Skild AI
Pittsburgh-headquartered embodied-AI company building “Skild Brain” — a general-purpose robotics foundation model designed to be “omni-bodied” (one brain, any robot, any task) — deployed initially on Foxconn × NVIDIA Blackwell assembly lines and through strategic robotics-platform partnerships with ABB Robotics, Universal Robots and Mobile Industrial Robots.
The Business
Skild AI builds “Skild Brain” — a general-purpose robotics foundation model designed to be “omni-bodied” (one brain, any robot, any task) — the central product thesis being that a single learned model trained across diverse robot embodiments, tasks and environments generalises faster than the per-vendor task-specific software stacks that have dominated industrial robotics to date. The model is positioned as hardware-agnostic infrastructure that ships across robot-platform partners rather than as a vertically-integrated robot product. Founded in 2023 in Pittsburgh by CMU Robotics Institute professors Deepak Pathak (CEO) and Abhinav Gupta (President), the company emerged from stealth on Sequoia Capital’s founding-round partnership and has scaled through Lightspeed-, SoftBank- and NVIDIA-led subsequent rounds. The headline product surfaces are the Skild Brain foundation model itself, the NVIDIA Isaac-anchored training and deployment compute stack, and the integration tooling that lets the same model run across ABB Robotics industrial robot arms, Universal Robots collaborative robots, Mobile Industrial Robots autonomous mobile robots and Foxconn manufacturing-line systems.
Customers and Distribution
Skild’s distribution sits across two tracks: robot-platform partnerships and direct manufacturing-customer deployment. The platform-partnership track anchors on ABB Robotics, Universal Robots and Mobile Industrial Robots, with the joint deployment programme announced in March 2026 positioning Skild Brain as a cross-vendor intelligence layer running on top of three of the most-installed industrial-robot, collaborative-robot and autonomous-mobile-robot fleets in the world. The direct-deployment track anchors on Foxconn, where Skild Brain runs on NVIDIA Blackwell GPU assembly lines at Foxconn’s Houston facility — documented as the first commercial use of a generalised physical-AI system per Technical.ly coverage. Beyond Foxconn, Skild has disclosed that deployments exist across warehousing, construction, inspections, logistics, manufacturing and data centres but has not publicly named the customers. Annualised run-rate revenue triangulates to approximately $30M through 2025 per Sacra coverage. Headcount sits in the approximately 50-100 range per LinkedIn disclosure, with the research stack anchored on the CMU Robotics Institute. The compute and channel layer runs through NVIDIA — training infrastructure on NVIDIA GPUs, deployment on NVIDIA Isaac, NVentures as a lead Series C investor, and the Foxconn deployment on Blackwell.
Model Strategy
Skild’s strategic bet is that a single general-purpose foundation model trained on diverse embodiment data — the “omni-bodied” thesis — generalises across tasks and robot hardware faster than per-vendor task-specific stacks, and that the resulting platform can be shipped to industrial-robot OEMs as the cross-vendor intelligence layer rather than competing with them on hardware. The infrastructure backstop is NVIDIA-anchored compute alignment — training on NVIDIA GPUs, deployment on NVIDIA Isaac, headline customer deployment on NVIDIA Blackwell, and NVIDIA NVentures as Series C lead investor. The brand positioning is “reindustrial revolution” — a deliberate frame that positions Skild as foundation-model infrastructure for the broader robotics industry rather than as a humanoid-robot company (the structural differentiation from Figure AI and 1X) or a captive in-house robotics-AI stack (the structural differentiation from Amazon Robotics post-Covariant, Tesla Optimus and Google DeepMind RT/Gemini Robotics). The closest direct ideological peer on technical thesis is Physical Intelligence; the closest cohort peer on foundation-model methodology applied to physical-world embodiment is Wayve in the AV vertical.
At A Glance
The Numbers
Annualised revenue
Headcount (FTE)
Funding to date
Leadership Team
Pittsburgh-headquartered with the founding research stack anchored on the CMU Robotics Institute; deep academic-faculty bench across the founders’ lab groups. CFO, CRO and CTO roles have not been publicly disclosed as separate named appointments at time of writing — the founder-led structure remains the principal operating spine. Headcount sits in the approximately 50-100 range per Skild AI LinkedIn disclosure as of May 2026, materially smaller than the humanoid-hardware-focused cohort (Figure AI, 1X) reflecting the software-foundation-model focus rather than vertically-integrated robot manufacturing.
