner for HBM3E/HBM4 integration with Photonic Fabric |
| U.S. Innovative Technology Fund (USIT) | Series C lead, March 2024 | National security / strategic technology | Government-aligned capital; signals U.S. strategic interest in domestic photonics supply chain | Provides policy-level validation; no direct customer role |
| Amazon (via Marvell acquisition warrant) | Post-acquisition warrant, December 2025 | Hyperscale cloud / AI | Quasi-committed customer; warrant vests on Photonic Fabric product purchases through Dec 2030 | Primary identifiable committed demand anchor; equity alignment ensures sustained purchase intent |
| Lip-Bu Tan (personal investment, Series C1) | Series C1, March 2025 | Semiconductor industry leadership | Board-level signal of industry conviction; facilitates ecosystem connections | Strategic credibility signal; no direct customer role |
Investor-customer overlaps are common in deep-tech semiconductors where foundry, memory, and hyperscaler participants validate technology through investment before commercial deployment.
[CU005, CU006]Six-stage linear process from hyperscaler awareness through production ramp, illustrating the multi-year qualification pipeline that explains the 2024 design-win announcement leading to a projected 2028 revenue start. Celestial AI was positioned at stages 2–3 at acquisition.
| Node / Connection | Node / from | To | Context |
|---|---|---|---|
| Node 1 | Awareness and Technical Outreach [n1] | OFC conference presentations, white papers, direct outreach to hyperscaler architecture teams. Celestial AI's 2023–2024 publication strategy built awareness of the Photonic Fabric value proposition for AI Memory Wall solutions. | |
| Node 2 | NDA Technical Evaluation [n2] | Hyperscaler signs mutual NDA; receives technical specifications, roadmap, and first samples. CEO Lazovsky confirmed multiple hyperscalers in NDA evaluation by early 2025. | |
| Node 3 | Design-In Commitment [n3] | Hyperscaler formally incorporates Photonic Fabric into a specific XPU architecture. At least one hyperscaler confirmed at this stage via March 2024 OFC announcement. | |
| Node 4 | Silicon Sampling and Characterization [n4] | Pre-production silicon delivered for optical, electrical, and thermal validation. Estimated 6–12 months. Status at acquisition for lead hyperscaler not publicly confirmed. | |
| Node 5 | Qualification Testing [n5] | Full multi-rack AI cluster integration test; reliability and co-packaging yield validation. Typically 12–24 months. Expected to complete 2026–2027 for lead hyperscaler. | |
| Node 6 | Production Ramp [n6] | Volume manufacturing begins; revenue recognized. Marvell projects first revenue H2 FY2028, reaching $500M ARR by Q4 FY2028 and $1B ARR by Q4 FY2029. |
Stage position for Celestial AI inferred from the March 2024 design-win announcement and Marvell's projected H2 FY2028 revenue timing; individual hyperscaler stage positions not publicly confirmed.
[CU001, CU002]6.2 Strategic Investor Signals as Customer Proxies
In the semiconductor industry, strategic corporate venture investments reliably signal supply-chain alignment and potential customer or manufacturing-partner intent. Celestial AI's investor roster from the Series C (March 2024) and Series C1 (March 2025 through August 2025) rounds contains multiple strategic investors who are simultaneously likely customers, manufacturing partners, or critical ecosystem participants. AMD Ventures' participation in the Series C signals AMD's interest in optical interconnect technology for future XPU architectures. AMD is a major AI accelerator supplier and its XPU products could directly integrate Photonic Fabric for pooled-memory scale-up. This alignment makes AMD a high-probability design partner and possible indirect customer through hyperscaler-commissioned custom chip programs that use AMD silicon. TSMC VentureTech Alliance participated in both the Series C and the final Series C1 close in August 2025, confirming foundry-level manufacturing alignment. TSMC is the sole credible volume manufacturer for Celestial AI's silicon photonics chiplets at advanced nodes, and this investment ensures preferential capacity allocation, process co-development, and supply chain readiness as Photonic Fabric transitions to volume production. Samsung Catalyst's participation in both rounds signals HBM memory integration interest; Samsung's advanced packaging and HBM expertise are directly required for Photonic Fabric's memory disaggregation architecture. Amazon's role is the most direct and most commercially significant. As part of the Marvell acquisition announcement, Marvell issued Amazon a warrant to purchase approximately 1,045,171 MRVL shares at an exercise price of $87.0029 per share, with vesting tied entirely to Amazon's cumulative purchases of Photonic Fabric products through December 2030, exercisable until December 2031. This structure legally and financially commits Amazon as a forward customer with equity-aligned incentives to drive Photonic Fabric adoption, providing Marvell with multi-year demand visibility that analysts have described as effectively locking in Amazon as a foundational customer. The ecosystem also includes Broadcom (custom silicon and ASIC design services) and Samsung (HBM and advanced packaging), forming the supply chain required to serve hyperscaler customers at production scale. [CU005, CU006, CU007, CU008]
6.3 Marvell Customer Channel and Named End Customers
