FieldAI

1.1 Company Identity and Business Model

FieldAI is an Irvine, California-based robotics AI software company founded in 2023 by Ali Agha, a veteran of NASA's Jet Propulsion Laboratory (JPL). The company's central thesis is that the binding constraint on industrial robotics adoption is not hardware but intelligence: robots lack the adaptive, risk-aware autonomy needed to operate reliably in unstructured real-world environments. FieldAI addresses this by building Field Foundation Models (FFMs), a class of "physics-first" foundation models trained on sensor data from physical deployments rather than internet text or images. FFMs serve as a universal software brain that can be installed across heterogeneous robot platforms—quadrupeds, humanoids, wheeled rovers, tracked vehicles, and passenger-scale autonomous vehicles—without requiring per-platform reprogramming, pre-mapped environments, GPS, or cloud connectivity. The robots make decisions entirely on-edge in real time. FieldAI markets the platform to enterprise industrial customers in construction, energy, mining, manufacturing, logistics, urban delivery, inspection, and defense. The company explicitly does not manufacture robots; it is a pure-play AI software vendor that partners with robot OEMs and deploys software on customer-owned or third-party robot hardware. This hardware-agnostic model positions FieldAI as the "brain layer" of the robotics stack, analogous to how an OS or middleware layer abstracts hardware in enterprise computing. Bill Gates described the company simply: FieldAI "develops AI software for other companies' robots that enables them to perceive their environments, navigate without GPS, and even communicate with each other." Primary revenue channels are B2B software licensing and hardware-integration services, with subscriptions reportedly ranging from tens of thousands to $500,000 per year depending on deployment scale. As of April 2026, sources familiar with the matter indicate the company has more than $100 million in booked revenue and partnerships with large industrial customers. [CO001, CO002, CO003, CO004, CO005, CO006]

1.2 Founders, Leadership, and Governance

FieldAI was founded by Ali Agha, who serves as CEO, drawing on nearly two decades of robotics AI expertise. Before founding FieldAI, Agha spent approximately seven years at NASA's Jet Propulsion Laboratory (JPL) as Principal Investigator for some of the nation's highest-profile autonomy programs, including the DARPA Subterranean Challenge, DARPA RACER (self-driving off-road vehicles), NASA's Autonomous Mars Cave Exploration, and Coordinated Autonomy for Prototype Mars Helicopter-Rover. He also worked as a researcher at Qualcomm and held a postdoctoral position at MIT. Critically, Agha led the JPL-MIT-Caltech-KAIST-LTU CoSTAR team that won the Urban Circuit of the 2020 DARPA Subterranean Challenge—a landmark achievement in field robotics that validated the risk-aware autonomy approach now central to FieldAI's technology. Agha earned his PhD in Computer Science and Engineering from Texas A&M University. His co-founders include Dr. Shayegan Omidshafiei (Chief Science Officer), who spent over five years as a Research Scientist at DeepMind/Google focused on large-scale AI models and multi-agent reinforcement learning, and who completed his PhD at MIT where he first collaborated with Agha; and Dr. David Fan (Chief Technology Officer), who was chief technologist for multiple DARPA SubT and DARPA RACER programs and a NASA JPL research fellow. Additional senior leaders include Dr. Eric Krotkov, who heads FieldAI Federal, the company's defense subsidiary, and is a former DARPA Program Manager and CSO of Toyota Research Institute; and Sebastian Scherer, Director of Fieldable Embodied AI, an Associate Research Professor at Carnegie Mellon's Robotics Institute. The founding team's shared JPL, DARPA, and MIT lineage gives FieldAI exceptional founder-market fit for deploying AI in safety-critical field environments. Board composition and formal governance structure have not been publicly disclosed. CEO Ali Agha is the company's primary public spokesperson and fundraiser, constituting a meaningful key-person dependency. [CO011, CO012, CO013, CO014, CO015, CO016]

1.3 Funding History and Investor Composition

FieldAI operated in stealth mode from its 2023 founding until August 20, 2025, when it simultaneously disclosed $405 million raised across two consecutive funding rounds and announced its $2 billion valuation—quadrupling from the $500 million valuation established in the earlier round. The most recent round raised $314 million and was oversubscribed; co-investors include Bezos Expeditions (Jeff Bezos' family office), Prysm Capital, and Temasek. Additional investors across both rounds include BHP Ventures, Canaan Partners, Emerson Collective, Intel Capital, Khosla Ventures, NVentures (NVIDIA's venture capital arm), and others. Gates Frontier (Bill Gates' investment arm) and Samsung are designated "previous investors," indicating they participated in the earlier round which set the $500 million valuation. By subtraction from the disclosed total, the first round raised approximately $91 million. Note: some pre-diligence references describe the breakdown as $150 million (seed) plus $255 million (Series A); this breakdown is inconsistent with Reuters' reporting of a $314 million latest round and cannot be confirmed from primary sources. The oversubscription and caliber of investors—spanning tech-venture leaders (Khosla, NVentures, Intel Capital), sovereign wealth and institutional funds (Temasek, BHP), philanthropic capital (Gates Frontier, Emerson Collective), and the world's wealthiest individuals' family offices (Bezos Expeditions)—reflects exceptionally broad investor conviction. Vinod Khosla stated: "FieldAI is at the forefront of the general-purpose robotics revolution, and its ability to rapidly deploy will unlock long-term economic and societal value." Intel Capital's public endorsement post formalizes a platform integration and strategic relationship with Intel's hardware stack. The funding is designated for global expansion, product development (locomotion, manipulation), and headcount scale to double by end of 2025. [CO021, CO022, CO023, CO024, CO025, CO026]

1.4 Scale, Partnerships, and Milestone Chronology

FieldAI's growth trajectory since its 2023 founding has been steep. The company had approximately 30 employees at the end of 2024, grew to approximately 130 by the time of its August 2025 funding announcement, and LinkedIn reported 201–500 employees as of early 2026—consistent with its stated plan to double headcount to approximately 260 by end-2025. The company relocated from roughly 13,000 square feet of Mission Viejo office space to a 41,000-square-foot R&D headquarters at 3 Morgan, Irvine Spectrum in late 2025, driven by hiring and lab expansion needs. FieldAI's production deployments now span thousands of missions across hundreds of sites on three continents—North America, Europe, and Asia-Pacific (including Japan). The company has attracted dozens of large enterprise customers across construction, energy, manufacturing, urban delivery, security, and defense, and has established a FieldAI Federal subsidiary for government and military applications. Two major strategic partnerships were announced in early 2026: in February, Certis Group (Singapore's leading integrated security provider) and FieldAI formed a partnership to deploy autonomous security robots globally; in March 2026, Boston Dynamics and FieldAI announced a collaboration combining Spot's mobility with FieldAI's FFMs for construction and other dynamic environments. Marc Raibert, founder of Boston Dynamics, endorsed the company specifically by referencing their 2020 DARPA SubT win: "We've known Ali and his team since they won the Urban-circuit of the DARPA Subterranean Challenge using Spot robots, while tackling similar challenges for NASA." A key risk metric noted by Construction Dive: active robotics deployments among surveyed contractors fell from 65% in 2024 to 46% in 2025 despite 95% positive evaluations—suggesting industry-wide friction between willingness and actual adoption that represents a near-term headwind for FieldAI's commercial scale targets. [CO031, CO032, CO033, CO034, CO035, CO036]

1.5 Exhibits

2.1 Market Boundary and Definition

FieldAI's addressable market is not the full industrial robotics market—the majority of deployed industrial robots reside in structured, repetitive manufacturing environments (automotive assembly, semiconductor fabrication, consumer electronics) that are already well served by traditional fixed-path automation vendors such as FANUC, ABB, and Yaskawa. FieldAI's distinct market is the segment of industrial robot deployments that require adaptive, risk-aware autonomy to operate in unstructured environments where layouts change daily, GPS and pre-mapped paths are unavailable, and static programming fails. This includes active construction sites (where terrain, obstacles, and workflows shift hourly), underground and open-pit mines (where rock falls, gas hazards, and confined spaces make human exposure dangerous), oil and gas topside and inspection environments (where access to assets is hazardous and remote), and flexible or high-mix manufacturing facilities not amenable to fixed-path robots. The market boundary excludes spend on: (a) structured factory automation hardware and software (e.g., traditional articulated arm robots on fixed assembly lines); (b) consumer or logistics warehouse robots operating in controlled environments; (c) robot hardware itself, as FieldAI is a pure-play software vendor; and (d) subsea ROVs and specialty underwater vehicles, which are addressed by a separate supplier ecosystem. The status-quo substitutes that industrial buyers currently rely on include manual human inspection and monitoring crews, traditional fixed-path or single-task robots that require extensive pre-programming, drone/UAV aerial surveys, and BIM/CAD modeling with manual data capture. Each of these substitutes is more expensive per unit of site intelligence, less real-time, and increasingly untenable as labor pools shrink and safety requirements tighten. FieldAI's market spans four primary verticals: (1) construction and infrastructure, (2) energy, mining, and extractives, (3) oil and gas inspection and operations, and (4) industrial manufacturing in unstructured or high-mix configurations. The company describes its target as "dirty, dull, and dangerous" environments where autonomous robots deliver compounding value across safety, documentation, and operational insight— a framing that aligns closely with the structural characteristics of its four served verticals. Adjacent markets that represent future expansion paths include defense and perimeter security, logistics in complex environments, and urban delivery in semi-structured settings. The relevant market is best sized through vertical-specific analyst estimates rather than a single broad robotics TAM, because scope definitions across analyst reports diverge materially. [CM001, CM002, CM003, CM004, CM005]