IM Framework Scoring
IM’s structured assessment of Skild AI’s competitive position. The summary below is the headline; expand “Show the full analyst-grade analysis” near the bottom for the per-dimension reasoning and evidence. Methodology →
Funding History
| Date | Round | Raised | Post-money | Lead investor(s) |
|---|---|---|---|---|
| Jan 2026 | Series C | $1.4B | $14.0B | SoftBank, NVIDIA (NVentures) |
| 2025 | Series B | $500M+ | $4.7B | SoftBank Group (Lightspeed, Coatue participating) |
| 2024 | Series A | $300M | $1.5B | SoftBank Group (Lightspeed, Coatue participating), SoftBank, Bezos Expeditions, Sequoia |
| 2023 | Seed | — | — | Sequoia (founding round) |
Cumulative external equity above $1.83B+ through the January 2026 Series C, one of the largest robotics fundraises on the public record. The Series C drew the deepest strategic-investor list in the robotics-foundation-model cohort — SoftBank and NVIDIA (NVentures) co-led; Jeff Bezos (Bezos Expeditions), Samsung, LG, Lightspeed, Sequoia and Coatue participated. NVIDIA is simultaneously a lead investor, the training-compute supplier, and a deployment-channel partner via the Foxconn Blackwell assembly-line collaboration — structural alignment but also concentration. Round dates and lead-investor identities are taken from Skild AI’s own press disclosures and corroborated by named-author TechCrunch, BusinessWire and Crunchbase News coverage.
Competitive Landscape
Skild AI’s competitive set sits in three concentric rings: the pure-play robotics-foundation-model cohort (Physical Intelligence as the closest ideological peer, plus the broader frontier-research robotics-FM bench), the humanoid-cohort hardware-plus-software stacks (Figure AI, 1X) that compete for the same industrial-deployment budgets, and the captive in-house robotics-AI builders (Amazon Robotics post-Covariant, Tesla Optimus, Google DeepMind RT/Gemini Robotics) that absorb potential customers into their own software stacks. Wayve sits as the AV-vertical cohort peer for foundation-model methodology comparison. Skild is unusual in the set because it positions itself as hardware-agnostic foundation-model infrastructure — “Skild Brain” running across ABB industrial robot arms, Universal Robots collaborative robots, Mobile Industrial Robots autonomous mobile robots and Foxconn manufacturing-line systems on the same model.
| Competitor | Positioning | Distribution edge | Threat profile |
|---|---|---|---|
| Physical Intelligence (PI) | San Francisco-headquartered robotics foundation-model company; the π0 and π0.5 models position PI as the most direct ideological peer to Skild Brain on hardware-agnostic, general-purpose robotics intelligence. Founded by Sergey Levine, Karol Hausman and a senior research cohort from Google Brain and Berkeley. | Foundation-model licensing across multiple humanoid and industrial-robot platforms; partnerships disclosed across the humanoid and warehousing cohort. | High — the closest direct substitute to Skild Brain on technical philosophy (general-purpose, hardware-agnostic, foundation-model first) and on customer profile (robot OEMs and industrial deployers). |
| Figure AI | Humanoid-robot company building vertically-integrated humanoid hardware (Figure 02) plus the Helix vision-language-action model. The closest humanoid-cohort competitor on the foundation-model layer, though structurally a hardware-plus-software play rather than a pure foundation-model layer. | BMW Spartanburg humanoid deployment (disclosed flagship customer); direct sales into manufacturing and warehousing customers; vertical hardware-plus-software stack. | High — the most-watched humanoid-deployment story in the cohort and a direct customer-budget competitor at the largest industrial deployers. |
| 1X Technologies | Norwegian-founded humanoid-robot company (NEO Gamma, EVE) with OpenAI as an early strategic investor; vertically integrated humanoid hardware plus learned control. Consumer-home and light-industrial deployment focus. | Direct robot sales plus pilot deployments; OpenAI strategic alignment as a model-research and capital partner. | Medium-high — vertical-integration logic differs from Skild’s hardware-agnostic platform play, but overlaps on the humanoid-cohort customer base and on the broader frontier-robotics narrative. |
| Covariant (Amazon (NASDAQ: AMZN) — talent and technology acquired Aug 2024) |
Berkeley-founded warehousing-robotics foundation-model pioneer; RFM-1 robot foundation model. Amazon hired the co-founders and licensed the technology in August 2024, effectively folding Covariant’s foundation-model capability into Amazon Robotics’ captive stack. | Amazon Robotics captive deployment surface across Amazon fulfilment-centre robotic-arm fleet; the largest single warehousing-robotics deployment in the world. | High and asymmetric — Amazon Robotics is the largest single warehousing-robotics customer in the world and, post the Covariant transaction, an in-house robotics-foundation-model builder rather than a Skild Brain candidate customer. |
| Wayve | London-headquartered embodied-AI company building an end-to-end neural-network “AI Driver” foundation model for autonomous driving — the closest cohort peer to Skild on foundation-model methodology applied to physical-world embodiment, but vertical to AV rather than to general robotics. | OEM-design-win lattice (Nissan ProPILOT FY2027, Mercedes-Benz, Stellantis) plus Uber robotaxi deployments in London 2026 and Tokyo late 2026. | Low-medium — not a head-to-head competitor on robot OEM channels, but a credibility benchmark on the broader “foundation models for embodied AI” investment frame and on capital-raised cadence. |
Valuation benchmark: Skild AI at $14B+ post the January 2026 Series C is the largest robotics-foundation-model private valuation in the pure-play cohort (Physical Intelligence and Figure AI also at multi-billion-dollar private valuations with different round structures). The Amazon-Covariant talent-and-licence transaction in August 2024 set the principal acquisition comparable in the warehousing-robotics-FM lane. Wayve at $8.6B post the February 2026 Series D is the closest cohort-peer comparable on capital position and foundation-model methodology, though vertical to AV. The competitive frame is foundation-model generality and robot-platform-design-win cadence rather than headline industrial-robot unit count today.
