DriveNets

1.1 Identity, product, and operating model

DriveNets is consistently described in the assigned evidence as an Israeli business-software company headquartered in Ra'anana and founded in 2015 by Ido Susan and Hillel Kobrinsky. Its core Network Cloud product separates networking software from standard white-box hardware, using a cloud-native architecture for communications service providers and cloud operators. The company-issued solution description says the platform provides shared compute and networking infrastructure, supports services from core to edge, and separates its control and data planes. The same material describes DNOS as containerized, distributed networking software and DNOR as the orchestration layer. The assigned evidence also indicates a later Network Cloud-AI direction for Ethernet-based AI fabrics. These are product and company descriptions, not independent proof of performance, cost savings, or customer-wide deployment.[CO001, CO002, CO003]

1.2 Founders, leadership, and key-person dependence

Public profiles identify Susan as co-founder and CEO and Kobrinsky as co-founder and chief strategy officer. The founders' prior exits are part of the company's external narrative: the assigned AIWiki profile says Susan previously co-founded Intucell, acquired by Cisco for $475 million, while Kobrinsky founded Interwise, acquired by AT&T for $121 million. Startup Nation Central's profile also lists CFO Mike Lilo, CMO Inbar Lasser-Raab, and VP R&D Tali Izhar, providing evidence of functional coverage beyond the founders. No assigned source supplies a board roster, ownership register, executive tenure history, succession plan, or a verified leadership change during the current refresh period. The combination of founder-led positioning and missing governance detail makes founder dependence a diligence item rather than a quantified conclusion.[CO005, CO006, CO007, CO008]

1.3 Financing, valuation, and disclosed scale

The current financing evidence agrees that DriveNets completed a $410 million Series D in June 2026, with Bessemer Venture Partners and Atreides Management leading and AMD, Red Dot Capital Partners, Pitango, and D1 Capital Partners participating or being named as investors. Valuation evidence is not fully aligned: Calcalist and Startup Nation Central report $8.5 billion, 650 Group states $8 billion, and Reuters says the company did not disclose the round valuation. Current sources place total primary capital at approximately $1 billion, while the 2025 AT&T transaction is described as a $650 million secondary purchase from existing holders rather than primary company financing. Public scale indicators conflict: assigned profiles range from roughly 350 to 643 employees, and one profile reports 201–500 employees. The company-related sources state cash-flow positivity and more than $1 billion in orders or project backlog, but these are not audited revenue, ARR, margin, or cash-available disclosures. Valuation therefore has a clear financing-event anchor but neither a fully consistent reported figure nor a public operating denominator.[CO004, CO009, CO010, CO011, CO012, CO013, CO016]

1.4 Milestones, strategic relationships, and unresolved diligence

The evidence supports a milestone arc from the 2015 founding, through early financing and the September 2020 AT&T core-network deployment, to the 2023 Network Cloud-AI announcement, the 2025 AT&T secondary transaction, and the 2026 Series D. The company-issued product material quotes an AT&T network-services executive saying more than 50% of AT&T traffic was running on an open disaggregated architecture where the network operating system was DriveNets; this is customer-attributed and should not be treated as independently audited market share. The 650 Group article describes the strategic logic of AMD reference designs and heterogeneous AI infrastructure, while the assigned sources do not independently establish broad customer counts, current revenue, gross margin, or the full investor/control structure. The practical company overview is consequently positive on product identity and financing momentum, but conditional on verifying commercial conversion, customer concentration, governance, and the distinction between backlog and recognized revenue.[CO003, CO014, CO015, CO016]

1.5 Exhibits

2.1 Market boundary and status-quo substitutes

The defensible boundary is narrower than total networking or AI-infrastructure spend. For carrier and cloud networks, it includes network operating software, disaggregated routing and fabric hardware, orchestration, integration and support needed to run core, aggregation, edge and peering workloads. For AI infrastructure, it includes the back-end Ethernet fabric, compatible switching systems and the software that schedules traffic among accelerators. It excludes GPUs and other accelerators, model software, storage, general data-center construction and unrelated enterprise LAN equipment. The status quo is also different by segment: carrier buyers can retain monolithic, vendor-locked routers, while AI buyers can retain InfiniBand or another vertically integrated GPU-network stack. Telecom Infra Project requirements show that disaggregation is an industry architecture rather than a DriveNets-only category, while the APNIC article describes both single-chassis and leaf-spine choices for AI fabrics. DriveNets therefore competes for specific network-control and fabric budgets, not every dollar attached to an operator or AI cluster.[CM001, CM002, CM003, CM004, CM005]