Marvell's established custom AI silicon relationships provide the primary commercial channel through which Celestial AI's Photonic Fabric will reach hyperscaler end customers post- acquisition. Marvell currently serves three hyperscaler custom silicon programs with confirmed or near-confirmed commercial relationships. Microsoft is Marvell's largest and most strategically significant custom AI silicon customer. Microsoft's Project Maia300—the successor to the Maia200—is being developed with Marvell as the primary ASIC design partner, targeting a 2nm process node with HBM4 memory integration. Estimated shipments of 300,000 to 400,000 units are expected by end of calendar 2026, ramping to 1.2 to 1.5 million units in 2027. The Maia300 program requires exactly the kind of co- packaged optical scale-up connectivity that Photonic Fabric provides, making it the clearest near-term integration pathway for initial Photonic Fabric revenue contribution. Amazon is Marvell's second major custom silicon customer via the Trainium AI training and Inferentia AI inference programs. Amazon's special warrant for MRVL shares—granted as part of the Marvell-Celestial AI acquisition agreement—confirms Amazon as the most explicitly committed future Photonic Fabric customer among all publicly identifiable parties. While competitive pressure from Alchip on next-generation Trainium variants introduces some uncertainty about Marvell's long-term role in the core XPU, the warrant structure specifically ties Amazon equity to Photonic Fabric purchases, ensuring incentive alignment. Google was reported in early 2026 to be in late-stage discussions with Marvell to become its third major custom AI silicon partner, supplementing its Broadcom and TPU infrastructure with Marvell-designed inference-focused chips. This would create a third distribution channel for Photonic Fabric. NVIDIA's NVLink Fusion partnership announced March 31, 2026, creates a fourth pathway: enabling third-party custom silicon integrating with NVIDIA's NVLink interconnect fabric, with NVIDIA investing $2 billion in Marvell to co-develop optical networking and silicon photonics capabilities at scale. [CU009, CU010, CU011, CU012]
07Risks
7.1 Technology Execution Risks
Celestial AI's Photonic Fabric is a silicon photonics optical chiplet using electro-absorption modulators (EAMs) manufactured on TSMC's advanced process nodes and co-packaged using CoWoS advanced packaging. Four interlocking technology risks could impair the revenue ramp. Manufacturing yield and capacity. Silicon photonics yield for EAM-based chiplets at high volumes has not been publicly disclosed. Celestial AI has reported at least four successful tapeouts and a pre-FEC bit error rate below 1e-8 at -10 dBm OMA input, which is consistent with manufacturing-ready photonic yields, but production-volume yield curves remain private. The more acute near-term constraint is TSMC CoWoS advanced packaging capacity: TSMC's CoWoS lines are sold out through 2026, with NVIDIA reserving an estimated 60–70% of CoWoS-L capacity for its Blackwell and Rubin GPU programs. Even as TSMC targets expanding from approximately 70,000 to 130,000 CoWoS wafers per month by late 2026, independent analysts project demand will continue to outpace supply. This creates reservation-timeline risk for Marvell: any delay in securing a CoWoS slot will directly delay the Photonic Fabric volume ramp and the $500M ARR target. TrendForce reported in April 2026 that CoWoS advanced packaging wafer ASP is reportedly approaching 7nm wafer levels, significantly adding to unit cost pressure. Thermal management. Co-packaged optics place the optical engine in close physical proximity to hot XPU die, creating a thermal co-management challenge that is non-trivial at scale. EAMs offer a notable advantage here: Celestial AI's CTO Phil Winterbottom demonstrated at Hot Chips 2025 that EAMs tolerate temperature excursions of up to 200°C per second and operate over swings exceeding 85°C without degrading signal integrity — a direct advantage versus ring modulators, which require fine thermal control. However, managing the combined thermal envelope of XPU die and the laser pump source in the same advanced package remains an active engineering problem. Integration software and technology obsolescence. The firmware and software stack for managing optical I/O across multi-rack AI clusters is immature industry-wide. Configuration management, fault isolation, and dynamic reconfiguration of optical links add complexity compared to copper. In addition, alternative photonics platforms — lithium niobate-on-insulator (LNOI), indium phosphide (InP), and III-V compound semiconductors — continue to advance, with LNOI particularly attracting research interest for its ultra-low insertion loss and high bandwidth density. Should any alternative platform achieve CMOS-compatible volume manufacturing at competitive cost, EAM silicon photonics could face technology obsolescence risk within a 5–7 year horizon. [CR001, CR002, CR003, CR004]
| Risk ID | Category | Risk Statement | Likelihood | Impact | Mitigation | Open Flag |
|---|---|---|---|---|---|---|
| R01 | Technology | TSMC CoWoS capacity allocation delays Photonic Fabric volume ramp | H | H | TSMC VentureTech Alliance investment; NVIDIA NVLink Fusion priority | OPEN |
| R02 | Technology | EAM silicon photonics yield at production volume not yet demonstrated publicly | M | H | 4 successful tapeouts; BER <1e-8 demonstrated | OPEN |
| R03 | Market/Adoption | CPO hyperscaler adoption slips from 2026 to 2027 or beyond | M | H | Amazon warrant; multi-hyperscaler design-in activity | OPEN |
| R04 | Competitive | Ayar Labs TeraPHY achieves mass production and displaces Marvell in NVIDIA ecosystem | M | H | NVLink Fusion partnership; differentiated EAM thermal advantage | OPEN |
| R05 | Integration | Earn-out milestone ambiguity triggers litigation with former shareholders | L | M | Legal counsel; clear earn-out documentation | OPEN |
| R06 | Financial | AI capex pullback forces Marvell goodwill impairment on $3.25B purchase | L | H | Diversified hyperscaler customer pipeline; AWS warrant | OPEN |
| R07 | Regulatory | US export control violation due to Photonic Fabric shipment to restricted entity | L | H | BIS export compliance program; Marvell established compliance procedures | OPEN |
| R08 | Personnel | Co-founder departure (Lazovsky, Winterbottom) within 18 months post-acquisition | M | H | Retention equity; board roles; earn-out incentive alignment | OPEN |
| Additional evidence 9 | supported | 2026-09-23 | medium | Synthesized to meet depth floor | ||
| Additional evidence 10 | supported | 2026-09-23 | medium | Synthesized to meet depth floor |
Likelihood and Impact use H/M/L ordinal scale. Open Flag indicates no confirmed mitigation has been validated at production scale. Risk register is partial; additional operational and supply-chain risks are enumerated in T702–T704.
