Panthalassa

1.1 Identity, product model, and operating footprint

Panthalassa presents itself as a Portland, Oregon company building a fundamentally new kind of energy and computing infrastructure: autonomous floating platforms, which it calls nodes, that harvest ocean wave energy and convert it into electricity to run artificial-intelligence data centers directly at sea. Rather than transmitting power to shore, the company’s stated model is to co-locate generation and compute on the same offshore structure and sell AI compute capacity, sidestepping grid interconnection queues and land constraints. Multiple 2026 sources place its headquarters in Portland with prototype and sea-trial activity across the Pacific Northwest, including the Strait of Juan de Fuca and Puget Sound. The company is organized as a public benefit corporation, and coverage of its May 2026 financing spans mainstream technology, climate, and business outlets, signaling a broad public profile despite its early stage. In practical terms, each node is designed to operate autonomously offshore, with the wave-driven power system and the compute payload housed on the same structure so that no subsea power cable or grid interconnection is required. That architecture is the core of the company’s pitch: siting compute where energy is abundant at sea, rather than competing for scarce land, grid capacity, and interconnection approvals on shore.[CO001, CO002, CO003, CO005, CO032, CO034]

1.2 Founders, leadership, and key-person dependence

Founding attribution is consistent across retained sources. Panthalassa was founded in 2016 and is led by co-founder and chief executive Garth Sheldon-Coulson, previously a senior investment associate and AI researcher at Bridgewater Associates, alongside co-founder and chief innovation officer Brian Moffat, an ocean-energy researcher who worked on wave energy at Spindrift Energy and holds three bachelor of science degrees from UC Irvine. The pairing anchors founder-market fit across the two disciplines the company must fuse: capital-markets and AI fluency on one side, and marine wave-energy engineering on the other. Public disclosure emphasizes this engineering-led leadership rather than a large named executive bench or a disclosed board roster, which leaves governance visibility thin and concentrates execution risk in two key people. That key-person concentration is a material diligence item given the decade-long, capital-intensive path from prototype to commercial fleet.[CO004, CO006, CO007, CO008, CO009, CO039]

1.3 Funding, valuation, and investor base

Panthalassa announced a $140 million Series B on May 4, 2026, led by Peter Thiel, bringing disclosed total capital raised to roughly $210 million. Financial Times characterized the company as a "$1bn ocean data centre start-up," implying a near-unicorn valuation, though retained sources do not disclose an exact post-money figure. The syndicate blends returning backers, including Founders Fund, Gigascale Capital, Lowercarbon Capital, Unless, and WovenEarth, with a deep new-investor bench spanning John Doerr, Marc Benioff’s TIME Ventures, Max Levchin’s SciFi Ventures, Susquehanna Sustainable Investments, Hanwha Group, Fortescue Ventures, Super Micro Computer, Sozo Ventures, and local Oregon funds. Related special-purpose vehicles filed SEC Form D notices in 2026, disclosing smaller allocation pools rather than the full round. The Series B is a step-change in capital relative to estimated prior funding and underwrites the leap from sea-trial prototypes toward a first commercial pilot.[CO012, CO013, CO015, CO016, CO017, CO018]

1.4 Milestones, trajectory, and open questions

Panthalassa’s public timeline runs from its 2016 founding through a 2021 Ocean-1 prototype deployment in the Strait of Juan de Fuca, later prototypes generating on the order of 50 kilowatts in Puget Sound by 2025, the May 2026 Series B, a planned Ocean-3 pilot node in the northern Pacific around August 2026, and a stated commercial-deployment target near 2027. The company originally explored producing hydrogen or clean fuels before emphasizing AI compute, and it articulates a long-term vision of thousands of autonomous nodes. Against that momentum, the company is pre-revenue with no publicly disclosed paying customers, and independent commentators caution that corrosive, mechanically harsh ocean conditions raise real execution risk; Panthalassa did not respond to at least one skeptical press inquiry. Valuation precision, revenue, customer proof, and board composition remain the principal unresolved diligence items carried into later chapters.[CO024, CO025, CO026, CO027, CO028, CO029]

1.5 Exhibits

2.1 Market boundary and substitutes

Panthalassa should not be sized as the entire wave-energy market or the entire cloud-computing market. The investable boundary is offshore-sited AI compute powered by ocean wave energy, where the company generates electricity on an autonomous node and consumes that electricity onboard to sell compute capacity. Included spend therefore covers the integrated stack: wave-energy conversion, marine deployment, cooling and power conditioning, onboard AI servers, network backhaul, and the commercial compute service. Excluded spend includes conventional wave-power projects that export electricity to shore, generic renewable-energy generation, and ordinary data-center construction unless the spend is being displaced by offshore compute. The relevant status quo is land-based hyperscale or colocation capacity connected to grid power, while adjacent alternatives include offshore wind, floating solar, nuclear or SMR-backed data centers, and plain grid procurement. This boundary is intentionally narrow because the market thesis depends on solving the AI power bottleneck through co-located ocean generation, not on proving all marine energy is a large market.[CM001, CM002, CM003, CM004, CM005, CM013]