Potential Risks
The case for Skild AI at IM Framework 7.45 rests on the deepest strategic-investor alignment in the pure-play robotics-foundation-model cohort (SoftBank, NVIDIA, Bezos, Samsung, LG, Lightspeed, Sequoia, Coatue), the first named commercial deployment of a generalised physical-AI system (Foxconn × NVIDIA Blackwell), and a hardware-agnostic Skild Brain platform thesis that converts model generality into platform optionality across the major industrial-robot OEMs. The case against splits into five risks of differing magnitude — with NVIDIA concentration the most structural and revenue-ramp-vs-valuation the most active.
NVIDIA supplier and channel concentration
Skild’s training-compute, the headline Foxconn × Blackwell deployment, and the lead-investor cap-table position all run through NVIDIA. The bull case is structural alignment — NVentures capital, NVIDIA Isaac compute platform, NVIDIA-anchored Foxconn deployment, and a credible roadmap on AV-grade GPU economics for industrial robotics. The bear case is single-stack concentration: a strategic shift inside NVIDIA (Isaac prioritisation, custom-silicon allocation, Foxconn deployment cadence) or a re-pricing of NVIDIA GPU supply would propagate through Skild’s training, deployment and distribution surfaces simultaneously. The D4a supplier-diversity sub-rubric was held at 4 on this evidence; NVIDIA is the single most-watched external variable in the thesis.
Revenue ramp lags valuation
Skild reportedly grew from $0 to approximately $30M annualised through 2025 per Sacra and named-press triangulation — a credible early-commercial signal for a pre-revenue frontier-robotics-FM company but materially smaller than the $14B valuation implies. The Foxconn anchor (documented as the first commercial use of a generalised physical-AI system per technical.ly coverage), the ABB / Universal Robots / Mobile Industrial Robots collaboration announced March 2026, and the existing deployments across warehousing, construction, inspection, logistics, manufacturing and data centres are the commercial-validation milestones. Skild has not publicly named the other deployment customers. Slippage on commercial-customer expansion through 2026 compresses the bridge from research-stage to platform-scale revenue directly.
Key-person concentration on the two CMU co-founders
Both founders — Pathak as CEO and Gupta as President — are tenured CMU Robotics Institute professors and the public-facing voices on the Skild Brain technical thesis. The academic-credibility anchor is real and was a structural input to Sequoia’s founding-round partnership decision and to the SoftBank / NVIDIA Series C lead position. The risk concentration is that no deep senior executive bench is publicly disclosed beyond the founders and Christopher Bruzzo at COO. The D4e key-person sub-rubric was held at 5 on this evidence; dual-founder structure prevents single-person risk but does not de-risk the founders-as-a-pair concentration.
Regulatory exposure across general-purpose physical AI in unstructured environments
Skild operates under the ANSI/A3 R15.06-2025 Part 3 industrial-robot safety standard (completed January 2026), the EU AI Act plus the EU Machinery Regulation dual-applicability frame for robotics deployments, and the EU Robotics Skills Qualification Centres piloting in 2026. General-purpose physical-AI deployment in unstructured environments (warehouses, construction sites, inspection routes, data centres) sits in a regulatory grey zone for embodied autonomy and product-liability allocation. The D4c regulatory-exposure sub-rubric was held at 5 on this evidence; regulatory clarification across the major operating regimes is a forward variable on the thesis.