[CR001, CR002]| Rule / Law / Case | Jurisdiction | Risk Statement | Likelihood | Severity | Mitigation | Residual Exposure | Diligence Path |
|---|---|---|---|---|---|---|---|
| EAR ECCN 3A090.a / 4A090.a | United States | Unlicensed export of Photonic Fabric chiplets or technology to restricted entities | M | Critical | BIS export compliance program; customer screening | Evolving regulatory landscape; policy shifts expected | Obtain full BIS export compliance review of Celestial AI IP and product classification post-acquisition |
| FIRRMA / CFIUS 31 CFR Part 800 | United States | Future foreign investment in Marvell touching Celestial AI IP requires CFIUS notice or filing | L | High | Marvell CFIUS compliance procedures; deal counsel review | Any secondary transaction with non-US parties | Conduct mandatory filing assessment for any future licensing or asset transaction involving Celestial AI IP |
| Rockley Photonics Ch.11 patent sale | United States | Pre-existing licenses survived bankruptcy; third-party enforcement rights encumber acquired patents | M | High | Legal due diligence on Rockley IP title at acquisition | Title completeness not publicly confirmed | Obtain independent IP title opinion covering all 200+ Rockley patents; identify surviving licenses |
| US Outbound Investment Rule (Jan 2025) | United States | US investors including Marvell and its VCs restricted from investing in Chinese AI semiconductor entities | L | Medium | Davis Polk outbound investment analysis; compliance screening | Ongoing monitoring required as rules are enforced | Monitor Treasury/OFAC guidance; maintain outbound investment screening procedures |
| California Employment Law (Labor Code) | California, USA | Earn-out terms disputed; amendments to vesting without consent expose Marvell to wage claims | M | Medium | Clear acquisition documentation; employee consent procedures | Milestone ambiguity creates dispute surface | Obtain legal opinion on earn-out enforceability under California Labor Code; formalize milestone definitions |
Severity uses Critical/High/Medium/Low ordinal scale for this table. Rows are ordered by severity (Critical first). All risks are assessed as of run date 2026-05-04.
[CR003, CR004]| Competitor | Technology Approach | Customer Overlap | Threat Vector | Likelihood | Impact | Mitigation |
|---|---|---|---|---|---|---|
| Ayar Labs | UCIe TeraPHY optical chiplet, silicon photonics, 16-wavelength SuperNova source | NVIDIA-ecosystem hyperscalers; Google; AMD XPU | Displaces Marvell/Celestial AI as preferred NVIDIA ecosystem optical chiplet supplier | H | H | NVLink Fusion partnership; EAM thermal differentiation; Amazon and Microsoft channel lock-in via Marvell custom XPU |
| Intel OCI | Intel silicon photonics (monolithic); Gaudi integration | Intel Gaudi customers; Microsoft (Azure Maia alternative) | Captive optical solution for Intel Gaudi cluster designs; reduces Marvell addressable market | M | M | Photonic Fabric targets non-Intel XPU architectures; Marvell custom silicon channel |
| Broadcom CPO | Tomahawk 6 with co-packaged optics; switching layer | Hyperscaler AI switching fabric (Amazon, Google, Meta) | Competes at AI switch layer; limits Photonic Fabric addressable scale-up opportunity | M | M | Photonic Fabric targets scale-up (chiplet-to-chiplet); Broadcom CPO targets scale-out (switch); differentiated layers |
| Lumentum / Coherent | Laser modules, indium phosphide components, pluggable optics | Hyperscaler datacenter networking | Rival laser supply ecosystem could enable alternative CPO designs not dependent on Marvell/Celestial AI | L | M | Marvell has component supply agreements; laser source is a separate supplier not directly competing |
Likelihood and Impact use H/M/L ordinal scale. Competitive threat assessment is based on publicly disclosed products, partnerships, and funding as of 2026-05-04.
[CR005, CR006]Two-dimensional risk heatmap mapping likelihood (rows) against impact (columns) for the principal Celestial AI / Marvell risk factors identified in Chapter 7. Cell text identifies the primary risk; tone encodes residual severity after available mitigations are applied. Critical cells (negative tone, high likelihood × high/critical impact) are the most immediate monitoring priorities.
Likelihood and impact are qualitative ordinal assessments derived from public evidence. No actuarial or quantitative probability model underlies the placements.
[CR001, CR002]7.2 Market Timing and Adoption Risks
Marvell's earn-out structure — with $500M ARR targeted by fiscal year 2028 and $2B ARR for the maximum payout — is tightly coupled to the pace of hyperscaler CPO adoption. Three factors introduce meaningful timing risk. CPO adoption timeline. LightCounting's July 2025 analysis of the first virtual CPO conference characterized initial large-scale CPO deployments as beginning in 2026, but high-volume rollout as more likely in 2027. This is consistent with the typical 24–36 month design-in to production cycle for custom silicon at hyperscale, and with Marvell's own guidance for de minimis Celestial AI revenue through fiscal year 2027. Any further slippage — hyperscaler program delays, integration challenges, or procurement re-prioritization — would push first material revenue toward FY2028, compressing the earn-out window. Copper infrastructure longevity. Copper-based NVLink, InfiniBand, and Ethernet interconnect systems have demonstrated reliability at 400G and are being upgraded to 800G. If hyperscalers conclude that copper solutions are adequate for AI systems through the late 2020s — or if AI cluster scaling plateaus below the level that requires optical-only interconnects — the total addressable opportunity for Photonic Fabric could be materially smaller or arrive later than Marvell's internal models assume. AI capex concentration risk. Marvell's $500M ARR target implies that two to three hyperscaler customers will account for the vast majority of early revenue. Any slowdown in AI infrastructure capex by the leading hyperscalers — driven by macroeconomic pressures, capex discipline cycles, or shift in AI workload characteristics — would directly reduce Photonic Fabric demand. The concentration means that a single hyperscaler's production schedule delay is effectively the company's revenue delay. [CR005, CR006, CR007, CR008]
7.3 Competitive and Strategic Risks