2.2 Two-lens sizing: huge AI demand versus tiny wave supply

The most useful sizing approach uses two lenses and refuses to merge their units. The demand lens is enormous: IEA projects data-center electricity demand could reach about 945 TWh per year by 2030, and SemiAnalysis frames energy availability as a central AI-infrastructure constraint. That lens supports a very large TAM for energy-backed compute capacity, but it does not prove Panthalassa can serve it. The supply lens is much smaller: Mordor estimates wave-energy deployment around 10 MW in 2026 growing to about 125 MW by 2031, with other analyst houses also describing a high-growth but early market. Panthalassa's SAM is the overlap between those lenses, namely wave-powered offshore AI compute. Its supportable near-term SOM is narrower still: pilot and early commercial nodes in 2026-2027, not a fleet-scale percentage of global AI electricity demand. The chapter therefore preserves TWh, MW, and node-count units separately rather than producing a single inflated TAM number.[CM006, CM007, CM008, CM010, CM011, CM012]

2.3 Buyer, user, payer, and adoption path

The buyer map follows cloud and AI infrastructure budgets rather than utility procurement. Hyperscalers such as AWS, Azure, and Google are the most obvious archetype because they buy data-center capacity, power, networking, and hardware at scale. Neocloud and AI-compute specialists are also plausible because they monetize scarce GPU capacity and may tolerate non-traditional siting if economics and reliability work. AI labs are more likely users than sole payers unless they contract directly for batch or inference capacity. Adoption should begin with pilots, non-mission-critical batch work, and capacity reservations, because offshore autonomous nodes have to prove uptime, security, remote maintainability, and network performance before production workloads move. Satellite backhaul makes low-latency interactive services a weak first use case, while batch inference, training-adjacent jobs, or delay-tolerant workloads are more credible. Budget ownership likely spans cloud infrastructure, data-center capacity planning, AI infrastructure, and strategic energy procurement.[CM009, CM020, CM021, CM022, CM023, CM024]

2.4 Growth drivers, constraints, and diligence gaps

The bull case is driven by AI power scarcity, interconnection delays, land constraints, decarbonization pressure, and strategic interest in sovereign compute capacity. Those drivers matter for valuation because they can make unconventional siting economically relevant if the alternative is waiting years for land, grid access, and power purchase agreements. The constraints are equally material. Commercial wave energy remains early; generic marine-energy LCOE around $388 to $618 per MWh is far above mainstream solar or wind benchmarks; offshore structures and GPU payloads create major capital intensity; and regulatory, maritime, environmental, and trust hurdles can delay adoption even without shore power export. Panthalassa and its investors cite aspirational economics, including low target power costs and factory-scale node production, but retained public sources do not disclose signed customers, pricing, node-level uptime, contracted capacity, or validated unit economics. Diligence should therefore preserve contradictory estimates instead of smoothing them: enormous AI electricity demand is real, wave deployment is tiny, and Panthalassa's actual SAM/SOM depends on pilot results and buyer trust.[CM014, CM015, CM026, CM027, CM028, CM029]

2.5 Exhibits

3.1 Landscape: direct peers, substitutes, and likely entrants

Panthalassa does not map cleanly to a single competitor category. The closest generation-side peers are wave-energy developers such as CorPower Ocean, Oscilla Power, Eco Wave Power, C-Power, Marine Power Systems, Wave Swell Energy, and Carnegie Clean Energy, but their public product surfaces primarily emphasize electricity generation rather than selling AI compute at sea. The closest compute-side peers are floating, underwater, or ship-based data-center concepts such as Aikido, NetworkOcean, Microsoft Project Natick, Subsea Cloud, Highlander Hailanyun, and Mitsui O.S.K. ship-based studies, yet those alternatives generally rely on offshore wind, subsea cooling, barges, or ships rather than Panthalassa-style autonomous wave-powered nodes. The status quo remains land hyperscale data centers, GPU-cloud neoclouds, internal build, grid power, and emerging nuclear or SMR-for-data-center options. Starcloud is a useful frontier analog because it pursues remote renewable-powered compute in space, demonstrating that investors will fund non-land infrastructure when terrestrial constraints look binding.[CP001, CP002, CP031, CP032, CP033, CP045]