2.2 Market Sizing Lenses

No single published analyst figure cleanly captures FieldAI's serviceable addressable market because most analyst reports either define markets too broadly (including hardware or all industrial automation) or too narrowly (covering only one vertical). The most relevant cross-cutting figure is Global Market Insights' estimate for the AI-powered industrial robot market: $17.9 billion in 2026 growing at 7.1% CAGR to $33.3 billion by 2035. This figure includes hardware, but it signals the scale of the AI-augmented robotics software layer that FieldAI targets. Separately, The Business Research Company estimates the industrial artificial intelligence market (all software, platforms, solutions) at $13.69 billion in 2026, with exponential growth to $73.54 billion by 2030 at a 51.1% CAGR—the widest set of use cases across manufacturing, energy, and logistics, not unstructured environments specifically. For FieldAI's four core verticals, vertical-specific sizing provides sharper signal: Construction robots: $7.79 billion in 2026 (TBRC) and $1.4 billion (Grand View Research, narrow definition for construction robots only); the range reflects definitional scope differences, with the lower figure likely more defensible for robotic platforms alone. Mining robotics: $1.7 billion in 2026 (PMR), growing at 9.8% CAGR to $3.3 billion by 2033. Oil and gas robotics: $3.8 billion in 2026 (MarkWide Research), growing at 14.6% CAGR to $12.96 billion by 2035; the inspection sub-segment alone is estimated at $0.93 billion in 2026 (Fortune Business Insights). The embodied AI market (cross-vertical, all physical AI systems) is forecast to reach $7.24 billion by 2030 at 17.5% CAGR (Research and Markets). A conservative bottom-up SAM estimate for FieldAI's software-only, unstructured-environment market—derived by applying a 15–25% software-share assumption to the combined $13.2 billion vertical hardware and services market (construction + mining + oil and gas robots in 2026)—yields a SAM range of approximately $2–3.3 billion for 2026 addressable software spend in FieldAI's current verticals. This estimate is inherently uncertain and carries low confidence; FieldAI has not published a SAM figure and no analyst has published a dedicated "AI software for industrial robots in unstructured environments" market estimate. The IFR's World Robotics 2025 report documents 542,000 industrial robot installations in 2024 and 4.66 million units in operational use globally; the fraction of these deployed in unstructured environments where FieldAI's FFMs add value is not separately quantified. The analyst estimate spread is wide (see TM002), and contradictory sizing is preserved as an evidence gap. [CM006, CM007, CM008, CM009, CM010, CM011]

2.3 Buyer, User, and Payer Segmentation

FieldAI sells B2B enterprise software, not hardware. The company's buyer structure follows a classic enterprise industrial technology pattern: the payer is the C-suite or VP Operations/Finance, the technical buyer is the operations or HSE (Health, Safety, Environment) leader, and the day-to-day user is the site superintendent, field engineer, or project manager. Budget ownership sits in one of two pools depending on vertical: (1) capital expenditure (capex) budgets for new automation investments, which are controlled by CFOs and VPs of Engineering or Operations; and (2) operational expenditure (opex) for recurring inspection, monitoring, and site-intelligence contracts, which are controlled by plant or project managers working within pre-approved service vendor agreements. FieldAI has positioned its offering as delivering immediate ROI on the opex side (90%+ reduction in manual documentation time, identification of rework earlier), enabling the initial sale to be framed as cost avoidance rather than capex transformation. The construction segment concentrates its most immediate buying power among the largest general contractors—top ENR-ranked firms such as the unnamed customers referenced in FieldAI's Boston Dynamics case study. These firms own large project backlogs, bear cost-overrun risk, and have internal innovation or digital-transformation budgets earmarked for automation pilots. Adoption triggers include inability to generate accurate as-built BIM data quickly, labor shortages for site documentation teams, and safety mandates requiring periodic inspection of hazardous areas. Mining and energy buyers are concentrated around major operators (BHP, Rio Tinto, and equivalents) whose HSE and operations leaders are incentivized by regulatory compliance, incident reduction, and fleet productivity metrics. BHP Ventures' investment in FieldAI constitutes direct market validation from this buyer segment. Oil and gas customers are driven by inspection compliance requirements, cost reduction in hazardous environment access, and asset integrity management mandates. Across all segments, the adoption path progresses from a limited proof-of-concept (typically one site or asset cluster), through validated ROI, to an enterprise framework agreement covering multiple sites or regions—the same pattern demonstrated by FieldAI's references to "fleet-wide deployments" and "enterprise-scale" expansion across global operations. [CM025, CM026, CM027, CM028, CM029, CM030]

2.4 Growth Drivers and Adoption Constraints

Structural growth drivers are multi-decade and well-documented. Labor shortages are the most immediate: over 50% of the Western US mining workforce is expected to retire by 2029, and the construction industry globally faces persistent skilled-labor deficits. The US Mine Safety and Health Administration (MSHA) recorded 31 mining fatalities in fiscal year 2024 at a fatal-injury rate of 0.0110 per 200,000 hours worked, sustaining regulatory pressure for autonomous alternatives in the most hazardous tasks. IFR's World Robotics 2025 data shows global industrial robot installations growing from 500,000+ per year through 2028, with software-defined autonomy increasingly the differentiating factor as hardware specifications converge. Deloitte's 2026 TMT Prediction confirms that specialized AI foundation models— distinct from consumer LLMs—represent the technology breakthrough that could shift robots from command-and-control to genuinely generalized autonomy between 2026 and 2030. Adoption constraints are equally substantive. McKinsey's Industrial Robotics Survey documents that 71% of industrial companies cite hardware capital cost and 61% cite lack of internal experience as the primary barriers to robotics adoption, and that complex activities such as assembly, surface treatment, and welding are unlikely to be automated in the short to medium term. Deloitte observes that annual industrial robot unit sales have been essentially flat at approximately 500,000 units since 2021 and warns that the market is likely to remain at relatively modest annual growth unless bottlenecks around data quality, integration complexity, and cybersecurity are resolved. Most critically, BuiltWorlds' 2025 Equipment and Robotics Benchmarking report—cited in Construction Dive's coverage of FieldAI's funding—found that the share of construction firms reporting active robotics use fell from 65% in 2024 to 46% in 2025, even as positive evaluations jumped from 74% to 95%. This divergence between enthusiasm and deployment is the defining tension in the near-term market and is FieldAI's most direct execution risk: translating strong pilot performance into durable enterprise revenue. High capex cycles, OT/IT integration friction with legacy plant systems, long payback period requirements, and the volatile capital environment for oil and gas operators are further structural headwinds. The gap between stated market size and deployed product is real and material. [CM031, CM032, CM033, CM034, CM035, CM036]

2.5 Exhibits

3.1 Competitive Landscape and Buyer Alternatives

Industrial enterprises that need autonomous robot operations have five materially distinct ways to solve the job in 2026: (1) deploy FieldAI's Field Foundation Models on existing hardware; (2) adopt an embodied-AI foundation model platform from a well-capitalized rival (Intrinsic/Google, Physical Intelligence, Covariant); (3) use an inspection-specialist platform with its own robotics stack (Gecko Robotics, Machina Labs); (4) rely on the robot OEM's bundled intelligence software (Boston Dynamics Orbit, KUKA, Fanuc native AI); (5) continue with status quo—manual inspection crews, traditional SLAM/GPS-dependent navigation, or an in-house software team. Most large enterprise buyers use more than one approach simultaneously, making multi-homing common and switching costs softer than they initially appear. FieldAI targets the largest unmet slice: dynamic, GPS-denied, unstructured field environments where status-quo and OEM-native solutions fail. The company's hardware-agnostic software model avoids the capital intensity of robot manufacturing while enabling presence across more platforms than any single OEM. As of May 2026, the eight direct and substitute players profiled below collectively represent substantially more VC and corporate capital than FieldAI, creating both validation of the market and a material risk of well-resourced competitive entrenchment. [CP001, CP002, CP003, CP004, CP005]