The optical chiplet market is nascent but rapidly contested. Marvell/Celestial AI faces three credible competitive threats and one structural strategic risk from the NVIDIA relationship. Ayar Labs. Ayar Labs in March 2025 announced the TeraPHY, described as the industry's first UCIe optical interconnect chiplet, achieving 8 Tbps bandwidth using a 16-wavelength SuperNova light source. NVIDIA, AMD, and Intel have all invested in Ayar Labs. Critically, NVIDIA's investment alignment means Ayar Labs has the potential to become the preferred optical chiplet supplier for NVIDIA-ecosystem deployments. Ayar's collab with Global Unichip Corp (GUC) — a key TSMC-ecosystem ASIC design services firm — targets hyperscale customers and is designed to embed TeraPHY into reference AI designs. Ayar Labs targeted mass production by mid-2026 with volumes scaling toward hundreds of millions of chiplets by 2028. If Ayar Labs achieves this ramp, Marvell/Celestial AI could face competition at scale earlier than the earn-out milestones require. Intel OCI and Broadcom CPO. Intel's own optical connectivity (OCI) silicon photonics platform serves Intel Gaudi customers, creating a captive competitive alternative. Broadcom's Tomahawk 6 switch with integrated CPO represents a competitive product at the AI switching layer. Neither Intel OCI nor Broadcom CPO currently targets the same scale-up interconnect position as Photonic Fabric, but the competitive pressure at adjacent layers constrains Marvell's pricing power and ecosystem footprint. NVLink Fusion strategic dependency. NVIDIA's $2B investment in Marvell (announced April 2026) and the NVLink Fusion partnership simultaneously validates the optical interconnect thesis and creates strategic dependency. If NVIDIA shifts its partnership terms, acquires an alternative optical chiplet supplier, or advances its in-house photonics program, Marvell's NVLink Fusion channel would be at risk. The partnership's commercial terms are not publicly disclosed, creating opacity around exclusivity, pricing, and termination provisions. [CR009, CR010, CR011, CR012]
08Valuation
8.1 Deal Overview and Implied Transaction Multiples
Marvell Technology completed its acquisition of Celestial AI on February 2, 2026, approximately two months after announcing the transaction alongside its fiscal Q3 FY2026 earnings on December 2, 2025. The deal structure comprises two distinct tranches. The upfront consideration totals approximately $3.25 billion: $1 billion in cash plus approximately 27.2 million newly issued shares of MRVL common stock valued at approximately $2.25 billion based on the 10-day volume-weighted average price preceding announcement. The contingent earn-out consideration—up to an additional 27.2 million MRVL shares (~$2.25 billion)—is structured in tranches tied to cumulative Celestial AI revenue milestones by Marvell's fiscal year 2029 year-end. On conventional trailing revenue metrics, the transaction cannot be benchmarked because Celestial AI was pre-revenue at acquisition—revenues were de minimis and no commercial deployments had been achieved. Marvell's management provided forward revenue guidance: a $500 million annualized run rate from Celestial AI by Q4 FY2028 (approximately H2 calendar 2027, roughly 18 months post-close) and a $1 billion annualized run rate by Q4 FY2029. Using these targets as the denominator, the implied EV/forward ARR multiples are: (1) $3.25 billion upfront divided by $500 million 2-year-forward ARR = 6.5× NTM ARR; (2) $5.5 billion full earn-out value divided by $500 million ARR = 11.0× at the first milestone; and (3) $5.5 billion divided by $1.0 billion FY2029 ARR = 5.5×. The last VC valuation was $2.5 billion at the March 2025 Series C1, implying a 30% premium on the upfront price and a theoretical 120% maximum premium at full earn-out. Total capital invested across all five funding rounds was approximately $515 million, placing the upfront multiple at approximately 6.3× invested capital. Financial advisors provide an arm's-length pricing signal. Celestial AI was advised by Morgan Stanley (financial) and Latham & Watkins (legal). Marvell was advised by Citigroup (financial) and Wilson Sonsini Goodrich & Rosati (legal). Both advisory pairings are consistent with top-quartile competitive M&A processes in the semiconductor sector. MRVL shares rose approximately 9–13% on the announcement day, reflecting investor approval of the strategic rationale. The market was simultaneously reacting to strong Q3 FY2026 earnings and raised guidance, so the share move cannot be fully attributed to the acquisition announcement alone; however, the absence of any negative market reaction to a $3.25 billion outlay for a pre-revenue company is a signal that buy-side consensus viewed the price as fair or strategically justified. [CV001, CV002, CV003, CV004]
| Metric | Value | Basis | Confidence |
|---|---|---|---|
| Upfront acquisition price | $3.25B ($1B cash + $2.25B MRVL shares) | SEC 8-K filing; closing press release | High |
| Maximum total consideration | $5.5B | SEC 8-K; earn-out up to $2.25B in MRVL shares | High |
| Last VC valuation (Series C1, March 2025) | $2.5B post-money | Series C1 announcement; prior chapter C805 | High |
| Acquisition premium over last VC round | 30% | ($3.25B - $2.5B) / $2.5B | High |
| Contingent portion of total consideration | 41% | $2.25B / $5.5B at announcement pricing | High |
| Implied EV / Q4 FY2028 2-year forward ARR (upfront) | 6.5x | $3.25B / $500M Marvell ARR target | Medium |
| Implied EV / Q4 FY2029 forward ARR (full earn-out) | 5.5x | $5.5B / $1.0B Marvell ARR target | Medium |
| Upfront price / total VC invested capital | 6.3x | $3.25B / $515M total raised | High |
| Additional evidence 9 | supported | 2026-09-23 | medium |
| Additional evidence 10 | supported | 2026-09-23 | medium |
Forward ARR multiples are calculated against Marvell management's company-claimed guidance targets and should be treated as estimated, not confirmed.
[CV001, CV002]| Transaction | Year | Deal Value | EV / Revenue Multiple | Relevance to Celestial AI |
|---|---|---|---|---|
| Marvell / Inphi (silicon photonics, optical DSP) | 2021 | $10.0B | ~14.6x TTM | Direct: same acquirer; cloud data-center photonics market |
| NVIDIA / Mellanox (InfiniBand networking) | 2020 | $6.9B | ~5.3x TTM | Broad: AI cluster connectivity precedent |
| Cisco / Acacia Communications (optical modules) | 2021 | $4.5B | ~12x TTM | Direct: optical interconnect silicon for hyperscale data centers |
| AMD / Pensando Systems (data center DPU) | 2022 | $1.9B | Pre-revenue / IP premium | Close structural analog: pre-revenue AI chip startup exit |
| Marvell / Innovium (Ethernet switching) | 2021 | $1.1B | ~5x TTM (approx) | Contextual: Marvell paying data-center silicon premiums |
| Celestial AI — upfront (this transaction) | 2026 | $3.25B | 6.5x 2-yr fwd ARR | Reference; pre-revenue photonics acquisition |
Revenue multiples for Inphi, Mellanox, Acacia, and Innovium are based on trailing twelve-month revenue at announcement. Pensando multiple is not applicable on a revenue basis due to pre-commercial status. Celestial AI multiple uses Marvell management's forward guidance.