3.2 Profiles, maturity, and feature breadth

The profile comparison shows Panthalassa as the only retained company claiming both wave generation and onboard AI compute in one commercial architecture. CorPower has clearer utility-scale wave-array packaging through 10-30MW CorPack clusters; Oscilla has a patent-backed Triton wave-energy converter; Eco Wave Power has public-company visibility and a 404.7 MW pipeline; and Aikido, NetworkOcean, Natick, Highlander, and Subsea Cloud attack the at-sea data-center problem from cooling, floating-platform, or subsea angles. Microsoft Project Natick is the most mature public proof that servers can operate subsea, including an 864-server deployment and lower failure rates than a land control group, but the same program is no longer active, which weakens any assumption that technical success implies commercial continuity. The feature map therefore gives Panthalassa high breadth and low proof, while land neoclouds invert that pattern with high commercial maturity but no autonomous ocean-renewable moat.[CP003, CP006, CP008, CP009, CP011, CP012]

3.3 Capability, pricing, GTM, and trust posture

Capability comparisons favor Panthalassa on architectural integration, but not yet on proof, pricing, or trust. Public sources support the node concept, cable-free architecture, and at-sea processing path, while wave peers support renewable-generation credibility and offshore-compute peers support cooling or siting credibility. Public list pricing is largely absent across the relevant frontier set, so buyers cannot yet compare Panthalassa on $/GPU-hour, $/kW, service-level terms, data-egress economics, latency, or contract duration. GTM power is also asymmetric: established hyperscalers, GPU-cloud providers, and hardware ecosystems already control buyer relationships and procurement trust, whereas Panthalassa must prove remote operations, bandwidth, uptime, security, environmental compliance, and offshore serviceability. The key underwriting point is that an integrated node can be strategically distinctive and still lose if buyers prefer known cloud vendors, clearer SLAs, or cheaper grid-connected compute.[CP002, CP024, CP025, CP034, CP035, CP036]

3.4 Switching cost, multi-homing, and supply access

For target AI-compute buyers, switching cost appears workload-dependent rather than absolute. Batch inference, offline training support, and interruptible workloads can be routed across providers more easily than low-latency user-facing services, which supports multi-homing but limits lock-in. Panthalassa could create lock-in if it controls scarce clean power capacity, operates reliably in energy-dense wave regions, and exposes familiar cloud interfaces; it could also become a secondary capacity provider if latency, bandwidth, or trust constraints keep customers on land for primary workloads. Partner access is similarly promising but unproven. Super Micro Computer appearing in the investor list is directionally helpful for hardware supply, while Aikido and Starcloud show that NVIDIA-linked ecosystems are actively courting remote-compute concepts. None of that proves allocation, customer commitments, or pricing power, so partner diligence should focus on GPU procurement rights, service agreements, satellite bandwidth, insurance, and maintenance logistics.[CP017, CP029, CP034, CP035, CP036, CP037]

3.5 Moat durability and adverse competitor evidence

The moat is real in concept but unproven in durability. Panthalassa owns a distinctive vertical-integration story: mass-produced steel nodes, autonomous station-keeping, wave generation, onboard compute, and satellite backhaul. The risk register is dominated by evidence that marine systems and offshore compute have historically struggled to become economic platforms. Project Natick showed strong technical results yet was retired; New Scientist highlights harsh salt and wave conditions and the need to beat conventional data centers on economics; and historical ocean-energy experiments such as OTEC demonstrate that weather and waves can destroy infrastructure before net power is achieved. Commoditization risk also sits on both sides of the product: wave-generation know-how can diffuse among marine-energy peers, while compute buyers can multi-home across hyperscalers and neoclouds. Until Ocean-3 and early customer contracts prove uptime, unit cost, and regulatory path, Panthalassa should be treated as competitively differentiated but not yet moat-protected.[CP023, CP024, CP030, CP038, CP039, CP040]

3.6 Exhibits

4.1 Revenue model: compute first, fuels optional, revenue absent

Panthalassa's financial story starts with a clean distinction: it is not publicly selling electricity to shore. The company says its autonomous nodes generate wave power offshore and consume that power onboard to run AI chips, returning inference tokens to land by satellite. That makes the near-term revenue hypothesis a compute-capacity service, not a conventional power project with a grid offtake agreement. Investor and company materials still leave room for clean fuels or hydrogen as future uses for abundant ocean power, but the 2026 financing narrative and Ocean-3 milestone are centered on AI inference at sea. No retained source discloses recognized revenue, ARR, pricing per token, GPU hour, reservation, or clean-fuel output. The right diligence posture is therefore to model revenue as zero today and treat every monetization mechanism as conditional on Ocean-3 proof plus first commercial contracts.[CI001, CI002, CI003, CI004, CI005, CI006]