Foundation-model technical thesis is contested by humanoid-cohort and captive-FM competitors
Skild’s “omni-bodied” thesis — one foundation model generalising across robot hardware, tasks and unstructured environments — is contested by both ends of the robotics-FM cohort. Physical Intelligence pursues the same hardware-agnostic foundation-model frame and is the closest pure ideological substitute. Figure AI and 1X argue that vertical hardware-plus-software integration is the path to humanoid-deployment scale. Amazon Robotics (post-Covariant), Tesla Optimus and Google DeepMind RT/Gemini Robotics absorb potential customers into captive in-house foundation-model stacks. The technical thesis is credible and academically well-published, but the in-market validation cycle (Foxconn deployment expansion, ABB / Universal Robots / Mobile Industrial Robots design-win cadence, additional named customers) remains forward.
Recent IM Coverage
- IM AI Tracker — Skild AI entry May 2026.
Show recent press coverage of Skild AI
- Mar 2026 — Skild AI expands generalised robot intelligence across industries with ABB Robotics, Universal Robots and NVIDIA.
- Mar 2026 — Skild AI and NVIDIA deploy generalised robot intelligence on Foxconn Blackwell GPU assembly lines.
- Mar 2026 — Skild AI deploys general-purpose AI in Foxconn, ABB and Universal Robots for US manufacturing.
- Jan 2026 — Robotic software maker Skild AI hits $14B valuation in $1.4B Series C.
- Jan 2026 — Skild AI raises $1.4B, now valued over $14B — SoftBank, NVIDIA NVentures, Bezos, Samsung, LG join.
- Jan 2026 — Robotics startup Skild AI triples valuation in latest funding round.
- 2025 — Reindustrial revolution — Skild AI on the omni-bodied robotics foundation-model thesis.
- 2024 — Partnering with Skild — Sequoia Capital on the founding-round thesis.
Curated feed of named-source coverage — Skild AI’s own blog and press disclosures, named-author tech and business press (TechCrunch, BusinessWire, GlobeNewswire, Crunchbase News, Technical.ly, IndexBox), and partner-issued press releases (ABB Robotics, Universal Robots, NVIDIA, Foxconn). Excludes wire-aggregator reposts and unsourced robotics round-up pieces.
Show the source register for the figures on this page
IM operates a primary-source-where-possible discipline. The figures above come from:
- Revenue — basis-disclosure note: Skild AI is private and does not disclose GAAP revenue. The relevant disclosure basis for a pre-commercial-scale robotics-foundation-model company is named-customer deployment plus triangulated annualised run-rate. Skild reportedly grew from $0 to approximately $30M annualised through 2025 per Sacra coverage and corroborating named-press triangulation; named customers include Foxconn (Blackwell GPU assembly line deployment per Technical.ly) plus ABB Robotics, Universal Robots and Mobile Industrial Robots platform partnerships per GlobeNewswire. Skild has stated additional deployment customers exist across warehousing, construction, inspections, logistics, manufacturing and data centres but has not publicly named them. We decline-to-publish a precise GAAP-equivalent revenue figure pending a primary disclosure; we use named-customer-deployment scale as the headline usage proxy.
- Usage — robot-platform deployment lattice: Skild’s disclosed deployment surface as of May 2026: Foxconn Blackwell GPU assembly line deployment in Houston (documented as the first commercial use of a generalised physical-AI system per Technical.ly); ABB Robotics, Universal Robots and Mobile Industrial Robots platform collaboration announced March 2026 per GlobeNewswire; existing deployments disclosed across warehousing, construction, inspections, logistics, manufacturing and data centres (customer names withheld by Skild). The “omni-bodied” thesis (one brain, any robot, any task) is articulated in Skild’s Reindustrial Revolution blog post.
- Headcount: Approximately 50-100 employees as of May 2026 per Skild AI LinkedIn company page. Pittsburgh-headquartered with the founding research stack anchored on the CMU Robotics Institute; headcount is materially smaller than the humanoid-hardware-focused cohort (Figure AI, 1X) reflecting the software-foundation-model focus rather than vertically-integrated robot manufacturing.
- Funding to date: Cumulative external equity above $1.83B+ through the January 2026 Series C. References: BusinessWire, Series C $1.4B at $14B+ valuation; TechCrunch, Series C corroboration; Crunchbase News, valuation tripling; Sequoia Capital, founding-round partnership post.
Methodology & Disclaimer
For metric definitions, source-tier hierarchy, and decline-to-publish rules, see the tracker methodology. Confidence dots (• green / • amber / • red) follow the same convention as the AI Tracker.
Spotted a figure you believe is wrong? Send corrections to info@informationmatters.net.
Information Matters Framework scores are the considered opinion of the IM team — human and AI — applied to publicly-available evidence under a disclosed methodology. They are not statements of fact about the companies scored and they are not investment advice.