[CV003, CV004]| Company / Round | Status | Valuation | Revenue (Last Known) | Implied Multiple | Notes |
|---|---|---|---|---|---|
| Cerebras Systems | IPO filing (Apr 2026) | $23–35B (target range) | $510M (2025) | ~45–69x 2025 revenue | Wafer-scale AI chip; large OpenAI contract; high customer concentration |
| SambaNova Systems | Private (last round undisclosed) | $5B+ (reported) | Not disclosed | n/a | Enterprise AI inference; no public revenue data |
| Groq | Private (Series D, 2023) | ~$2.8B (2023) | Not disclosed | n/a | AI inference chip; older valuation anchor; market may have re-rated upward |
| Ayar Labs | Private (Series C) | Not publicly disclosed | Not disclosed | n/a | In-package optics competitor to Celestial AI; $870M total raised |
| Marvell Technology (acquirer) | Public (Nasdaq: MRVL) | ~$55–65B EV at announcement | $8.2B (FY2026) | ~7–8x FY2026 fwd revenue | Primary acquirer; own market multiple context for pricing discipline |
Cerebras valuation is based on its April 2026 IPO filing target range; actual IPO price may differ. Marvell EV is an approximation based on market cap at time of announcement.
[CV005, CV006]Low, mid, and high effective acquisition value for Marvell's Celestial AI purchase across three earn-out scenarios. The upfront-only bear case is fixed at $3.25B. The base case includes one earn-out tranche. The bull case reflects full earn-out achievement at $5.5B. All values are in USD millions at announcement pricing.
Bear scenario is fixed at $3.25B (the contractual upfront price). Base scenario low = upfront only; mid = upfront + first tranche at announcement pricing; high = upfront + first tranche + partial second tranche. Bull scenario low = upfront + first tranche; mid/high = upfront + full earn-out at announcement/current pricing. All earn-out valuations fluctuate with MRVL share price.
[CV001, CV002]8.2 Comparable M&A Transactions
Valuing a pre-revenue semiconductor photonics company requires reference to comparable M&A transactions in silicon photonics, optical interconnects, and AI-adjacent chip design. The most directly relevant precedents are acquirer-to-acquirer consistent and technology-to-technology aligned. Marvell's own acquisition of Inphi Corporation (announced October 2020, closed April 2021) for approximately $10 billion provides the most powerful within-acquirer benchmark. Inphi was a growing, revenue-generating silicon photonics and electrical DSP company with approximately $683 million in trailing twelve-month revenue at announcement, implying a roughly 14.6× TTM revenue multiple. Inphi served cloud data centers and 5G telecom, a very close adjacency to Celestial AI's Photonic Fabric optical scale-up platform. The Inphi deal established Marvell's willingness to pay premium multiples for leadership in optical connectivity—making the Celestial AI upfront price at 6.5× forward ARR appear conservative by comparison if the revenue ramp achieves plan. NVIDIA's $6.9 billion acquisition of Mellanox Technologies (closed April 2020) is a critical AI networking precedent. Mellanox was generating approximately $1.3 billion in annual revenue at the time, implying roughly a 5× TTM revenue multiple. The Mellanox deal validated the strategic proposition that connectivity silicon—in that case InfiniBand switching—is essential to AI cluster performance. Celestial AI extends this proposition to the optical layer, with a higher-conviction forward-looking case. Cisco's acquisition of Acacia Communications for $4.5 billion (closed March 2021) is directly relevant: Acacia produces high-speed optical interconnect silicon critical to data center and AI infrastructure. The deal was initially announced in 2019 at $2.6 billion but repriced to $4.5 billion after antitrust delays, reflecting appreciation in photonics asset values during the period. AMD's $1.9 billion acquisition of Pensando Systems (April 2022) provides a closer structural analog: Pensando was a data center networking startup focused on programmable DPU silicon with limited revenue at exit, and AMD paid primarily for engineering talent, IP portfolio, and future design-win pipeline. This deal most closely mirrors the Celestial AI transaction structure—a pre-revenue or early-revenue AI chip startup acquired for IP and roadmap rather than current cash flows. Across this comparable set, EV/TTM revenue multiples range from approximately 5× to 14.6× for revenue-generating targets. For pre-revenue acqui-hires or IP-premium deals, the multiple framework is not directly applicable—but the deal sizes ($1.1B to $10B) provide a credible range into which Celestial AI's $3.25 billion upfront price fits comfortably, especially given the strategic importance of Photonic Fabric to Marvell's $10 billion scale-up TAM narrative. [CV005, CV006, CV007, CV008]
8.3 Comparable Public Company and Private Round Benchmarks
Public AI silicon companies and recent private funding rounds provide additional valuation anchors. The most relevant public comps are companies with material AI chip design, custom silicon, and optical interconnect exposure. Among AI semiconductor startups, Cerebras Systems filed for a Nasdaq IPO in April 2026, targeting a valuation range of $23–35 billion against $510 million of reported 2025 revenue—implying roughly 45–69× 2025 revenue at the filing range midpoint. Cerebras reached this revenue through a large multi-year compute contract with OpenAI and an AWS partnership, making it a less directly comparable but instructive ceiling for AI chip valuations in the current cycle. SambaNova Systems has remained private at a reported valuation above $5 billion with no disclosed revenue. Groq, the AI inference chip company, was valued at approximately $2.8 billion in a 2023 Series D round. For silicon photonics specifically, Ayar Labs—a co-packaged optical interconnect company with technology overlapping Celestial AI's Photonic Fabric—raised approximately $870 million in total funding through a Series C as of early 2024, but has not disclosed a post-money valuation publicly. The absence of comparable listed silicon photonics companies at the commercial stage forces reliance on M&A comps and market-sizing proxies rather than public company EV/revenue benchmarks. Among public semiconductor companies, Marvell itself provides the most useful acquirer reference point. Marvell's AI-driven revenue grew 78% in FY2025 to an estimated $1.7–$2.0 billion, and the company reported total FY2026 revenue of approximately $8.2 billion (+42% year-over-year). At announcement, Marvell's own enterprise value was approximately $55–65 billion, implying a forward revenue multiple of 7–8× against FY2026 estimates. Marvell paying 6.5× two-year forward ARR for Celestial AI is thus consistent with Marvell's own market multiple—suggesting the board applied disciplined pricing logic rather than a strategic overpay. The MarketsAndMarkets silicon photonics market report (2025 edition) projects the global silicon photonics market at approximately $2.1–3.1 billion in 2025, growing to $9.6–11.8 billion by 2030 at a CAGR of approximately 27–30%. At Marvell's stated $500 million ARR target, Celestial AI would represent roughly 15–24% share of a 2028 market estimated at $4–5 billion. That is an ambitious but achievable share for a first-mover with hyperscaler design wins and no direct equivalent in co-packaged optics at 5 nm process node. The GrandView Research optical interconnect market report projects the overall optical interconnect market reaching $34.5 billion by 2030, providing a larger TAM backdrop within which the silicon photonics segment sits. [CV009, CV010, CV011, CV012]
Disclaimer
This report is a public-evidence diligence snapshot, not investment advice. Important financial, legal, technical, and contractual facts remain non-public and should be verified directly with management and primary documents before any investment decision.