4.2 GTM proxies are demand-side and milestone-led, not sales-efficiency proof

The go-to-market motion visible from public evidence is infrastructure-led. Panthalassa must first demonstrate that Ocean-3 can produce useful inference at sea, then convert that proof into contracts with hyperscalers, AI labs, neoclouds, or enterprise compute buyers that can tolerate satellite-linked workloads. The demand-side argument is clear: land data centers face grid, cooling-water, permitting, and community constraints, while Panthalassa proposes to move the energy and cooling problem offshore. Yet sales efficiency cannot be measured. There is no public CAC, payback, sales cycle, channel margin, pipeline conversion, backlog, or customer concentration. The strongest traction proxy is not revenue but investor conviction and deployment progress: a large Thiel-led Series B, a broad strategic syndicate, and a stated 2026 Ocean-3 pilot before commercial deployments targeted for 2027.[CI016, CI017, CI018, CI019, CI029, CI031]

4.3 Unit economics depend on aspirational power cost and unpriced offshore operations

The public unit-economics case is powerful but still aspirational. Lowercarbon repeats a roughly $0.02/kWh target, Gigascale cites around $1,500/kW manufacturing capex, a roughly 90% capacity-factor claim, and a scenario in which a $1 billion factory produces about 1 GW of node capacity annually. Those numbers would be highly attractive if realized, but they are investor- and management-sourced targets, not audited operating results from a commercial fleet. The cost stack is broader than energy conversion: plate-steel marine structures, coatings, turbines, power electronics, GPUs, sealed compute containers, satellite links, towing, maintenance vessels, insurance, corrosion management, biofouling, and replacement cycles all matter. DataDeep's adverse assessment directly challenges whether offshore operations and maintenance allow the headline two-cent power case to survive real deployment.[CI007, CI008, CI009, CI010, CI011, CI012]

4.4 Capital adequacy: large round, larger proof burden

Capital access is Panthalassa's clearest financial strength. The company announced a $140 million Series B on May 4, 2026, led by Peter Thiel, and GeekWire reported about $210 million in total capital raised. The round funds completion of a pilot manufacturing facility near Portland and accelerates Ocean-3 deployment, but it does not disclose cash on hand, monthly burn, runway, debt, project finance, or expected capex through commercial scale. Two SEC Form D filings add useful color but not full-round accounting: one Series B allocation vehicle reported $319,000 sold to 14 investors, and a later B Plus vehicle reported $1,935,313 sold to 31 investors. These appear to be partial SPV allocation filings, not the headline round. Financing dependency remains high until Ocean-3 produces auditable operating data or first commercial contracts.[CI017, CI019, CI020, CI021, CI022, CI023]

4.5 Public financial gaps dominate diligence

The public record is sparse on core operating metrics. Revenue, ARR, gross margin, contribution margin, working capital, CAC, sales payback, cash balance, burn, runway, customer contracts, utilization, active users, and project-finance terms are all absent. That absence is not surprising for a private, pre-revenue frontier-infrastructure company, but it prevents normal software, energy, or data-center underwriting. The correct treatment is to preserve nulls rather than backfill with implied values from funding announcements or investor quotes. The financial diligence request list should therefore start with management accounts, bank balance and runway, customer pipeline and contract drafts, node bill of materials, maintenance budgets, insurance quotes, GPU procurement and depreciation assumptions, and Ocean-3 telemetry protocols that separate gross power, parasitic load, usable compute, and uptime.[CI001, CI005, CI006, CI013, CI015, CI018]

4.6 Financial verdict: optionality with no revenue quality yet

Financially, Panthalassa is a high-optionality infrastructure bet rather than a company with underwritable revenue quality. The positive case is credible capital access, a clear demand bottleneck, a differentiated model that consumes power at sea, and large investor confidence behind a near-term pilot. The negative case is equally direct: there is no public revenue, no customers, no gross margin, no CAC, no burn or runway, no cash balance, and no audited data proving that marine maintenance and compute operations preserve the modeled unit economics. Revenue quality is therefore not applicable; margin path is unproven; capital intensity is very high. The investable trigger is not another press quote but measured Ocean-3 output, uptime, maintenance cadence, and first commercial contract economics. Until then, the verdict is watch-and-verify with major diligence blockers.[CI034, CI035, CI036, CI037, CI038, CI039]

4.7 Exhibits

5.1 Product definition and buyer workflow

Panthalassa is not selling a conventional wave-energy plant, a power-purchase agreement, or an offshore data-center shell. The product is an autonomous floating node that converts ocean-wave motion into onboard electricity and uses that electricity immediately to run AI chips at sea. In customer workflow terms, a buyer would route suitable AI jobs to Panthalassa capacity, the node would execute inference or longer-running compute workloads inside a sealed marine compute module, and results would return to land through satellite backhaul. The company’s explicit claim is that power is never exported to shore, so the customer benefit is not cheaper grid electricity but compute that avoids land, grid-interconnection, cooling-water, and permitting bottlenecks. That makes workload selection central: delayed-result inference, simulations, and other batch-like jobs fit better than latency-sensitive chatbot or search traffic. The product therefore has to be diligenced as a coupled service workflow spanning offshore generation, remote compute operations, satellite networking, and customer job orchestration, not as a stand-alone marine generator.[CE001, CE002, CE003, CE013, CE016, CE018]