2.2 Evidence-constrained sizing and demand signals

The assigned evidence does not support a conventional top-down TAM/SAM/SOM stack with common geography, period and spend denominator. A company-issued release quotes 650 Group forecasting that AI networking will surpass $200 billion by the end of the decade, but the release supplies neither the market definition nor methodology needed to translate that number into DriveNets' addressable revenue. More reliable boundary checks come from operating evidence. Telecom Infra Project specifies disaggregated aggregation routers from 0.3-1.0 Tbps at the XSmall tier through 12 Tbps or more at Large, showing that deployments span materially different capacity and budget classes. Telstra International separately reported an incremental 200 Tbs and a 30% increase in capacity across key Asia-Pacific routes in October 2025. The Fast Mode reported that more than 52% of AT&T's North American core traffic was already transported by DriveNets-powered infrastructure. Those data points support demand and production adoption, but they are capacity and penetration signals, not interchangeable market values.[CM006, CM007, CM008, CM009, CM016]

2.3 Buyers, users, payers and adoption path

Communications service providers and cloud providers form the established buyer group. Network architecture and procurement leaders sponsor the decision, engineering and operations teams use the system, and core, aggregation or transport budgets pay for capacity and lifecycle support. The AI-fabric buyer set is newer and includes foundation-model labs, hyperscalers, NeoClouds and large enterprises, according to the company release; cluster, network and performance engineers are the users, while the AI-infrastructure owner controls the budget. Across both segments, adoption is likely to begin with design and interoperability validation, proceed through a controlled production deployment, and expand only after reliability and operational responsibility are proven. The economic trigger differs by segment. Carrier buyers seek incremental scaling, faster provisioning and less proprietary lock-in, whereas AI buyers seek higher accelerator utilization, resilient scale and a multi-vendor alternative. Public sources do not disclose standardized pricing, buyer conversion rates or the software share of total deployment spend.[CM010, CM011, CM012, CM013]

2.4 Growth drivers, adoption constraints and valuation relevance

Capacity growth, network cloudification and the cost of idle accelerators create a credible demand backdrop. Telstra's shorter provisioning cycle illustrates why operators value software-controlled capacity, and AT&T's reported production penetration shows that disaggregated routing can move beyond a pilot. The counterweight is execution. DriveNets' own operations FAQ acknowledges the perceived risk of a multi-vendor blame game and the need for clear end-to-end responsibility. The APNIC article, written by an Arista engineer, explains that large leaf-spine AI fabrics require advanced congestion and flow-control mechanisms; it is useful technical context, not an independent competitive benchmark. These constraints can lengthen certification, migration and support cycles even when the architecture is attractive. For valuation, the key issue is conversion rather than theoretical market breadth: investors need segment revenue, gross margin, pricing, win rates, backlog conversion and customer concentration before treating the broad AI-networking forecast as DriveNets' obtainable market.[CM007, CM009, CM013, CM014, CM015, CM016]

2.5 Exhibits

3.1 Competitive landscape and buyer alternatives

DriveNets addresses two overlapping buying jobs. In service-provider routing, Network Cloud is a cloud-native network operating system running across white-box hardware, positioned as an alternative to monolithic integrated routers. In AI infrastructure, AI Fabric is an Ethernet-based fabric for multi-vendor accelerator clusters and is positioned against Nvidia's InfiniBand and AI-Ethernet systems, as well as merchant-switch offerings from Cisco and Arista. Independent coverage describes the field as crowded: telecom incumbents include Cisco, Juniper, and Nokia, while AI back-end alternatives include Nvidia, Arista, and Cisco. The practical status quo remains an integrated incumbent stack or an internal build assembled from merchant silicon, switching hardware, NICs, and orchestration. That status quo can win on procurement simplicity, installed-base familiarity, and single-vendor accountability even when DriveNets offers lower lock-in. Public sources identify the alternatives and architecture, but do not provide a comparable competitor-by-competitor share or realized price dataset.[CP001, CP002, CP003, CP004]