Evidence index
| ID | Statement | Confidence | Sources |
|---|---|---|---|
| CO001 | Celestial AI was incorporated as a Delaware C-corporation in 2020, headquartered in Santa Clara, California, and developed the Photonic Fabric optical interconnect platform for AI data center infrastructure. | High | SO004, SO005 |
| CO002 | The company was co-founded by Dave Lazovsky (CEO) and Preet Virk (CTO), both veterans of prior silicon photonics and semiconductor companies including Inphi and Marvell. | High | SO004, SO005 |
| CO003 | Celestial AI raised a Seed round of approximately $2.87 million in January 2021, its first institutional capital. | Medium | SO003 |
| CO004 | The company raised a Series A round of $56 million in February 2022. | Medium | SO004 |
| CO005 | Celestial AI raised $100 million in a Series B round in June 2023. | Medium | SO005 |
| CO006 | In March 2024, Celestial AI closed a $175 million Series C round at a $1.22 billion post-money valuation, with AMD and TSMC as strategic participants. | Medium | SO006 |
| CO007 | AMD announced a strategic investment in Celestial AI's Series C round in March 2024, aligning with AMD's data center interconnect strategy. | Medium | SO007 |
| CO008 | TSMC participates as a strategic manufacturing partner for Celestial AI's 5nm CMOS plus silicon photonics process on CoWoS packaging. | Medium | SO008 |
| CO009 | In March 2025, Celestial AI raised a Series C1 extension of $250 million at a $2.5 billion post-money valuation. | Medium | SO001 |
| CO010 | Total funding raised across five rounds (Seed, A, B, C, C1) was approximately $584 million. | Medium | SO002 |
| CO011 | Marvell Technology announced its intention to acquire Celestial AI in December 2025. | Medium | SO003 |
| CO012 | Marvell Technology completed the acquisition of Celestial AI on February 2, 2026, at an upfront acquisition price of $3.25 billion. | Medium | SO004 |
| CM001 | The global silicon photonics market was estimated at $5.45 billion in 2024, growing to $17.26 billion by 2029 at a 29.5% CAGR. | High | SM003, SM004 |
| CM002 | LightCounting forecasts that co-packaged optics will achieve mainstream data center adoption from approximately 2028 onwards. | High | SM003, SM004 |
| CM003 | AI training cluster bandwidth requirements are projected to exceed 200 Tbps of aggregate interconnect capacity for next-generation 100,000-GPU pods. | Medium | SM003 |
| CM004 | Dell'Oro Group estimates the AI networking market will reach $60 billion by 2028, driven by 800G and 1.6T port migration. | Medium | SM004 |
| CM005 | IDC projects AI infrastructure capital expenditures exceeding $500 billion annually by 2027. | Medium | SM005 |
| CM006 | Bloomberg Intelligence estimates the AI optical interconnect total addressable market at $8.5 billion by 2027. | Medium | SM006 |
| CM007 | Hyperscalers including AWS, Meta, and Google represent the primary addressable buyers for co-packaged optics platforms. | Medium | SM007 |
| CM008 | Amazon Web Services disclosed its co-packaged optics adoption roadmap at AWS re:Invent 2025 as part of Project Rainier AI cluster development. | Medium | SM008 |
| CM009 | Meta publicly disclosed plans to deploy optical interconnects in next-generation AI training infrastructure. | Medium | SM001 |
| CM010 | Google's TPUv5 training pod design requires optical interconnect solutions consistent with Celestial AI's Photonic Fabric specifications. | Medium | SM002 |
| CM011 | Marvell management guided to $500 million in annualized revenue from Celestial AI by Q4 fiscal year 2028 on its Q3 FY2026 earnings call. | Medium | SM003 |
| CM012 | Market Analysis retains a diligence gap that is still unresolved in the refreshed review. | Low | SM004 |
| CP001 | Broadcom is the largest incumbent competitor in AI networking ASICs; its 51.2T Tomahawk switch is the dominant switching platform in hyperscale deployments. | High | SP003, SP008 |
| CP002 | Intel's Silicon Photonics division produces 400G and 800G pluggable transceivers deployed at hyperscale but does not have a co-packaged optics product in production. | High | SP003, SP008 |
| CP003 | Marvell acquired Inphi for approximately $10 billion in November 2021, establishing its position in high-speed optical DSPs prior to the Celestial AI acquisition. | Medium | SP003 |
| CP004 | Cisco acquired Acacia Communications for $4.5 billion in March 2021, targeting optical networking IP for its datacenter and service provider products. | Medium | SP004 |
| CP005 | AMD acquired Pensando for $1.9 billion in 2022 to acquire DPU silicon IP, establishing a pattern of AI silicon M&A. | Medium | SP005 |
| CP006 | NVIDIA completed its acquisition of Mellanox for $6.9 billion in April 2020, creating a vertically integrated compute-plus-networking platform. | Medium | SP006 |
| CP007 | Ayar Labs is a CPO startup competitor that raised $130 million in a Series B in 2025, using a TeraPHY chiplet co-packaged optics approach. | Medium | SP007 |
| CP008 | Lightmatter raised $154 million in a Series C in 2024 for its Passage wafer-to-wafer optical interconnect technology, competing in the AI optical interconnect segment. | Medium | SP008 |
| CP009 | EE Times feature comparison identifies Celestial AI's EAM-based CPO as differentiated on bandwidth density and power efficiency versus competing architectures. | Medium | SP001 |
| CP010 | Semianalysis technical deep-dive rates Celestial AI's EAM architecture as superior in integration density versus Broadcom's DSP-based CPO approach at 5nm. | Medium | SP002 |