3.2 Competitor Profiles and Capability Comparison

Intrinsic, now folded into Google (February 2026), offers Flowstate, a web-based development and simulation environment for industrial robots. The platform is hardware-agnostic across KUKA, Fanuc, Universal Robots, and others, and leverages Google's Gemini AI models for vision and reasoning. The May 2026 Fanuc partnership gives Google access to 1.1 million deployed industrial robots—a distribution channel no startup can match. Intrinsic's laid-off 20 percent of staff in January 2023 after five years of development and has not disclosed commercial revenue, but its absorption into Google and the Foxconn joint venture (October 2025) signal accelerating commercial intent. Physical Intelligence (Pi) is a research-first foundation model company founded in 2024 with no commercial product or revenue as of March 2026. Its π0.7 model (April 2026) shows emergent generalization across manipulation tasks, and it is reportedly raising $1 billion at an $11 billion valuation. Pi's VLA (Vision-Language-Action) architecture targets lab-to-production transfer. Covariant was an early mover in robotics foundation models (RFM-1, 8B parameters, launched March 2024), but its founders and approximately 25 percent of staff were acqui-hired by Amazon in August 2024 in exchange for a non-exclusive technology license. The remaining company, led by new CEO Ted Stinson, continues to serve non-Amazon logistics customers (apparel, grocery, pharma) but faces corporate uncertainty and narrow moat from competing data accumulators. Gecko Robotics pairs proprietary TOKA inspection robots with the Cantilever AI platform for critical-infrastructure digital twins, reached unicorn status ($1.25 billion valuation) in June 2025, and in March 2026 won a $71 million five-year U.S. Navy IDIQ contract. Gecko's market is deep in energy, defense, and manufacturing inspection but is structurally narrower than FieldAI's horizontal ambition. Boston Dynamics (Hyundai-owned) offers Spot, Stretch, and Atlas robots plus the Orbit software platform for fleet intelligence. In March 2026, Boston Dynamics announced a direct partnership with FieldAI to extend Spot into uncharted, dynamic environments—making Boston Dynamics simultaneously a hardware partner and a potential software competitor as Orbit expands its AI integration with Google Gemini. Machina Labs focuses on AI-driven robotic sheet-metal forming (RoboCraftsman) for defense and aerospace, with a $124 million Series C (February 2026). It is not a direct competitor to FieldAI's inspection and autonomy focus but competes for defense budget and OEM robotics mind-share. Viam is an open-source robotic software platform ($117 million total raised) targeting developers who want hardware abstraction across IoT, robots, and industrial machines. Viam competes for the same "brain layer" framing but positions itself as a developer-SDK approach versus FieldAI's pre-trained AI model. Internal build remains the incumbent: large industrial companies (energy, mining, construction) maintain internal robotics and software engineering teams. The unit economics of internal build are difficult to quantify but become the default when no commercial platform meets the required safety certifications or data-residency constraints. [CP006, CP007, CP008, CP009, CP010, CP011]

3.3 Pricing and Packaging Comparison

Pricing transparency across the competitive field is low; most platforms sell into enterprise procurement without public list pricing. FieldAI's subscriptions reportedly range from tens of thousands to $500,000 per year per deployment, with the company citing $100 million or more in booked revenue and $140 million in ARR for 2025 per a third-party estimate. Intrinsic does not publish Flowstate pricing; Google's cloud-first model suggests enterprise SaaS tiers layered on top of Google Cloud compute fees, with the Fanuc partnership implying volume licensing. Physical Intelligence has no commercial product or pricing. Gecko Robotics bundles inspection services, Cantilever platform access, and engineering consulting into multi-year contracts; the Navy IDIQ illustrates a $71 million five-year envelope for 18 vessels. Boston Dynamics Orbit is sold as a software layer on top of hardware; annual fees have not been disclosed but are referenced by Michelin and other enterprise case studies as ongoing operational software. Viam offers a developer-friendly free tier and enterprise pricing upon request; the open-source model drives developer adoption before upselling managed infrastructure. Machina Labs prices on manufacturing-service contracts rather than software subscriptions. Status-quo manual inspection costs are estimated by industry at $13–$20 billion per year for the U.S. Navy alone, and FieldAI-enabled systems reduce documentation time by more than 90 percent on construction sites per the Boston Dynamics partnership announcement. [CP021, CP022, CP023, CP024, CP025, CP026]

3.4 Differentiation Durability, Moat Analysis, and Adverse Evidence

FieldAI's claimed moat rests on four pillars: (1) physics-first, risk-aware foundation models trained on proprietary real-world deployment data; (2) on-edge deployment without cloud or GPS dependency, enabling markets inaccessible to cloud-first competitors; (3) data flywheel from a growing fleet across construction, energy, mining, and defense; and (4) partner ecosystem anchored by Boston Dynamics, Certis Group, and Carnegie Mellon's Robotics Innovation Center. Independent assessment of this moat is structurally difficult at this stage: deployment data volume and model performance benchmarks are not publicly disclosed, and the hardware-agnostic positioning that is a strength also means competitors can pursue the same robot partners. Several adverse signals merit attention. First, Physical Intelligence is raising capital at a $11 billion valuation—5.5× FieldAI's—despite having no commercial products. If Pi commercializes before FieldAI achieves a dominant data advantage, the superior capital base could compress FieldAI's differentiation window. Second, the Google/Intrinsic/Fanuc integration (1.1 million deployed robots receiving AI upgrades) represents an installed-base leverage that FieldAI cannot replicate organically. Third, UpsideList's March 2026 analysis assigns a 50-percent probability to a bear case in which NVIDIA Isaac or Boston Dynamics internalizes the autonomy layer, wiping FieldAI's common equity through a down-round. Critically, the same Boston Dynamics that is FieldAI's most prominent partner runs the Orbit software platform, backed by Google Gemini, and could elect to develop competing autonomy capabilities. Fourth, an independent analysis of Covariant—a warehouse-AI analog—rates its data moat as "narrow, not wide," citing competing accumulation by Physical Intelligence, Nimble, and Amazon; the same logic applies to any embodied AI platform. Foundation-model commoditization risk is real: as VLA models proliferate through open-source releases (Pi has open-sourced early models) and hyperscaler APIs, the physics-first differentiation may compress to a trade secret rather than a structural moat. FieldAI's defense vertical (FieldAI Federal, led by former DARPA PM Eric Krotkov) and its GPS-denied edge deployment requirement may prove the most defensible niches, as they require hardware certifications and security clearances that create regulatory switching costs independent of model quality. [CP027, CP028, CP029, CP030, CP031, CP032]

3.5 Exhibits

4.1 Revenue Streams and Monetization Strategy

FieldAI's revenue model is B2B enterprise, centered on two primary streams: (1) upfront hardware integration fees collected when a customer first deploys FieldAI's sensor-compute payload and firmware onto existing third-party robots; and (2) recurring software licensing fees for continued access to Field Foundation Models (FFMs), fleet management, and analytics services. The company does not manufacture robots and explicitly avoids hardware OEM positioning, keeping the model closer to an enterprise software subscription than a robotics hardware vendor. All on-device inference runs at the edge with latency under 100ms; fleet data is uploaded to FieldAI's cloud platform for analytics and federated model improvement. List pricing is not publicly disclosed. Third-party sources characterize the subscription range as "tens of thousands to $500,000 per year" depending on deployment scale, with CEO Ali Agha citing "multi-million dollar contracts" in the US, Europe, and Asia as evidence of large-enterprise deal sizes. Construction Dive and Sacra confirm the B2B hardware integration plus software licensing structure. There is no evidence of a consumer-facing revenue stream, marketplace model, or hardware OEM revenue. A defense-oriented subsidiary, FieldAI Federal, is operational and could eventually generate government contract revenue, though no contract awards have been publicly disclosed. The only public revenue signal is an April 2026 Orange County Business Journal report citing anonymous "sources familiar with the matter" indicating the company has more than $100M in booked revenue. GetLatka's analyst platform estimates $140M ARR as of November 2025. Neither figure is audited, company-confirmed, or broken down by stream. These data points are treated as directional signals rather than confirmed financial metrics. [CI001, CI002, CI003, CI004, CI005, CI006]