3.2 Capability, distribution, and trust comparison

DriveNets' claimed capability advantage is architectural rather than a published price advantage. Its Network Cloud separates control-plane software from a data-plane cluster, supports multiple white-box vendors, and exposes open management interfaces. The company FAQ acknowledges a material tradeoff of disaggregation: multi-vendor support can create a root-cause and certification "blame game"; it answers with either an end-to-end responsibility model or a shared-responsibility model. This is a meaningful trust and operations issue, not a cosmetic caveat. On distribution, the Dell AI Factory relationship places DriveNets inside a recognized infrastructure channel, while AMD and Broadcom relationships support heterogeneous accelerator and switching choices. Those are ecosystem signals, not proof of independent customer preference or repeatable channel economics. The capability matrix therefore marks unsupported cells as unknown rather than inferring that every competitor lacks an equivalent feature.[CP005, CP006, CP007, CP008]

3.3 Switching costs, moat durability, and displacement risk

DriveNets has credible entry barriers from carrier-grade deployment experience, operational tooling, and integration knowledge. Its FAQ says the architecture has been deployed at multiple service providers and carriers, and that legacy equipment interoperability is necessary because the company enters environments with incumbent vendors. Independent reporting also links the company to AT&T's core routing deployment and describes its AI pivot. These references support a field-learning and reference-account moat, but not a quantified switching-cost moat. Buyers can multi-home with a second or third network vendor, and incumbent vendors can respond with their own Ethernet, AI, or disaggregated offers. The same standards and merchant silicon that make DriveNets' proposition open can lower hardware lock-in for buyers and lower differentiation durability over time. The largest displacement risk is that Nvidia's integrated ecosystem, or an incumbent's support and procurement bundle, wins despite DriveNets' openness. A second risk is execution: the public record does not disclose pricing, gross margin, deployment conversion, or competitor win rates.[CP009, CP010, CP011, CP012]

3.4 Evidence-backed positioning

The positioning map uses ordinal coordinates rather than pretending that public sources provide a common benchmark. The horizontal axis reflects architectural openness and multi-vendor flexibility; the vertical axis reflects disclosed evidence of carrier or AI-scale deployment maturity. DriveNets scores high on both dimensions because its sources describe white-box disaggregation, carrier deployments, and AI-fabric activity, but the score is an analyst synthesis rather than a reported market ranking. Nvidia scores lower on openness because the cited competitive description characterizes InfiniBand as proprietary, while its deployment maturity is high. Cisco, Juniper, Nokia, and Arista are placed conservatively at unknown-to-midpoint positions where this assigned pool lacks comparable evidence. This makes the map useful for diligence framing, not a claim of product superiority. The priority diligence questions are realized pricing, referenceable production deployments, competitive win/loss data, and the extent to which open standards narrow DriveNets' differentiation.[CP013, CP014, CP015]

3.5 Exhibits

4.1 Monetization and public traction

DriveNets sells enterprise infrastructure software and associated network solutions to service providers, cloud providers, and AI-infrastructure operators. The public materials describe Network Cloud as cloud-native software running on standard white-box hardware, while Network Cloud-AI extends the model to Ethernet fabric for large AI clusters. This supports a software-led, infrastructure-linked revenue model, but the sources do not disclose list price, contract duration, recurring-versus-services mix, recognized revenue, ARR, or gross margin. The strongest traction disclosure is company-issued: the June 2026 financing release says the company had more than $1B in secured business and had been cash-flow positive since 2025. That is an order/backlog and cash-flow assertion, not revenue or cash-on-hand evidence. Independent reporting confirms the $410M round and notes that valuation was not disclosed to Reuters, reinforcing the need to separate financing facts from operating KPIs.[CI001, CI002, CI003, CI004, CI005]

4.2 Pricing, sales motion, and efficiency proxies

The buying motion appears to be high-value, technical enterprise selling rather than self-serve software. DriveNets' public surfaces describe production deployments with tier-one operators, partnerships with hardware and systems vendors, and a pipeline tied to large AI fabrics. This can support sizeable contracts and ecosystem distribution, but it also implies long qualification, integration, and deployment cycles. No retained source supplies realized price per router, fabric, port, GPU, site, or contract; no source supplies CAC, payback, renewal, expansion, implementation margin, or sales-cycle length. The 2021 Heavy Reading survey, sponsored by DriveNets, reported cost and openness as operator pain points and described adoption of disaggregated networks, but it is demand context rather than DriveNets' own conversion or pricing data. Underwriting should treat pricing and efficiency as private-data gaps, not infer them from the financing size or from official product positioning.[CI006, CI007, CI008, CI009]