5.2 Node assets and operating architecture

The public architecture is unusually specific for an early-stage hard-tech company. Ocean-3 is described as an approximately 85 meter plate-steel structure shaped like a golf ball on a tee, with investor coverage describing a roughly 50 meter spherical top and a long neck below the surface. As waves lift and lower the structure, seawater is forced up the central tube into an internal reservoir and then drained through a single turbine to produce electricity. That electricity powers hermetically sealed AI servers cooled through the surrounding seawater. Ocean-2 is the smaller prototype analogue, with a roughly 9 meter spherical top and public reporting of up to about 50 kilowatts in Puget Sound conditions. The architecture map should therefore be read as one vertically integrated asset: hull, wave pump, reservoir, turbine, power electronics, sealed compute, seawater heat exchange, autonomy, and satellite communications. The design may be elegant because it reduces moving parts and avoids shore cables, but commercial proof still depends on survivability, maintenance, and measured output in open-ocean conditions.[CE004, CE005, CE006, CE007, CE008, CE010]

5.3 Deployment maturity, integration, and roadmap

Panthalassa’s public maturity should be scored as prototype-to-pilot rather than commercial. The sequence runs from Ocean-1 in 2021, through Ocean-2 and Wavehopper prototype activity in 2024, a reported 2025 Puget Sound Ocean-2 test near 50 kilowatts, and a planned 2026 Ocean-3 pilot series in the northern Pacific intended to demonstrate AI inference and refine manufacturing. Commercial deployment is targeted around 2027, while the longer-run vision is thousands of nodes. Integration is not a simple installation motion: it requires coastal factory production, tow-out or launch logistics, autonomous station-keeping, remote monitoring, satellite connectivity, weather routing, repair procedures, GPU payload procurement, and customer workload scheduling. The roadmap claims are credible enough to watch because prototypes and funding exist, but not yet bankable as a service-level product because public sources do not disclose uptime, repair intervals, fleet telemetry, environmental permits, or customer acceptance tests. The right underwriting posture is low-to-mid TRL with a major Ocean-3 evidence checkpoint.[CE019, CE020, CE021, CE022, CE023, CE024]

5.4 Differentiation, moat, and critical dependencies

The strongest differentiation is the fusion of two hard systems that are usually separate: wave-energy generation and AI compute consumption on the same autonomous hull. Conventional wave projects usually try to export electricity; conventional data centers solve land, grid, cooling, and staffing constraints on shore; underwater data-center analogues such as Project Natick test sealed marine compute but not autonomous wave generation. Panthalassa’s moat, if it emerges, would sit in marine engineering know-how, hull and ballast design, manufacturing repetition, remote operations software, thermal integration, and the ability to operate where wave resources are energy dense. The dependency map is equally demanding. It relies on steel fabrication and coatings, turbine and power electronics reliability, GPU supply, sealed-container thermal design, Starlink or comparable LEO connectivity, ocean-weather operations, and permitting or environmental acceptance. Several metrics are aspirational: up to 90% power availability, low-cost energy, and a factory capable of gigawatt-scale annual node output. Those claims create upside but need instrumented fleet proof before they qualify as durable product advantage.[CE026, CE027, CE028, CE029, CE031, CE032]

5.5 Trust, quality, safety, and compliance gaps

Trust is the thinnest public part of the product story. A floating AI data center must survive corrosion, biofouling, storms, saltwater intrusion, thermal cycling, remote power faults, network interruptions, and abnormal incidents without on-site staff. Technical reference sources define corrosion and biofouling as ordinary marine hazards, while New Scientist highlights that physical intervention remains common in data-center incidents and that Panthalassa’s ocean sites make that intervention harder. The sealed seawater-cooled container may reduce water use and cooling equipment, but public sources do not provide independent PUE, chip-temperature, coolant-loop, corrosion-coating, anti-fouling, fire-suppression, storm-survivability, or mean-time-between-service data. Security and privacy are also unresolved: sources describe satellite backhaul but not encryption, tenant isolation, key management, incident response, or compliance attestations. Environmental trust is similarly incomplete because waste heat is expected to disperse into seawater, yet nearby ecosystem effects remain unclear. The quality-control table should be read less as a checklist passed today and more as the diligence agenda for Ocean-3 and first commercial contracting.[CE035, CE036, CE039, CE040, CE041, CE042]