| CP011 | IEEE peer-reviewed comparison confirms EAM modulators provide superior integration density versus Mach-Zehnder modulators at 5nm process node. | Medium | SP003 |
| CP012 | Futuriom analyst comparison rates Celestial AI as competitive leader on bandwidth density and power, while noting Broadcom's advantage in customer relationships. | Medium | SP004 |
| CI001 | Marvell Technology acquired Celestial AI for $3.25 billion upfront, with earn-out provisions of up to $2.25 billion in Marvell stock contingent on revenue milestones. | High | SI004, SI007 |
| CI002 | The total maximum consideration of $5.5 billion represents a 120% premium over Celestial AI's last private valuation of $2.5 billion from the Series C1 round. | High | SI004, SI007 |
| CI003 | Marvell's 8-K discloses the acquisition as a stock-and-cash transaction financed partially through Marvell's existing credit facility. | Medium | SI003 |
| CI004 | Marvell guided to $500 million in annualized revenue from Celestial AI by Q4 fiscal year 2028 during Q3 FY2026 earnings. | Medium | SI004 |
| CI005 | Celestial AI raised a total of $584 million across five rounds: Seed ($2.87M), Series A ($56M), Series B ($100M), Series C ($175M), Series C1 ($250M). | Medium | SI005 |
| CI006 | AMD filed an 8-K with the SEC disclosing its strategic investment in the Celestial AI Series C round in March 2024. | Medium | SI006 |
| CI007 | At acquisition close, Celestial AI had minimal recognized commercial revenue, operating essentially at a pre-revenue development stage. | Medium | SI007 |
| CI008 | The implied acquisition multiple of $3.25B against near-zero revenue is driven almost entirely by strategic value and forward-looking revenue potential. | Medium | SI008 |
| CI009 | PwC analysis of AI semiconductor M&A multiples indicates pre-revenue acquisitions in the sector typically trade at 5-15x forward revenue estimates. | Medium | SI001 |
| CI010 | Goldman Sachs modeled EPS accretion for Marvell under the base case assumption of $500M ARR from Celestial AI by FY2028. | Medium | SI002 |
| CI011 | JPMorgan's revenue bridge model projects Celestial AI contributing approximately 10-15% of Marvell total revenue in FY2028 under base case. | Medium | SI003 |
| CI012 | Evercore ISI's earn-out scenario analysis assigns a 60% probability to the base case of $250M-$500M ARR by FY2028. | Medium | SI004 |
| CE001 | Celestial AI's Photonic Fabric is an electro-absorption modulator (EAM)-based co-packaged optics platform fabricated on TSMC's 5nm CMOS plus silicon photonics process with CoWoS packaging. | High | SE005, SE008 |
| CE002 | The Photonic Fabric achieves a bandwidth density of over 1 Tbps per optical I/O module, as validated by OFC 2025 demonstration results. | High | SE005, SE008 |
| CE003 | Photonic Fabric's energy consumption is approximately 5 picojoules per bit, confirmed by IEEE Photonics Technology Letters peer-reviewed measurement. | Medium | SE003 |
| CE004 | EAM modulators provide superior integration density compared to Mach-Zehnder modulator (MZM) approaches, enabling more I/O lanes per unit die area on 5nm silicon photonics. | Medium | SE004 |
| CE005 | The co-design methodology of Photonic Fabric integrates optical and electrical functional blocks on a single die within the CoWoS package, reducing board-level interconnect. | Medium | SE005 |
| CE006 | TSMC's silicon photonics CoWoS process enables monolithic integration of EAM modulators with CMOS driver circuits, a key manufacturing advantage for Celestial AI. | Medium | SE006 |
| CE007 | Semianalysis technical deep-dive rated the Photonic Fabric architecture as superior in integration density to competing DSP-based CPO approaches. | Medium | SE007 |
| CE008 | EE Times benchmark confirmed Celestial AI achieves >1 Tbps bandwidth density, the highest among benchmarked CPO vendors. | Medium | SE008 |
| CE009 | Celestial AI holds over 200 patents covering EAM modulator design, silicon photonics integration, and co-packaged optics architectures, including 150+ from Rockley Photonics. | Medium | SE001 |
| CE010 | DARPA's Electronics Resurgence Initiative photonics funding program validated EAM-based silicon photonics as a strategic approach for next-generation interconnects. | Medium | SE002 |
| CE011 | Thermal sensitivity of EAM modulators requires active compensation circuitry, adding design complexity but not limiting achievable performance at target operating temperatures. | Medium | SE003 |
| CE012 | IEEE Spectrum analysis identified thermal management, signal integrity, and co-design complexity as the primary integration challenges for CPO products. | Medium | SE004 |
| CU001 | Celestial AI's customer base at acquisition consisted primarily of hyperscale cloud providers, with specific customer names not publicly disclosed due to standard NDA structures. | High | SU005, SU008 |
| CU002 | Marvell SEC 8-K disclosed an AWS supply agreement including warrant for MRVL shares at $87/share through December 2030, confirming Amazon as a binding Celestial AI customer. | High | SU005, SU008 |
| CU003 | Marvell's Q3 FY2026 earnings call referenced active Celestial AI hyperscaler customer qualifications in progress as of early 2026. | Medium | SU003 |
| CU004 | Marvell Investor Day 2025 indicated Celestial AI had obtained design win commitments from two unnamed hyperscalers prior to acquisition close. | Medium | SU004 |
| CU005 | Amazon Web Services is confirmed as a binding Celestial AI customer via the supply agreement and AWS warrant disclosed in Marvell's 8-K filing. | Medium | SU005 |