4.3 Cost structure and capital adequacy

The disclosed cost profile is consistent with a technically intensive infrastructure vendor: the company is scaling inventory for its AI-fabric pipeline, while its products require engineering, deployment, support, and ecosystem work across software and white-box hardware. Revelio Labs reports 643 employees as of March 2026 and identifies engineering, finance and operations, and sales and marketing as material workforce groups; this is a workforce proxy, not a payroll or cost statement. The company says the new financing will scale inventory and expand heterogeneous-AI infrastructure, so working capital and supply commitments may rise even if software gross margin is attractive. Capital adequacy cannot be tested from public evidence: cash on hand, monthly burn, runway, debt, purchase commitments, inventory turns, and next-round triggers are not disclosed in the retained sources. The $410M Series D and reported cash-flow-positive status reduce near-term financing dependency as a public hypothesis, but neither proves liquidity nor eliminates execution and dilution risk.[CI010, CI011, CI012, CI013, CI014]

4.4 Financial underwriting verdict

Financial quality is promising but not yet underwritable at the level required for a valuation conclusion. Positive signals include a large financing round, strategic participation from AMD, reported secured business above $1B, and a stated cash-flow positive position since 2025. Counterweights are equally material: the evidence does not establish how much secured business is backlog versus recognized revenue, how quickly it converts, what portion is hardware or services, whether customer concentration is high, or whether inventory investment will absorb cash. Independent reporting explicitly records that DriveNets did not disclose the valuation to Reuters, while third-party databases and company-distributed materials report differing employee counts and funding-round presentations. The financial verdict is therefore track/research-more rather than a numeric return claim: request the revenue bridge, backlog aging, margin waterfall, cash reconciliation, debt schedule, customer concentration, and unit economics before relying on the $8.5B valuation.[CI003, CI004, CI005, CI012, CI015, CI016]

4.5 Exhibits

5.1 Product and customer workflow

DriveNets sells networking software rather than a fixed router appliance. Its Network Cloud model separates the network operating system from hardware and lets operators assemble a shared infrastructure from standard white boxes. The operating model is visible in the KDDI deployment: Network Cloud was used as an internet-gateway peering router after several years of lab testing, with additional applications planned. DriveNets also describes Network Cloud-AI as an Ethernet fabric for AI infrastructure, positioned around higher GPU utilization and faster deployment. These are distinct workflows: carrier routing and peering are a more mature proof surface than the newer AI-fabric use case, whose customer breadth and realized outcomes are not fully disclosed.[CE001, CE003, CE004, CE006, CE013]

5.2 Architecture, integration, and maturity

The technical proposition is a distributed, disaggregated chassis: DNOS is described as container- and microservice-based software that runs across white boxes from multiple vendors, while TIP's aggregation-router requirements show the surrounding ecosystem's expectations for open hardware, routing, telemetry, automation, and security. DriveNets has also demonstrated an integration path beyond basic packet forwarding: the Ciena collaboration covered configuration, monitoring, troubleshooting, and 400ZR/ZR+ coherent pluggables in a tested open-line-system environment. This suggests a product that can absorb optical and management dependencies, but public material does not establish uniform support across every hardware combination or quantify operational reliability at scale.[CE002, CE005, CE010, CE011, CE012]

5.3 Differentiation, trust, and diligence risks

DriveNets' differentiation is the combination of software/hardware disaggregation, shared infrastructure, and an Ethernet-based AI-fabric thesis that can span multiple accelerator and networking vendors. The company reports a 512-GPU benchmark advantage over InfiniBand, while third-party coverage repeats claims of fabrics designed for up to 32,000 GPUs; neither source is an independent production audit. The strongest public maturity evidence remains carrier deployment and interoperability work, not a broad, independently documented AI installed base. TIP's requirements provide a useful external checklist for management interfaces, provisioning, authentication, logging, and data-plane protections, but they are not proof that DriveNets complies with every requirement. Diligence should therefore focus on customer-level AI production references, failure and recovery data, supported hardware matrices, security attestations, and the boundary between current capability and roadmap.[CE007, CE008, CE009, CE010, CE011, CE012, CE014, CE015]