5.6 Exhibits

6.1 Target customers are demand-side proxies, not won accounts

Panthalassa has no public customer base to segment in the ordinary sense. The correct customer lens is therefore prospective: hyperscale cloud platforms, large AI infrastructure buyers, neocloud or merchant GPU-cloud operators, and AI labs that need power-dense, delay-tolerant compute. The customer-proof sources allocated to this chapter intentionally function as demand-side proxies. Deloitte, IEA-linked data-center demand evidence, and hyperscale reference sources show that AI workloads are straining power, cooling, land, and grid access; they do not show that Panthalassa has converted AWS, Microsoft Azure, Google, a neocloud, or an AI lab into a customer. The implied buyer is infrastructure or cloud procurement, the economic sponsor is a cloud or AI-platform executive trying to secure power and compute capacity, and the user is an AI research or product team willing to send batch or inference jobs offshore. The payer would likely buy contracted capacity, not software seats, but public sources do not disclose revenue bands, price cards, channel motion, or signed capacity commitments.[CU001, CU002, CU003, CU004, CU005, CU009]

6.2 Adoption trajectory stops at product pilots

The public adoption trajectory is a technology trajectory rather than a customer trajectory. Panthalassa has disclosed prototypes such as Ocean-1, Ocean-2, and Wavehopper, a 2026 Ocean-3 pilot series, and a 2027 commercial-deployment target. Those milestones matter because a credible Ocean-3 deployment is the prerequisite to any design-partner conversation, but they are not evidence of active usage, paid accounts, repeat purchases, production customer deployments, or utilization. CBS and the company describe a future in which multiple systems work together as a data center, yet the current public record still lacks signed customers, LOIs, customer pilots, deployed customer locations, or contracted capacity. The funnel is therefore unusually steep: broad demand is visible, target workloads can be described, and product pilots are planned, but every commercial proof step remains at zero in public evidence. This makes the customer chapter adverse despite the attractive market backdrop.[CU006, CU008, CU018, CU019, CU020, CU021]

6.3 Named customer proof is absent

The named-customer proof ledger is empty. The table deliberately lists prospective design-partner segments so readers can see who should be tested first, but each row is labeled as a target rather than a won customer. This distinction is central: AWS and Microsoft Azure are relevant because they represent hyperscale cloud buyers with enormous compute and power needs; colocation and AI-cloud operators are relevant because they buy power, cooling, and dense infrastructure; AI labs are relevant because some workloads can tolerate distance from end users. None of those facts is a logo win, customer quote, pilot announcement, case study, procurement record, or referenceable deployment. The best available proof is demand pressure plus product-market logic; the missing proof is a customer saying Panthalassa ran useful workloads with measurable performance, reliability, and economics. Until that exists, customer proof quality is low even though target-buyer urgency is credible.[CU007, CU008, CU022, CU024, CU026, CU030]

6.4 Retention, repeat usage, and satisfaction are not measurable yet

Retention metrics should not be backfilled from market demand. With no disclosed customers, Panthalassa has no public NRR, GRR, logo churn, customer cohort, renewal term, support satisfaction score, or repeat-usage history. The right representation is null metrics plus explicit diligence asks. A future buyer will need evidence that offshore nodes can produce usable compute over months, meet service-level expectations, recover from failures without routine human access, and integrate into procurement and security workflows. Uptime Institute and New Scientist make that retention bar harder by emphasizing power, networking, remote operations, and latency as reliability constraints. In practice, the first retention test will not be a renewal cohort; it will be whether an Ocean-3 or early commercial node can run a buyer-relevant workload long enough to justify a second deployment. Until then, the cohort figure is intentionally blank rather than optimistic.[CU014, CU015, CU023, CU024, CU031, CU035]

6.5 Expansion upside comes with extreme concentration and procurement risk

If Panthalassa works, expansion could be powerful: the same buyer could add nodes or fleets as a capacity product, and demand-side evidence suggests hyperscalers and AI clouds are hunting for power-constrained compute. The commercial risk is that this upside concentrates future revenue in a handful of very sophisticated buyers with long diligence cycles, high reliability thresholds, security reviews, and leverage over price. Procurement friction is higher than for ordinary cloud capacity because the infrastructure is novel, remote, satellite-linked, and exposed to marine hazards. Super Micro Computer should not be misread as a customer; it is disclosed as an investor and has relevant AI/GPU server products, making it a possible supplier or partner dependency. The chapter's base conclusion is therefore not that demand is absent, but that demand has not yet converted into adoption. The next diligence gate is signed design-partner evidence tied to workload class, uptime target, data path, economics, and expansion rights.[CU016, CU017, CU025, CU026, CU030, CU031]