4.2 Capital Adequacy and Forward Runway

FieldAI's two-round funding history is detailed in the Company Overview chapter. For capital adequacy purposes, the key inputs are: $405M total capital raised as of August 2025, with the second round ($314M) closing oversubscribed in August 2025, co-led by Bezos Expeditions, Prysm Capital, and Temasek, with participation from Khosla Ventures, NVentures (NVIDIA), Intel Capital, Canaan Partners, BHP Ventures, and Emerson Collective. The first round (~$91M) closed in 2024 at a $500M valuation; Gates Frontier and Samsung are designated prior-round investors. Cash on hand is not disclosed. Monthly burn rate is not disclosed. The company's headcount trajectory—growing from approximately 30 employees at the end of 2024 to approximately 130 at the August 2025 announcement, with explicit plans to double to approximately 260 by end of 2025—implies a rapid ramp in the fixed-cost base. At 200–260 employees at an average fully-loaded cost of $250K–$350K per employee (consistent with Irvine, CA and distributed engineering teams), annualized compensation expense alone is $50–90M/year, implying a monthly run-rate labor cost of approximately $4–8M. Adding global deployment expenses, cloud infrastructure, hardware payload costs, and G&A, total monthly burn is estimated at $5–10M. At the lower bound, $405M total raised at $5M/month burn implies approximately 81 months of gross runway from inception; at $10M/month, approximately 40 months. Netting against an assumed $100M+ in cumulative revenue offset, net runway from the August 2025 close is estimated at 24–48 months, placing the next-round trigger in late 2027 to early 2029 absent revenue acceleration or unexpected cost escalation. FieldAI has not announced any debt facilities, project finance obligations, revenue-based financing, or lines of credit. The company's capital structure appears to be pure equity. The absence of hardware manufacturing (unlike vertically integrated robotics companies) materially reduces working capital and inventory financing needs compared to peers. [CI009, CI010, CI011, CI012, CI013, CI014]

4.3 Unit Economics and Cost Structure

FieldAI's unit economics are largely undisclosed. No audited financials have been published, and the company has not released gross margin, customer acquisition cost (CAC), net revenue retention (NRR), annual contract value (ACV), or payback period data. The following analysis is based on comparable B2B industrial AI software companies and the observable structure of FieldAI's revenue model. Gross margin is estimated to be in the 55–80% range on the software licensing component, based on typical enterprise AI software benchmarks. The hardware integration service component—which involves physical sensor-compute payload deployment and field support— likely carries a lower margin (20–45%), pulling blended gross margin below pure-software comparables. At scale, the software licensing stream should dominate and improve blended margins. The on-edge architecture (no cloud inference dependency) reduces ongoing variable costs per deployment compared to cloud-heavy AI platforms, which is a favorable structural feature. Sales efficiency for enterprise industrial robotics is structurally challenged: sales cycles are typically 12–24 months, require field demonstration and integration testing, and involve multi-stakeholder procurement. CEO Agha cited multi-million-dollar contract values, suggesting deal sizes that could support CAC payback in 1–3 years at moderate churn rates—but this cannot be verified without disclosed CAC and NRR data. Revenue per employee is approximately $900K–$1M if the $140M ARR estimate and 150-employee figure are correct, which is strong for an enterprise software company of this stage. FieldAI's GTM motion focuses on direct enterprise sales, hardware-partner relationships (Boston Dynamics, OEMs), and geographic expansion partnerships (Certis Group, Asia), which reduces the addressable market ceiling through a pure direct-sales channel. [CI016, CI017, CI018, CI019, CI020, CI021]

4.4 Financial Opacity and Diligence Blockers

FieldAI is a private company with no public financial reporting obligations. As of May 2026, key financial metrics that remain undisclosed or unverifiable include: (1) audited ARR or total revenue; (2) gross margin by revenue stream; (3) CAC and payback period; (4) net revenue retention rate; (5) customer count and concentration; (6) cash on hand; and (7) monthly or quarterly burn rate. The available revenue signals are conflicting. GetLatka estimates $140M ARR (November 2025), while other analyst platforms have cited figures as low as $5M—a 28x discrepancy that reflects the challenge of estimating private-company metrics from outside. UpsideList's community model flags -2% upside and notes the company as effectively pre-revenue from a secondary-market pricing perspective, suggesting liquidity and valuation uncertainty at the private secondary level. The $2B valuation implies a 14x revenue multiple against the $140M ARR estimate, and a 20x multiple against the $100M booked revenue signal. Both multiples are aggressive for a company with undisclosed profitability, no public financial track record, and operating in a market where at least 54% of robotics deployments in the construction vertical (the company's largest disclosed market) remain as pilot or stalled projects (per BuiltWorlds 2025 data). These multiples are defensible only if ARR growth is rapid and margins are high—neither of which can be confirmed from available data. FieldAI has not filed a Form D with the SEC (confirmed via EDGAR full-text search), meaning the company either relied on an exemption that does not require a Form D, used a different legal entity, or has not yet filed. The California entity "Field Ai, Inc." (document #6436098, incorporated October 24, 2024, Delaware formation, CEO Aliakbar Aghamohammadi) is active. The adverse valuation signal from secondary-market analysts and the wide revenue estimate range are not disqualifying, but they mean underwriting the $2B valuation requires accessing non-public financial information through formal due diligence. [CI022, CI023, CI024, CI025, CI026, CI027]

4.5 Financial Verdict

FieldAI's financial picture is that of a well-capitalized, early-revenue software business growing fast enough to attract $405M in 24 months but still too opaque for external underwriting. The $2B valuation is carried entirely by investor signal, growth trajectory, and team pedigree—not by verified financial fundamentals. If the $140M ARR estimate is accurate and growing at 50–100% year-on-year, the valuation is in range for a category-creating AI software company; if ARR is materially lower or growth is slower, the multiple becomes difficult to justify. Forward capital adequacy appears solid: $405M raised against a pure-software operating model (no hardware manufacturing capex, no inventory financing) and a disclosed >$100M in bookings implies at least 2–3 years of operating runway before the next fundraise. The largest diligence risk is not solvency but information asymmetry—the gap between what is publicly verifiable and what is needed to underwrite the valuation is unusually wide for a $2B company. [CI030, CI031, CI032]

4.6 Exhibits

5.1 Product Definition and Module Map

FieldAI is a pure-play AI software company that does not manufacture robots. Its commercial offering is the Field Foundation Model (FFM) platform: embodiment-agnostic autonomy software installed on partner or customer-owned robot hardware. The FFM acts as a universal "brain" that enables robots to perceive, reason, and act in unstructured industrial environments without requiring pre-mapped layouts, GPS, cloud connectivity, or per-task reprogramming. The platform is hardware-agnostic across quadrupeds (e.g., Boston Dynamics Spot), wheeled robots, humanoid platforms, and passenger-scale autonomous vehicles. FieldAI's application portfolio spans eight solution categories listed on its official Solutions page: site mapping and documentation; facility and equipment inspection; condition anomaly detection; large-scale data capture; unmanned material transport; security and threat detection; telepresence and teleoperation; and search and ISR. The company develops applications both in-house and through a partner payload model with third-party providers. Near-term commercial focus is inspection and data collection (revenue-generating now), with manipulation and general-purpose operations on the mid- and long-term roadmap. A distinguishing architectural principle is "Context over Training": instead of requiring exhaustive scenario pre-training, FFMs use contextual cues and learned physical priors to make inferences about unforeseen situations. CEO Ali Agha described the robot's risk-awareness as giving it a sense of "how confident I am" in any action, with customer-configurable risk thresholds that adjust behavior in safety-critical zones. The result is a deployment model where robots can be activated at new sites quickly and scale to enterprise fleets without extensive per-site customization. [CE001, CE003, CE006, CE007, CE008, CE009]