5.4 Exhibits

6.1 Customer base and segmentation

The public customer footprint splits into two commercially different groups. The established base is communications-service and broadband operators buying core, edge, and network-virtualization software; named examples include AT&T, Comcast, and KDDI. Independent secondary profiles report roughly 100 customers, but that rounded count has no disclosed definition, date-consistent denominator, or split between production, pilot, and evaluation accounts. The newer group comprises hyperscalers, neoclouds, enterprises, and other AI-infrastructure operators evaluating Ethernet fabric for GPU clusters. That segment is strategically attractive but public proof remains mostly company-described trials and proofs of concept rather than named production references. Buyers therefore appear concentrated among technically sophisticated, high-scale network operators, while users are network engineering and operations teams and the economic payer is the infrastructure owner.[CU001, CU002, CU006, CU007, CU008]

6.2 Production adoption and expansion

AT&T is the strongest public adoption proof. DriveNets says its software has provided AT&T's production core-routing layer since 2020 and carried more than 52% of North American core traffic by January 2023; a separate company profile repeats the traffic share, while an independent profile characterizes the deployment as production rather than pilot. Comcast broadens the evidence: Futuriom reported that DriveNets supported the initial Janus trial in Atlanta and would help expand the software-based network nationwide. Futuriom also reported that Janus is intended to support Comcast's 63 million locations, but that is Comcast's addressable network footprint, not a count of live DriveNets sites. KDDI is repeatedly named, although the retained evidence does not expose deployment scale or outcome metrics.[CU003, CU004, CU005, CU006, CU015]

6.3 Durability, expansion, and customer economics

The evidence indicates unusually deep strategic relationships but does not quantify durability. A secondary data provider says DriveNets sells software and systems under multi-year contracts, and Olam reports that AT&T later bought shares from existing holders; neither item substitutes for renewal, churn, gross-retention, or net-retention data. Futuriom described the Comcast opportunity as reportedly worth hundreds of millions of dollars, while two sources repeated DriveNets' statement that secured business exceeded $1 billion. Those figures signal commercial demand but are not recognized revenue, ARR, or cash and do not reveal customer concentration. Diligence should therefore reconcile signed backlog to deployment milestones, acceptance terms, cancellations, and cash collections, then obtain revenue concentration and cohort renewal schedules.[CU009, CU010, CU011, CU012, CU013, CU016]

6.4 AI customer proof and procurement friction

DriveNets' AI-customer evidence remains earlier-stage than its telecom record. The company said leading hyperscalers validated AI Fabric in trials and reported up to a 30% reduction in idle time, while TelecomTV described participation in proofs of concept in very large AI clusters. These claims establish evaluation activity, not named production adoption, recurring consumption, or realized customer savings. TelecomTV also identifies a material commercial hurdle: AI-factory operators must be persuaded to choose DriveNets' Ethernet fabric over Nvidia's proprietary InfiniBand. The resulting sales motion likely requires technical benchmarking, interoperability validation, and procurement confidence before production rollout. A named production AI customer, measured utilization outcome, and renewal or expansion record are therefore the most important missing customer proofs.[CU007, CU008, CU014]

6.5 Exhibits

7.1 Risk ranking and thesis-break triggers

The highest-consequence risk is execution during the AI-infrastructure pivot. DriveNets is extending a carrier networking platform into large AI clusters while supporting heterogeneous accelerators, NICs, ASICs, and white-box hardware. Futuriom reports that the company is using the new financing to scale inventory and broaden heterogeneous infrastructure, while the company says it has more than $1 billion in secured business and has been cash-flow positive since 2025; those statements are company-linked and should not be treated as audited operating proof. A thesis break would be a material failure to convert signed orders into accepted production deployments, a sustained inability to meet performance or reliability requirements, or a reversal of the reported cash-flow-positive position. Monitoring should focus on production references, deployment acceptance, backlog conversion, gross margin, working capital, and customer renewal evidence.[CR001, CR002, CR003, CR004, CR005]

7.2 Regulatory, legal, and security exposure

No assigned source documents a DriveNets-specific enforcement action, lawsuit, breach, certification failure, or safety incident, so absence of a recorded event is not evidence that the exposure is absent. The product is networking software deployed in communications and AI infrastructure, and the public materials describe cloud-native software, white-box hardware, telemetry, automation, and customer deployment work. That creates diligence exposure around secure development, privileged access, vulnerability response, data handling, export controls, contractual service levels, and allocation of responsibility between DriveNets, hardware suppliers, cloud operators, and customers. The FTC's business guidance states that companies holding sensitive information need a sound security plan and that security obligations can apply after a breach; this is a regulatory baseline, not evidence of DriveNets' compliance. The kill criterion is a confirmed material incident or an inability to produce current control evidence, incident history, and contractual allocation for a critical deployment.[CR006, CR007, CR008, CR009, CR010]