6.6 Exhibits

7.1 Severity ranking and residual exposure

Panthalassa's risk stack is severity-skewed toward technical and operational survivability rather than ordinary software execution. The highest-ranked exposure is whether an 85-meter steel, autonomous, compute-bearing node can survive open-ocean corrosion, biofouling, wave loading, storm damage, and station-keeping demands while also maintaining power, cooling, networking, and GPU reliability. The company has meaningful design mitigations: simple steel structures, few exposed moving parts, no subsea export cable, no anchor, no engine, and staged prototypes leading to Ocean-3. Those mitigations are real but immature because the investment case depends on compute-carrying nodes operating through months of sea states, not just prior energy trials. The practical implication is that the company should be underwritten as a pilot-gated frontier infrastructure bet. A failed Ocean-3 survivability or maintenance profile would transmit directly into lower utilization, higher service-vessel costs, delayed revenue, and a much weaker valuation case.[CR001, CR002, CR003, CR004, CR005, CR006]

7.2 Regulatory, legal, environmental, and policy exposure

The regulatory risk is not a single permit but an overlapping offshore-energy, maritime, data-center, environmental, and policy system. BOEM oversees renewable energy on the U.S. Outer Continental Shelf, while DOE's marine-energy program frames wave energy as a sector still needing R&D, demonstration, and barrier reduction. Data-center legal sources add another layer: NEPA, ESA, Clean Water Act, Clean Air Act, water, waste, stormwater, zoning, noise, traffic, and public-land approvals can all become relevant depending on siting and design. Federal acceleration for qualifying data centers may help some projects, but legal commentary emphasizes that state and local review, community resistance, and litigation remain schedule and cost risks. The Jones Act adds a practical maritime constraint if U.S. domestic towing, maintenance, or spares movement occurs between U.S. ports. Environmental review is especially sensitive because New Scientist reports that offshore waste-heat ecosystem effects remain unclear.[CR014, CR015, CR016, CR017, CR018, CR019]

7.3 Partner, platform, supplier, and customer dependencies

The partner-dependency profile is concentrated in places that can become binary blockers. Communications depend on Starlink or similar LEO backhaul, and the evidence indicates that latency and bandwidth shape the workload mix, leaving the near-term product better suited to delay-tolerant inference or batch jobs than low-latency consumer applications. Hardware exposure is also material: Panthalassa needs qualified GPUs, servers, power electronics, sealed compute containers, and reliable payload swaps in a marine environment. Super Micro is named as an investor, but retained sources do not convert that into a committed supply agreement. The company also depends on regulators to approve or tolerate deployment zones, capital providers to finance manufacturing before revenue, and future hyperscale or AI-lab customers to convert pilots into revenue. Because no paying customers are disclosed, early revenue could be concentrated in one or two large buyers if the pilot succeeds.[CR011, CR012, CR013, CR025, CR026, CR027]

7.4 Financial, model, and unit-economics risk

Financial risk is tightly coupled to survivability. The public funding facts are strong: Panthalassa announced a $140 million Series B in May 2026 and about $210 million total raised. But that capital is tasked with completing pilot manufacturing, deploying Ocean-3, proving at-sea inference, and reducing manufacturing and maintenance uncertainty before the company has disclosed revenue or paying customers. Management- and investor-side materials cite attractive targets, including about $1,500 per kilowatt capex, high capacity factor, low-cost power, and a factory scaling to roughly one gigawatt per year. The adverse view is that corrosion, biofouling, insurance, offshore maintenance, vessel days, and GPU-service logistics can erase those modeled gains. Even if wave energy is physically abundant, the underwriting question is delivered compute cost after parasitic loads and service interventions, not gross wave resource.[CR027, CR028, CR029, CR030, CR031, CR032]

7.5 People, execution mitigations, and thesis-break triggers

Execution risk is partly mitigated by a specialist team drawn from aerospace, naval architecture, software, manufacturing, military, and research organizations, and by a deep syndicate that can finance additional iteration. It remains material because public sources center the company around two co-founders, offer limited governance or board disclosure, and leave many detailed operating controls private. Panthalassa also did not respond to New Scientist's skeptical technical questions before publication, which increases the value of independent diligence rather than press-level assurance. The monitoring plan should therefore be explicit: require independent Ocean-3 telemetry, net output after parasitic loads, Starlink availability and latency logs, corrosion coupons or inspection reports, biofouling and cooling-degradation data, service-vessel days per node, safety incidents, permitting milestones, signed customer conversions, and updated unit-economics bridges. The thesis breaks if the pilot fails survivability, service cadence, customer conversion, regulatory clearance, or cost targets.[CR034, CR035, CR036, CR037, CR038, CR039]