5.2 Architecture and Technical Design

The FFM platform is organized into three layered model components. The core FFM handles perception, world modeling, and decision-making using multimodal sensor streams including vision, LiDAR, text, and audio. The Dynamics Foundation Model (DFM) is a sub-model that integrates the robot's intrinsic kinematic and dynamic properties, enabling the same cognitive layer to drive mechanically distinct robot bodies. The Multiagent Foundation Model (MFM) is a coordination layer that allows multiple robots in a fleet to share environmental representations and reason collectively at site scale — turning a robot swarm into a distributed autonomous system rather than isolated tools. Architecturally, FieldAI operates in "belief space": the system continuously tracks probability distributions over environmental states rather than assuming perfect information, a design lineage traceable to CEO Ali Agha's academic work on FIRM (Feedback-based Information RoadMap) belief-space planning, as well as the NeBula system that won Phase II of the DARPA SubT challenge. This probabilistic approach means robots can function in GPS-denied, low-visibility, and sensor-degraded conditions. When FFMs encounter unfamiliar conditions, they explicitly slow down or choose more conservative paths — an admitted throughput cost that the company treats as a safety feature. The real-world data flywheel means each deployment improves model performance over time, creating a compounding data advantage. NVIDIA integration deepens the engineering stack. FieldAI uses NVIDIA Omniverse NuRec to reconstruct high-fidelity 3D digital twins from sensor data captured as a byproduct of normal robot operations. These digital twins load into NVIDIA Isaac Sim and Isaac Lab for robot policy training and validation, and FieldAI is adopting the NVIDIA Physical AI Data Factory Blueprint via Microsoft Azure, using NVIDIA Cosmos open-world foundation models and NVIDIA OSMO for synthetic data pipeline automation. The company's GitHub organization (field-ai) maintains active forks of robotics infrastructure libraries — Spot SDK, rosbridge_suite, spconv, unitree_sdk2 — with commits as recent as April 2026, providing developer-side evidence of the platform's integration surface, though no proprietary model weights or SDK documentation is publicly published. The private Squarespace release notes page further limits external auditability of shipped capabilities. [CE002, CE004, CE005, CE021, CE022, CE014]

5.3 Deployment Workflow and Use Cases

FieldAI's deployment model is B2B software: robots owned or leased by the customer receive the FFM brain as a software installation, with FieldAI providing integration services and ongoing model updates. No dedicated hardware needs to be purchased from FieldAI. Deployment time is a key differentiator claim: because robots do not require pre-mapped environments or GPS infrastructure, site activation is described as rapid, with the company and its investors describing processes that "once took months now take hours." One unnamed global industrial manufacturing executive cited a specific 3.5-month-to-12-hours outcome (over 200x improvement) when using FieldAI's technology plus NVIDIA NuRec digital twin generation — though this claim is unverified by an independent third party. Construction is the leading commercial vertical. Big-D Construction has been a FieldAI customer for over two years and in April 2026 announced expansion across multiple projects, with one project catch reportedly saving $1.2 million through early defect detection. In the security vertical, Certis Group (Singapore) deploys FFMs integrated with its proprietary Mozart orchestration platform across public infrastructure, transport hubs, and industrial facilities. FieldAI's Boston Dynamics partnership, formalized March 12, 2026, targets construction inspection; Spot robots equipped with FFMs perform 3D scanning, hazard identification, and overnight monitoring autonomously. The company also has federal and defense use cases handled via FieldAI Federal, led by Dr. Eric Krotkov, a former DARPA Program Manager. Geographically, FieldAI has documented deployments in the US, Japan, Europe, and across Asia-Pacific through Certis. Customers are generally not named publicly (TechCrunch confirmed the company declined to disclose customer names at August 2025 funding announcement), with Big-D Construction as the only construction-sector customer publicly identified by name as of May 2026. [CE011, CE012, CE018, CE023, CE024, CE032]

5.4 Trust, Safety, and Compliance

FieldAI's safety architecture is based on probabilistic risk management rather than explicit rules: the FFM operates in belief space, maintaining confidence estimates about the environment and taking conservative actions when uncertainty exceeds a customer-configured threshold. CEO Agha has described this as the robot knowing "how confident it is" and slowing or rerouting to avoid unsafe actions. This design addresses the hallucination risk of conventional LLMs applied to robotics, as FieldAI's models were designed from the ground up for physical safety rather than retrofitting language or vision models. However, no publicly documented safety certifications, regulatory approvals, or compliance standards (e.g., ISO 10218, IEC 62443, ANSI/RIA, FedRAMP, or equivalents for industrial robotics) have been disclosed as of May 2026. The Certis partnership agreement mentions that both companies will "work together on training, safety validation and operational frameworks to support responsible deployment," implying that formal safety validation frameworks are still in development rather than completed. Similarly, the Boston Dynamics partnership focuses on practical deployment outcomes rather than certifications. FieldAI's role in defense/federal contexts through FieldAI Federal implies some level of government security review, but no specific clearances or FedRAMP status have been published. The private release notes page prevents external parties from auditing shipped versus roadmap capabilities, and no formal bug bounty or security disclosure policy has been identified. [CE006, CE026, CE031, CE033, CE034]

5.5 Roadmap, Maturity, and Developer Ecosystem

FieldAI's product maturity is strongest in inspection and data collection, where the company has paying customers and documented production deployments across hundreds of sites. The three-phase roadmap, articulated in analyst coverage and investor materials, is: (1) near-term through 2026 — autonomous inspection, monitoring, and data collection, already commercialized; (2) mid-term 2026–2028 — intervention and manipulation, expanding beyond sensing to physical actions; (3) long-term 2028+ — general-purpose autonomy across all platforms including humanoids. This roadmap aligns with FieldAI's stated use of $405M raised to fund development across locomotion and manipulation capabilities. The CMU Robotics Innovation Center tenancy (announced February 2026, 2,500 sq ft at Hazelwood Green) formalizes the company's academic research partnerships, giving access to 1.5-acre outdoor testing grounds and specialized robotics labs. FieldAI sponsors research at CMU's AirLab and LeCAR lab, and supports the VectorRobotics student team on humanoid loco-manipulation, indicating active investment in the capability gaps on the mid-to-long-term roadmap. The Field AI Research Institute (FAIRI) publishes peer-reviewed research, with the NeBula DARPA SubT paper and 2024 publications on risk-aware AI and MPPI control providing academic credibility. Developer ecosystem access is limited. FieldAI's GitHub organization (field-ai) hosts 10+ forks of standard robotics libraries but publishes no proprietary FFM code, model weights, or SDK. No public developer portal, API documentation, or community forum (Discord, Slack, Stack Overflow tag) has been identified. The release notes page is a private Squarespace site. This limits the developer-signal surface to the observed GitHub repository activity rather than a genuine practitioner community. The partnership with NVIDIA and CMU provides an indirect proxy for engineering credibility, but independent technical evaluation of FFM capabilities is not yet possible from public sources. [CE019, CE025, CE028, CE029, CE036, CE035]

5.6 Exhibits

6.1 Customer Segments, Buyer/User/Payer Logic

FieldAI sells embodied AI software and sensor-compute payloads to enterprise industrial organizations across six named verticals: construction, energy/oil & gas, mining, manufacturing, urban delivery and inspection, and federal/defense. It does not manufacture robots; the company's product is the autonomy software and payload that retrofits hardware the buyer already owns or procures from an OEM partner such as Boston Dynamics. This clean software-licensing model means the payer is always the industrial enterprise, the buyer is typically a technology officer, VP of operations, or innovation director, and the daily user is the field superintendent, safety manager, or operations lead who deploys and monitors the robot. Subscriptions are reported to range from tens of thousands of dollars to $500,000 per year depending on deployment scale, with hardware integration services layered on top for first-time deployments. The construction vertical is FieldAI's most commercially mature segment — it has the only publicly named customer case studies (Big-D Construction, DPR Construction) and the only confirmed multi-year production deployments as of mid-2026. Energy/oil & gas and industrial manufacturing are served by a dedicated product page covering equipment anomaly detection, gauge and thermal monitoring, and digital-twin creation, but no named enterprise customers in these segments have been disclosed publicly. Security operations entered the addressable market via the Certis Group strategic partnership announced February 2026, and federal/defense is a separately branded segment (FieldAI Federal) with a Detroit Defense partnership noted in prior chapters. The buyer/user split creates an important sales dynamic: the champion in the purchase decision is often a technology or innovation director, but production adoption hinges on buy-in from superintendents and safety managers who use the system daily. At Big-D, organic demand from superintendents reinforced top-down investment decisions — a positive signal for stickiness — but also implies that a single project team's resistance could slow site-level expansion. [CU001, CU002, CU003, CU004, CU005, CU006]