7.3 Dependencies, mitigations, and diligence priorities

DriveNets' architecture reduces dependence on a single proprietary networking stack, but it does not eliminate dependency risk. The product materials describe support for multiple vendors, while the AMD relationship and references to Broadcom, Dell, and other ecosystem participants show that the commercial proposition still depends on partner compatibility, supply availability, integration quality, and shared go-to-market execution. A current solution-engineering role calls for rack design, proof-of-concept execution, onsite deployment, troubleshooting, customer training, and work across multiple vendors; that is evidence of implementation complexity rather than proof of a failure. The immediate diligence path is to obtain a deployment-by-deployment dependency map, supplier and partner concentration, service-level and indemnity terms, security attestations, incident records, backlog-to-revenue conversion, and customer-level renewal or expansion data. Any one customer, chip, ODM, or strategic partner becoming operationally irreplaceable would raise residual exposure.[CR011, CR012, CR013, CR014, CR015]

7.4 Exhibits

8.1 Recommendation and entry discipline

The recommendation is research-more rather than buy. Independent reports consistently place the June 2026 financing at $410 million, while several sources estimate an $8.5 billion valuation; Reuters separately reported that DriveNets did not disclose the valuation or answer its request for that detail. That distinction matters: the round is verified, but the price is a third-party mark rather than a company-confirmed term. The retained evidence also lacks audited revenue, ARR, gross margin, retention, free cash flow, and preference terms, so a defensible sales multiple or fully diluted ownership return cannot be calculated. Entry should therefore be contingent on primary financial diligence, customer concentration evidence, and cap-table terms, with staged funding or valuation protection if those materials remain unavailable.[CV001, CV002, CV003, CV009, CV013, CV014]

8.2 Thesis, anti-thesis, and evidence quality

The thesis rests on three signals: a large financing led by specialist investors, management's claim of positive cash flow and more than $1 billion of orders and project backlog, and a product strategy aligned with open, multi-vendor AI infrastructure. A joint AMD reference architecture adds technical validation, although its benchmark results remain company- and partner-issued rather than an independent production audit. The anti-thesis is equally material. The reported mark rose roughly 3.4 times from the 2022 financing level, fresh capital is partly intended to support inventory, and no verified operating denominator explains the re-rating. AT&T's 2025 secondary purchase demonstrates shareholder liquidity and strategic interest, but it did not provide primary capital and does not substitute for recurring-revenue quality.[CV004, CV005, CV006, CV007, CV008, CV011, CV012]

8.3 Bull, base, and bear framework

Scenario valuation should remain conditional rather than falsely precise. A bull case can support the reported mark only if audited growth, software-like margins, durable backlog conversion, and diversified AI production customers become visible. The base case assumes strategically valuable technology but incomplete financial proof; it favors a discount to the reported mark, milestone-based funding, or downside protection until revenue quality and dilution are known. The bear case is triggered if inventory requirements expand faster than converted cash, backlog proves concentrated or cancellable, or open Ethernet adoption does not translate into repeatable deployments. These are decision rules, not probability estimates: the available sources do not support numeric scenario probabilities or a precise valuation range.[CV006, CV009, CV013, CV015, CV016]

8.4 Comparables, exit readiness, and final diligence

Public networking and infrastructure software companies offer useful business-model references, but the retained comparable set cannot be applied mechanically because DriveNets has no verified revenue denominator and spans software, integration, and inventory-backed deployment. The independent market outlook for AI networking is supportive, yet market size does not establish DriveNets' capture rate or exit multiple. The 2025 secondary transaction shows some private liquidity, not IPO readiness or a realizable return at the 2026 mark. Final diligence should request audited revenue and margin bridges, backlog aging and cancellation terms, top-customer concentration, deployment-level gross margin, inventory commitments, fully diluted capitalization, liquidation preferences, debt, and a board-approved exit plan. Failure to provide these materials should break the valuation thesis even if product evidence remains positive.[CV007, CV010, CV011, CV013, CV014, CV016]

8.5 Exhibits