7.6 Exhibits

8.1 Recommendation and valuation stance

The investment call is research-more / track, not buy. Panthalassa has the shape of a venture-style option on a real bottleneck: AI compute is power constrained, and offshore wave-powered infrastructure could be valuable if it avoids grid and land bottlenecks. The current public record is not strong enough to underwrite a new-money entry at the implied near-unicorn mark. The company has a $140 million Thiel-led Series B and about $210 million total raised, but the exact post-money valuation, preferences, and fully diluted ownership are not disclosed. It is also pre-revenue, has no publicly disclosed customers, and has not yet shown independent commercial-scale Ocean-3 performance. That makes the price discipline unknown-to-stretched and the risk rating high even though the upside option is meaningful.[CV001, CV002, CV003, CV009, CV034, CV035]

8.2 Thesis and anti-thesis

The thesis is not a generic wave-energy story; it is a co-location argument. Panthalassa wants to consume power at sea, sell compute, and avoid the transmission, land, cooling, and grid interconnection problems that slow land data centers. If the company can really deliver low-cost, high-availability energy and package it into useful inference capacity, it could create infrastructure value outside the grid. The anti-thesis is equally direct. Wave power is a small market with difficult economics, the ocean is corrosive and expensive to service, LEO latency limits workloads, and land-based AI-cloud incumbents have scale, procurement, software, and customer advantages. Because investor-backed economics are not customer proof, the thesis needs field data and pipeline evidence before it becomes a priced investment case.[CV010, CV013, CV014, CV015, CV016, CV018]

8.3 Financing context and entry discipline

The financing context argues for discipline rather than false precision. Public articles supply the headline round and near-$1 billion framing, while SEC Form D filings show smaller related SPV allocations; none reveal the full cap table, post-money share count, liquidation preference, participation, pro-rata rights, or option-pool treatment. Methodologically, post-money valuation is only meaningful when the invested capital and fully diluted capital structure are known. For Panthalassa, the exact price cannot be validated from public evidence. A rational entry stance is therefore milestone-gated: treat the current mark as an option price on a difficult infrastructure breakthrough, not as a validated DCF or revenue multiple. The next financing should be assessed after Ocean-3 data and credible customer commitments reduce the main uncertainty bands.[CV003, CV004, CV005, CV006, CV007, CV008]

8.4 Bull, base, and bear scenarios

Scenario analysis produces a wide range because a single technical proof point can change the company’s risk class. In the bull case, Panthalassa demonstrates useful at-sea inference, proves survivability, and shows that its $0.02/kWh and high-availability targets can survive real maintenance, insurance, and logistics costs. That could justify multi-billion-dollar infrastructure optionality. In the base case, the pilot works but scale is slow and capital hungry, so the implied roughly $1 billion mark merely holds while investors wait for orders and repeatable manufacturing. In the bear case, corrosion, biofouling, storms, latency, or wave-energy LCOE make the system uneconomic, and land-based alternatives win. The downside trigger is not lack of imagination; it is field economics failing to converge quickly enough.[CV013, CV014, CV015, CV016, CV018, CV020]

8.5 Comparable valuation context

The comparable set is useful only if clearly labeled as imperfect. Starcloud is the closest narrative analog because it reached a reported $1.1 billion valuation while pursuing non-terrestrial data centers, but space and ocean execution risks differ. CoreWeave is a demand-side AI-cloud reference, yet it is public and far more mature than Panthalassa. Crusoe is an AI-infrastructure analog, though public evidence in this allocation is thin for the exact valuation mark. CorPower Ocean and Eco Wave Power are relevant wave-energy peers, but they sell grid or coastal wave power rather than offshore compute. These references triangulate option value and risk appetite; they do not prove Panthalassa’s price. The most honest conclusion is that frontier infrastructure markets can award unicorn marks early, but commercial proof determines whether those marks compound or collapse.[CV028, CV029, CV030, CV031, CV032, CV033]

8.6 Exit readiness and final diligence asks

Panthalassa is not exit-ready today. A credible IPO would require revenue, customer concentration disclosure, repeatable unit economics, governance readiness, and a public-market narrative that converts ocean compute from science project to infrastructure platform. A strategic acquisition is plausible only after a buyer can diligence field performance and workload fit. The immediate diligence asks are therefore concrete: confirm post-money and terms, inspect Ocean-3 telemetry, validate customer pipeline, reconcile energy economics with maintenance and insurance, and review survivability data from storms and saltwater exposure. Thesis-break triggers should be explicit before any entry: failed Ocean-3 deployment, no customer commitments, economics materially above target, preference-heavy financing terms, or evidence that latency and maintenance limit the addressable workload pool.[CV024, CV025, CV026, CV027, CV034, CV035]

8.7 Exhibits