6.2 Adoption Trajectory and Deployment Scale

FieldAI's August 2025 funding announcement stated "successful testing and deployments across hundreds of complex real-world industrial environments" in Japan, Europe, and the United States, with "some of the world's largest companies" across construction, energy, manufacturing, and urban delivery and inspection. The round was described as oversubscribed following "rapid customer adoption and multiple expansion contracts," and by March 2026 the company referenced a "rapidly growing fleet" with customers expanding across North America, Europe, and Asia in the NVIDIA collaboration announcement. These are company-originated claims without third-party verification of active site count, customer count, or fleet size. Independent tracking by robotsinconstruction.com as of April 2026 counts exactly four confirmed construction deployments: Big-D Construction (production, 2+ years), DPR Construction (production, ~1.5 years as of November 2025), and two unnamed top-10 ENR general contractors. The platform notes FieldAI has been shipping since 2024. The discrepancy between "hundreds of environments" and four confirmed construction deployments may be partly explained by non-construction verticals, pre-commercial testing environments, and demo or evaluation deployments that are not tracked by independent sources. The Big-D and DPR deployments are the only cases with customer-identified outcomes (named executive and superintendent quotes, specific square-footage data, elapsed deployment duration). The Boston Dynamics partnership (March 2026) added a credible third-party validator, with Boston Dynamics citing customers "from Asia to Europe to North America" expanding to enterprise-scale deployments. FieldAI stated plans to expand the Spot deployment to "one of the largest third-party quadruped fleets in the world." Neither claim carries independent site-level corroboration. [CU010, CU011, CU012, CU013, CU014, CU015]

6.3 Named Customer Proof

Big-D Construction and DPR Construction are FieldAI's two fully named, production-deployed customers with publicly attributed outcomes. Big-D, a US contractor with 25+ years in commercial construction, has been in active deployment for over two years as of April 2026. C-level executive Shaun Orr described a breakthrough moment when the robot walked a project site in real time, linked to the BIM model and project schedule, identified missing work and safety issues, and correlated findings back to the company's project management platforms — a capability that no single human team could replicate alone. Orr projected that "every project will have some representation of FieldAI tools." Big-D superintendent Bronson Dupaix described replacing five-hour manual job-walk sessions using handheld cameras and scanners. Big-D director Chantelle Menlove independently noted that a superintendent who had seen the robot on one site called to request it on his next site, indicating organic bottom-up demand. DPR Construction, one of the ten largest US contractors (and a data-center and healthcare specialist), hosted FieldAI robots at a Santa Clara, California data center site for approximately 1.5 years as of November 2025. The robot captured 45,000+ photos, walked 100+ miles, scanned 500,000 square feet of interiors and 125,000 square feet of roofing. Superintendent Justin Schreiner said the system "makes us better at what we do" by removing mundane documentation tasks and enabling teams to focus on critical work. Certis Group, Singapore's largest security and operations company, announced a strategic partnership with FieldAI in February 2026 to deploy autonomous security robots across its global operations. Certis' Mozart orchestration platform integrates with FieldAI's FFMs for autonomous patrol, incident detection, and human-robot coordination at multi-site security environments. Certis president Ng Tian Beng confirmed the partnership is structured for "live, mission-critical environments." The partnership is not a direct customer sale but a channel/OEM-style arrangement where Certis is both partner and operator-customer for the security vertical. An unnamed global industrial manufacturing executive was quoted in a March 2026 FieldAI/NVIDIA press release saying a process that "once took three and a half months now takes just twelve hours" — a 200× improvement. The customer is not named and the metric is unverifiable, but the quote is attributed by a named company executive (CEO Ali Agha) in a formal press release. Two additional unnamed top-10 ENR firms appear in the robotsinconstruction.com independent tracker with confirmed production deployments but no disclosed outcomes. [CU019, CU020, CU021, CU022, CU023, CU024]

6.4 Retention, Expansion, and Concentration Risk

FieldAI has not disclosed any retention metrics — no NRR, GRR, cohort data, churn rate, contract renewal rate, or customer satisfaction score has appeared in any public statement, press release, or investor communication as of May 2026. The company is pre-revenue-disclosure and private; no regulatory filing mandates these disclosures. Qualitative retention signals are modestly positive for construction. Big-D's multi-year active deployment and explicit plans to expand FieldAI to every future project are the strongest available signal of retention and expansion. DPR's corporate venture arm (WND Ventures) called the August 2025 funding raise "a sign that people can expect larger funding rounds for robotics," and noted that DPR sites helped test proof of concept — language that implies an ongoing relationship. The oversubscribed funding round, explicitly attributed to "multiple expansion contracts," provides financial-signal corroboration of repeat purchase, though specific contract values and customer names remain undisclosed. Structural concentration risk is elevated. Based on publicly confirmed production deployments, FieldAI's construction customer base consists of four accounts: Big-D, DPR, and two unnamed ENR firms. Loss of either named anchor customer would remove more than 50% of the visible evidence base. The energy, mining, and manufacturing verticals have no named public customers, meaning outside construction the proof-of-production is a gap. Certis provides vertical diversification into security but is a partner-operator, not a direct enterprise SaaS account. A broader macro headwind exists for the primary vertical: the BuiltWorlds 2025 Equipment & Robotics Benchmarking report found that active construction robotics usage among contractors dropped from 65% to 46% year-over-year, even as positive sentiment rose to 95%. The decline was attributed to a move away from experimental pilots toward selective, proven implementations — a dynamic that may slow FieldAI's new customer acquisition while benefiting its existing production accounts. Long enterprise sales cycles, multi-stakeholder approval workflows, and NDA-restricted customer naming mean FieldAI's real customer base is almost certainly larger than public evidence suggests but no independent method exists to verify this. [CU031, CU032, CU033, CU034, CU035, CU036]

6.5 Exhibits

7.1 Regulatory and Legal Risk

FieldAI deploys AI systems that make autonomous physical decisions in safety-critical industrial environments. This places the company squarely within the scope of several evolving 2026 AI regulatory frameworks. Under the EU AI Act, robotics AI systems used as safety components in industrial settings are classified as high-risk, requiring risk management systems, technical documentation, human oversight design, CE marking, and post-market monitoring. Baker Botts notes that the original August 2, 2026 deadline for Annex III high-risk systems is subject to a European Commission Digital Omnibus proposal that could extend it to six months after harmonised standards are published (backstop December 2027); however, the extension depends on adoption before August 2026, meaning companies must maintain parallel compliance preparation under the existing deadline. Colorado's AI Act (SB 24-205), effective June 30, 2026 — the first comprehensive US state statute on high-risk AI systems — requires impact assessments, consumer disclosures, and reasonable care to prevent algorithmic discrimination. California SB 53, effective January 2026, requires developers of large frontier AI models (>10²⁶ FLOPS training) to publish risk frameworks and report critical safety incidents within 15 days; whether FieldAI's training compute crosses this threshold is undisclosed. Federal direction remains unsettled: the Trump administration's December 2025 executive order signals intent to pre-empt state AI laws but does not preempt them today, leaving a patchwork compliance burden. On product liability, courts are consolidating around a products-liability framework for AI deployments. KL Gates (March 2026) documents a growing body of pleadings treating deployed AI systems as "products" subject to design-defect, failure-to-warn, and foreseeable-misuse theories. The EU revised Product Liability Directive extends strict liability across the supply chain to upstream AI component providers, which would encompass FieldAI even if the deploying customer or robot OEM owns the end-product relationship. Garcia v. Character Technologies set a precedent that consumer-facing AI can be treated as a product under strict liability at the pleading stage. For industrial robotics AI, physical harm causation (crushing, striking, unexpected motion) would be even more direct than in consumer chatbot cases. No public enforcement actions, regulatory investigations, lawsuits, or IP disputes involving FieldAI have been identified. FieldAI Federal, the defense subsidiary, would be subject to ITAR/EAR export controls if its autonomy systems cross into dual-use military technology; no public disclosure of export authorisations has been found. IP protection relies primarily on trade secrets and publication velocity rather than granted patents: USPTO patent application search finds one filed patent application (US 2025/0252306) but no granted patents as of May 2026, leaving the core FFM technology potentially exposed to imitation if personnel depart. [CR001, CR002, CR003, CR004, CR005, CR006]