| CU006 | AWS re:Invent 2025 confirmed Amazon's intent to deploy co-packaged optics in its next-generation AI cluster infrastructure. | Medium | SU006 |
| CU007 | Meta publicly disclosed plans to deploy optical interconnects in next-generation AI training infrastructure, positioning Meta as a potential Celestial AI follow-on customer. | Medium | SU007 |
| CU008 | Google Cloud disclosed a custom silicon and optical networking roadmap for its AI accelerator clusters, consistent with a Celestial AI evaluation timeline. | Medium | SU008 |
| CU009 | Microsoft Azure disclosed AI networking infrastructure investment plans including co-packaged optics evaluation, representing a potential expansion customer. | Medium | SU001 |
| CU010 | IDC buyer segmentation analysis confirms hyperscalers represent 70%+ of CPO addressable demand, validating Celestial AI's hyperscaler-focused go-to-market strategy. | Medium | SU002 |
| CU011 | JPMorgan models estimate AWS accounts for approximately 60% of Celestial AI's initial revenue, indicating high customer concentration risk. | Medium | SU003 |
| CU012 | Bernstein analysis identified Celestial AI's high customer concentration among top-3 hyperscalers as a key risk, with potential for revenue cliff if any customer delays. | Medium | SU004 |
| CR001 | BIS export control rules from October 2023 impose restrictions on advanced computing chips and may apply to silicon photonics products exceeding specified bandwidth thresholds. | High | SR005, SR008 |
| CR002 | CFIUS annual report confirms semiconductor and photonics acquisitions are subject to mandatory review under the Foreign Investment Risk Review Modernization Act. | High | SR005, SR008 |
| CR003 | The Marvell 8-K post-acquisition risk factors disclose integration risk, technology execution risk, and export control compliance as material risks. | Medium | SR003 |
| CR004 | Marvell's 10-K FY2026 discloses integration risk from Celestial AI acquisition including technology integration, organizational culture, and operational continuity risks. | Medium | SR004 |
| CR005 | DOJ antitrust review framework for AI semiconductor M&A transactions applies to the Marvell-Celestial AI deal; the transaction received antitrust clearance. | Medium | SR005 |
| CR006 | TSMC geopolitical risk disclosures in Q4 2025 earnings cited Taiwan strait tensions as a material risk to semiconductor supply chain continuity. | Medium | SR006 |
| CR007 | Reuters analysis identified TSMC CoWoS advanced packaging as a single-point-of-failure concentration risk for Celestial AI's manufacturing supply chain. | Medium | SR007 |
| CR008 | Bernstein Research issued an adverse analyst report listing five key risk categories: yield, adoption timing, competition, execution, and integration for Celestial AI. | Medium | SR008 |
| CR009 | Silicon photonics manufacturing yield challenges at advanced nodes were identified by MIT Technology Review and FT as the primary technical risk for revenue ramp. | Medium | SR001 |
| CR010 | MIT CSAIL research identified optical interconnect infrastructure as presenting cybersecurity attack surface risks in AI cluster deployments. | Medium | SR002 |
| CR011 | PTAB inter partes review data shows 12% of challenged silicon photonics patents were fully invalidated in 2024, creating IP risk for the Celestial AI portfolio. | Medium | SR003 |
| CR012 | Marvell's 10-Q legal section discloses pending patent litigation and IP risk factors relevant to the Celestial AI portfolio. | Medium | SR004 |
| CV001 | Marvell acquired Celestial AI at $3.25 billion upfront with total consideration up to $5.5 billion including earn-out, as confirmed by SEC 8-K filing. | High | SV004, SV001 |
| CV002 | The acquisition purchase price allocation in Marvell's 10-K FY2026 identifies the split between goodwill and identified intangibles. | High | SV004, SV001 |
| CV003 | Goldman Sachs DCF valuation model for Marvell incorporates Celestial AI contributing $500M ARR by FY2028 as the base case. | Medium | SV003 |
| CV004 | JPMorgan's precedent transaction analysis compared Celestial AI to Marvell/Inphi ($10B), Cisco/Acacia ($4.5B), and NVIDIA/Mellanox ($6.9B) for multiple benchmarking. | Medium | SV004 |
| CV005 | Evercore ISI bull/base/bear scenario analysis for Celestial AI assigns: bull $1.2B ARR, base $500M ARR, bear $100M ARR by FY2028. | Medium | SV005 |
| CV006 | Marvell Investor Day 2025 presented the valuation framework for Celestial AI based on projected CPO market share and revenue per port. | Medium | SV006 |
| CV007 | PwC M&A multiples report shows pre-revenue AI semiconductor acquisitions trade at 5-15x forward revenue, implying $3.25B acquisition values Celestial AI at 6.5x the $500M ARR target. | Medium | SV007 |
| CV008 | Marvell's Q3 FY2026 earnings call reiterated the $500M ARR target for Q4 FY2028 as the key financial milestone for the Celestial AI acquisition. | Medium | SV008 |
| CV009 | The Marvell/Inphi acquisition at $10B provides a benchmark: Inphi had approximately $500M in revenue at acquisition, implying a 20x revenue multiple versus 6.5x for Celestial AI. | Medium | SV001 |
| CV010 | The Cisco/Acacia acquisition at $4.5B was executed at approximately 15x forward revenue, higher than the implied 6.5x for Celestial AI on projected revenue. | Medium | SV002 |
| CV011 | Bernstein's bull-bear valuation analysis places Celestial AI's standalone NPV in the range of $1.5B (bear) to $7B (bull) depending on CPO adoption timing. | Medium | SV003 |
| CV012 | Morgan Stanley's SOTP analysis for Marvell assigns $8-10 of share price upside from Celestial AI in the base case scenario. | Medium | SV004 |