7.2 Operational, Safety, and Security Risk

FieldAI's core value proposition — autonomous robots operating in unstructured, safety-critical environments — is simultaneously its highest operational risk surface. Robots powered by FFMs make decisions entirely on-edge without GPS, maps, or cloud connectivity. A model inference error in a novel environment (e.g., a debris configuration not represented in training data, an unexpected human crossing a pre-authorised robot path) could result in physical harm, property damage, or mission failure in high-consequence settings such as construction, mining, or energy operations. The 2026 International AI Safety Report identifies "loss-of-control scenarios" and "inadequate real-world reliability testing" as two of the most pressing risks in rapidly advancing autonomy systems, directly applicable to FieldAI's product domain. A 2025 widely-reported incident in which a Unitree H1 robot malfunctioned near a human worker, and the CES 2026 safety analysis from SRES noting that "existing industrial safety protocols often fall short when robots share open workspaces with humans," illustrate how the broader competitive cohort — not FieldAI specifically — has suffered public safety failures that will shape customer due diligence requirements, procurement standards, and potential litigation expectations. No safety incidents involving FieldAI-powered robots have been publicly reported as of May 2026. However, the absence of published performance benchmarks, safety validation test reports, or third-party safety certifications (ISO 10218 for industrial robots, IEC 62443 for industrial cybersecurity, or SOC 2 for data handling) means that enterprise customers and investors cannot independently verify the residual risk level. The NIOSH Center for Occupational Robotics Research and OSHA both document a documented baseline of serious robot-related workplace injuries in industrial settings; FieldAI's deployments in active construction zones and mining operations place its robots in precisely the settings where these incidents are most common. The company's own deployment narrative — hundreds of sites on three continents — increases the probability of encountering a first adverse event as deployment scale grows. On cybersecurity: FieldAI's robots upload fleet data to a cloud analytics platform and communicate with NVIDIA Omniverse and Isaac infrastructure. No public DPA, security whitepaper, penetration test report, or SOC 2 certification has been found. An adversarial attacker capable of spoofing on-device sensor data or injecting commands into the fleet management plane could potentially cause physical harm. Industrial IoT and operational technology (OT) security standards (IEC 62443) are not confirmed as implemented. These gaps create material residual exposure if a security incident precedes formal certification. [CR013, CR014, CR015, CR016, CR017, CR018]

7.3 Partner and Technology Dependency Risk

FieldAI's go-to-market strategy and technology stack are concentrated across a small number of high-value partners and suppliers. The March 2026 Boston Dynamics partnership is the most visible commercial dependency: the company's construction-focused autonomy story is built around Spot quadruped hardware, and Marc Raibert's public endorsement citing the DARPA SubT history ties both brand and market credibility to the relationship. Boston Dynamics is owned by Hyundai Motor Group, which separately invested in FieldAI in February 2026 — creating a potential misalignment if Hyundai decides to invest in competing proprietary autonomy software for Boston Dynamics hardware, or if the partnership is restructured as part of a Hyundai-level strategic realignment. The terms of the partnership (exclusivity provisions, revenue sharing, minimum commitment, termination rights) are not publicly disclosed, making it impossible to assess contractual protection. NVIDIA is a deeper, more pervasive dependency. NVentures (NVIDIA's corporate venture arm) is a financial investor, and FieldAI uses NVIDIA Omniverse NuRec for 3D reconstruction, Isaac Sim / Isaac Lab for simulation training, NVIDIA Cosmos for synthetic data generation, and OSMO for pipeline automation. This creates a multi-layer strategic and commercial lock-in. While the investor relationship reduces the probability of abrupt supply disruption, NVIDIA's own priorities (hyperscaler allocations, data center GPU supply bottlenecks, competitive dynamics) could compress FieldAI's compute access or raise training costs. VerticalData reports that high-bandwidth memory (HBM) bottlenecks persist in 2026, with data-centre GPU lead times extending 9–12 months for the latest models; smaller AI firms without multi-year cloud commitments are most exposed. Beyond NVIDIA and Boston Dynamics, the Certis Group security partnership (February 2026) adds a third integration dependency. Certis uses its Mozart platform for fleet command-and-control, and FieldAI's FFMs must inter-operate with this proprietary infrastructure. Partner diversification is an explicit part of FieldAI's hardware-agnostic narrative — the company claims FFM compatibility with quadrupeds, humanoids, wheeled robots, and AVs — but named commercial partnership depth (Boston Dynamics for construction, Certis for security) concentrates revenue and reputational exposure in two partners as of May 2026. [CR023, CR024, CR025, CR026, CR027, CR028]

7.4 People, Talent, and Execution Risk

FieldAI's founding team carries exceptional domain credibility — Ali Agha's leadership of the DARPA SubT-winning CoSTAR team and his prior NASA JPL principal investigator role are the primary basis for investor and customer confidence. This creates a concentrated key-person risk: Agha is simultaneously the company's public face, its primary technical thought leader, and its most visible commercial relationship owner. No named COO, Head of Engineering, or parallel commercial leadership figure with equivalent public standing has been publicly identified. David Fan (CTO) and Shayegan Omidshafiei (CSO/President) provide some leadership depth, but their public profiles are substantially less prominent than Agha's. An involuntary departure, extended incapacitation, or reputational event involving Agha would likely disrupt fundraising, enterprise customer confidence, and partnership relationships simultaneously. Beyond the founder, FieldAI faces talent risk inherent to rapid scaling in one of the most competitive labour markets in technology. Headcount reportedly grew from founding in 2023 to 130+ by August 2025, implying annualised growth that makes culture-maintenance, engineering coordination, and product execution discipline difficult to sustain. The company competes for robotics-AI engineers with Google Intrinsic (now integrated into Google AI), Physical Intelligence (raised $700M+), Amazon Robotics, and Tesla Optimus — all of whom are hiring from the same narrow pool of PhDs and senior engineers who can work on embodied AI at scale. No evidence of post-funding key technical departures has been identified, but the absence of public evidence is not confirmation of stability. Execution risk also manifests in the commercial dimension. FieldAI was in stealth until August 2025; less than one year of demonstrable enterprise sales capacity has elapsed. The company has not published customer names, deployment case studies with verified outcomes, or retention metrics. The transition from a research-lineage team (NASA JPL, DARPA) to a revenue-accountable enterprise software vendor is a well-documented organisational inflection point that many deep-tech spinouts fail to navigate quickly enough to justify their valuation trajectory. [CR033, CR034, CR035, CR036, CR037, CR038]

7.5 Financial, Valuation, and Thesis-Break Risk

FieldAI's $2B Series A valuation is the largest in robotics-AI history for a company at this stage, and it creates a high bar for future capital events. The only independent revenue signal — "$100M+ in booked revenue" from anonymous OCBJ April 2026 sources — implies a revenue multiple of roughly 20× at the $2B mark, consistent with the highest-multiple cohort in enterprise AI software but exposed to multiple compression if growth deceleration, AI sector sentiment shifts, or public-market comps reprice. GetLatka's $140M ARR estimate (November 2025) is unaudited and unconfirmed by the company. Gross margin, NRR, CAC, cash on hand, and burn rate are entirely undisclosed, preventing underwriting of the valuation with conventional SaaS or deeptech metrics. Capital intensity is a hidden risk: industrial robot deployments require on-site integration, custom payload installation, and ongoing model fine-tuning, all of which carry cost-of-delivery overhead that could compress gross margins significantly below pure-software levels. If gross margins are substantially below 60–70%, the $2B valuation becomes difficult to support even at accelerating revenue growth. A second risk is burn concentration: at 130+ employees and with a leadership team from high-cost academic and technology institutions, monthly burn could reach $10–20M, implying a cash runway of 2–4 years from the $405M raise. Market or macroeconomic conditions that disrupt growth before breakeven could force a down round. Thesis-break triggers are defined at the intersection of evidence that undermines the core investment thesis: (1) a material safety incident involving FieldAI-powered robots that leads to customer withdrawal, regulatory action, or litigation; (2) EU AI Act enforcement action blocking EU market access; (3) Boston Dynamics partnership termination or Hyundai strategic pivot that removes the primary commercial case study; (4) Ali Agha departure or reputational event; (5) a valuation reset or down round indicating the market will not sustain the $2B entry valuation. Each of these triggers is independently sufficient to change the risk-adjusted return profile materially. [CR041, CR042, CR043, CR044, CR045, CR046]

7.6 Exhibits

8.1 Valuation Context and Financing History

FieldAI's last disclosed valuation is $2.0 billion post-money, established when the company announced $405 million raised across two consecutive rounds in August 2025. The most recent round raised $314 million, oversubscribed and co-led by Bezos Expeditions, Prysm Capital, and Temasek; the preceding round raised approximately $91 million at a ~$500M valuation. The company thus quadrupled its valuation in roughly twelve months. GeekWire and Reuters sources describe the two rounds as "Series A and A1." Some pre-diligence references describe the breakdown as $150M (seed) plus $255M (Series A), but this is inconsistent with Reuters' reporting of a $314M latest round and cannot be confirmed from primary sources; this chapter uses the canonical figures from sibling chapters. PremierAlts (accessed May 2026) shows $506.1M total funding across five rounds with a last round of $315M closed in February 2026—potentially reflecting post-announcement secondary SPV activity or a new primary round—at the same $2.0B valuation mark. Form D filings at the SEC confirm investment vehicles "HII Field AI Series I, II, and III, a Series of HII Field AI, LLC," filed in November 2025 and April 2026 (CIKs 0002095306, 0002095305, 0002126354), consistent with special-purpose pooled investment fund structures that commonly aggregate secondary and LP interest in high-profile private companies. No Form D has been filed by "Field Ai, Inc." directly under the California entity, consistent with the company relying on an exemption or using a different legal entity structure. Secondary-market activity is active: FieldAI shares trade on the Nasdaq Private Market and at Notice.co (~$34.32/share as of May 2026), and UpsideList lists the company with limited upside signal, suggesting secondary pricing is at or near the $2B headline valuation. PitchBook (archived June 2025) shows 47 employees and "Early Stage VC" deal type, reflecting data lag common to PitchBook's private-company coverage. The company's own disclosures of 201–500 LinkedIn employees and >$100M booked revenue (April 2026) suggest material growth since the PitchBook snapshot. [CV001, CV002, CV003, CV004, CV005, CV006]

8.2 Investment Thesis and Anti-Thesis

The investment thesis for FieldAI rests on five interlocking pillars. First, the embodied AI software layer is structurally positioned to capture a large share of enterprise value as industrial robot hardware commoditizes: a hardware-agnostic software brain with network effects from federated learning across diverse deployments is a natural monopoly attractor. Second, the founding team—Ali Agha (NASA JPL/DARPA SubT winner), Shayegan Omidshafiei (DeepMind/MIT), and David Fan (DARPA RACER/JPL)—has credible prior-art proof that the approach works in field conditions, not merely in simulation. Third, the investor syndicate (Bezos Expeditions, Khosla Ventures, NVentures/NVIDIA, Temasek, Intel Capital, Gates Frontier) is arguably the highest-conviction collection of strategic and financial investors assembled around a robotics AI company, and they have conducted private diligence not available externally. Fourth, the company exited stealth with already-live production deployments across hundreds of sites on three continents, and reported >$100M in booked revenue within nine months of the announcement—faster commercial traction than any comparable embodied-AI software company. Fifth, the total addressable market across construction, energy, mining, logistics, and defense is large and underpenetrated. The anti-thesis challenges each of those pillars. The hardware-agnosticism advantage could erode if robot OEMs vertically integrate their own AI software (Boston Dynamics, Figure AI, Agility Robotics), or if hyperscalers (Google/Intrinsic, NVIDIA Cosmos) commoditize the foundation model layer. Revenue numbers remain unverified: the $100M booked revenue figure is from anonymous sources and "booked" may include pipeline or uncommitted orders; the $140M ARR estimate is from a third-party analytics platform with undisclosed methodology; a separate estimate puts revenue as low as $5M, a 28x discrepancy. Governance and financial transparency are near-zero for an external investor: no audited financials, no Form D from the operating entity, no disclosed gross margin, CAC, or NRR. The $2B valuation prices in substantial execution: at even $140M ARR, the 14x revenue multiple requires high margin, high retention, and continued rapid growth to be justified as a secondary entry. Veteran investor Howard Morgan (First Round Capital/B Capital) has publicly warned that "buy high, sell higher works only inside a bubble," and Seattle-area VCs in a December 2025 GeekWire roundtable observed that private-market valuations at seed and Series A stages have run far ahead of fundamentals. Multiple compression in Series B+ AI software is a real risk in the 2026–2027 environment. [CV009, CV010, CV011, CV012, CV013, CV014]

8.3 Comparable Valuations and Market Context

The embodied-AI and industrial-robotics software sector provides a useful but imperfect comparable set. Within the embodied-AI software peer group, Physical Intelligence is in talks for $11B at near-zero disclosed revenue (March 2026)—an extreme outlier driven by research prestige and optionality value, not current commercial traction. Skild AI raised at $14B in January 2026 with $30M ARR growing exponentially, implying a ~467x forward multiple on a small base—again, an early-stage optionality multiple rather than a fundamental one. These comparables confirm that the investor community is placing high optionality premiums on embodied-AI software platforms, but the wide range ($14B–$5.6B at near-zero revenue) reflects enormous uncertainty in terminal-value assumptions. FieldAI's $2B at $100–140M ARR (if those estimates are correct) actually implies a more moderate multiple (14–20x) than its two closest embodied-AI software peers, making it look attractively priced within the category if revenue is real. However, if the true revenue figure is $5–20M, the implied multiple (100–400x) would be among the most aggressive in the sector. The hardware-integrated comparables—Apptronik ($5.3B, humanoid hardware+software) and Gecko Robotics ($1.25B, critical infrastructure inspection) — are structurally different business models with lower software margins but more verifiable revenue, pulling comps downward for a pure-software valuation. Houlihan Lokey's Q1 2026 report on AI vertical software found that AI-native software companies commanded a median EV/Revenue of 21.2x in VC rounds, versus 8.5x for AI-enabled software and 5.5x for legacy SaaS. At $100M ARR and 21x, FieldAI would support a ~$2.1B valuation—broadly consistent with its current mark. The key risk is that this median applies to companies with verified financials, whereas FieldAI's revenue is entirely unverified externally. [CV017, CV018, CV019, CV020, CV021, CV022]

8.4 Bull, Base, and Bear Scenarios

The bull case requires that FieldAI's revenue is at the high end of estimates ($140M ARR growing at 80–100% year-over-year), that gross margins on the software licensing stream approach 70–80%, that NRR exceeds 120% driven by fleet expansion, and that the company closes a Series B in 2027 at 15–20x forward ARR on a $350–500M ARR base—implying a $5.25–10B valuation and a 2.6–5x return from the $2B mark. The key risks to this scenario are that the revenue figure is unverified, the growth rate is not independently confirmed, and the market environment needs to remain supportive of high AI-software multiples. The base case assumes $100M ARR, 40–60% growth, blended gross margins in the 55–70% range, and NRR of 110–120%. On a 2027–2028 Series B at 10–12x forward ARR on a $200–300M ARR base, the valuation would be $2.0–3.6B—a 0–1.8x return from the $2B mark. This implies roughly flat to modest appreciation, appropriate for the risk level. The bear case assumes revenue is materially lower ($20–50M ARR), growth slows due to sector headwinds (enterprise adoption friction, macro capex pullback), and multiple compression in AI software drags the implied valuation to $0.4–1.0B at the next round—a 50–80% loss from the $2B mark. Down-round risk is highest in this scenario. Adverse triggers include: a robotics sector-wide valuation reset if Physical Intelligence or Skild AI raise at lower-than-expected multiples; FieldAI customers reducing deployments or not renewing; or evidence that "booked revenue" significantly overstates recognized ARR. [CV025, CV026, CV027, CV028, CV029, CV030]

8.5 Recommendation, Kill Triggers, and Final Diligence Asks

The recommendation is TRACK (medium confidence, high risk, stretched valuation). The investment thesis is structurally sound: embodied-AI software with physics-first architecture and field-proven deployments, elite founder-market fit, and the strongest possible investor syndicate. But no external investor can underwrite the $2B price without access to audited financials, verified ARR, and disclosed gross margin—the information asymmetry is the dominant risk, not the technology or team quality. At the current $2B mark, the entry price is priced for execution on the high end of the revenue range; any downward revision to revenue or a sector-wide multiple compression would result in material dilution. This recommendation would upgrade to BUY if three conditions are met: (1) audited or independently verified ARR of ≥$100M with ≥40% year-over-year growth is disclosed; (2) blended gross margin of ≥60% is confirmed; and (3) net revenue retention of ≥110% is evidenced by cohort data. Absent these, TRACK remains appropriate: monitor for company-disclosed financial milestones, Series B announcement and implied valuation step-up, strategic partnership revenue disclosures, and customer concentration risk signals. The recommendation would downgrade to AVOID if: (1) a Series B is announced at a valuation below $2B (down-round); (2) a major customer departure or deployment failure becomes public; or (3) the reported >$100M booked revenue is materially revised downward. Governance is a notable gap: board composition is undisclosed, liquidation preferences and option pool size are unknown, and no independent financial reporting exists. For any potential investor, these must be diligenced before any commitment. The company's strategic positioning as the "software brain" layer of industrial robotics is genuinely differentiated and difficult to replicate quickly, but the combination of governance opacity and valuation stretch warrants a disciplined TRACK posture until commercial proof points are publicly verifiable. [CV031, CV032, CV033, CV034, CV035, CV036]

8.6 Exhibits