Physical AI Market Report: October 2026

On 24 September 2026, the International Federation of Robotics released World Robotics 2026: factories installed about 603,000 industrial robots in 2025 (+11%), and global operational stock crossed 5.079 million units (+9%). That is the first year above 600,000 installs and the first public five-million stock mark.

The same September window delivered three humanoid signals that change diligence, not just demos. Figure published Helix 2.5 zero-shot results across 30 unseen homes (17 September). Boston Dynamics opened the Robotics Metaplant Application Center at Hyundai's Georgia Metaplant and restated a plan that begins with 25,000 Atlas units across Hyundai Motor and Kia plants (21 September). Agility unveiled Digit 5 after more than 65,000 Digit 4 operating hours and more than $300 million in conditional multi-year Digit 5 orders (15 September).

This is the October 2026 edition of Inside Deep Tech's Physical AI Market Report series, continuing after the September 2026 report. Below: five takeaways, four snapshot tables with sources, segment deep-dives, named movers, honest limits, Inside Deep Tech's take, methodology, and what to watch in November.

Key Takeaways

• IFR World Robotics 2026 (24 Sep): ~603,000 industrial installs in 2025 (+11%); operational stock 5.079 million (+9%); China 354,000 (59% of global).
• Figure Helix 2.5 (17 Sep): Index pretraining lifted zero-shot home success from 9% to 56% across 30 unseen homes; Nscale still stands at $3.5B initial Vera Rubin compute.
• Boston Dynamics RMAC opened 21 Sep at Hyundai HMGMA; Group plan begins with 25,000 Atlas units at Hyundai/Kia plants; U.S. factory cited at 30,000 robots/year.
• Agility Digit 5 (15 Sep): Digit 4 logged >65,000 hours; GXO hit 100,000 totes; >$300M multi-year Digit 5 orders as of May 2026 (milestone-conditioned); early access H1 2027.
• Interact Analysis (17 Jun 2026): industrial robot shipments 549,555 in 2025 (+5.1%) toward 761,303 by 2030; cobot shipment CAGR 17.4%. Mobile robots still on a 19% CAGR path to $14B by 2030.

October 2026 snapshot tables

Numbers below are dated and sourced. Where IFR and Interact Analysis differ (install census vs shipment model), both are shown. Treat research-house forecasts as research-house views, not filings.

Table 1. September signals: models, factories, and orders

Source: Figure Series C; Figure × Nscale; Helix 2.5; Boston Dynamics RMAC; Agility Digit 5; Apptronik Series A-X. Agility's $300 million order figure is company-stated as of May 2026 and subject to contractual milestones. Hyundai's 25,000-unit plan is a Group deployment intention stated in the Boston Dynamics RMAC release, not audited deliveries. Unitree shipment and mix figures: see Unitree H2 Plus / NVIDIA reference announcement and the STAR listing coverage cited in Inside Deep Tech's Unitree filing post.

Table 2. IFR World Robotics 2026 industrial census

Here's why that matters. September's humanoid headlines now sit against a classical industrial base that just cleared five million units. Physical AI software that cannot attach to that base is selling a different product than warehouse AMRs and cobots already clearing cash.

Table 3. Interact Analysis industrial + mobile + cobot forecasts

IFR's ~603,000 installs and Interact's 549,555 shipments are different methodologies. This edition does not invent a bridge. Use IFR for the authoritative install census and Interact for shipment/CAGR product-structure views.

Table 4. Named humanoid / platform comparison (October update)

Source: Figure Helix 2.5; Figure Series C ($39B); Agility Digit 5; Boston Dynamics RMAC; Unitree / NVIDIA H2 Plus.

Start with last month's baseline, then layer September's new primary sources:

Physical AI Market Report: September 2026

September edition: Figure×Nscale $3.5B, Unitree STAR IPO, AMR/cobot forecasts, and IFR 2024 install baseline.

Inside Deep Tech

Segment deep-dives

Let's start with the four buckets that actually move budgets in October 2026: humanoids, warehouse AMRs, industrial robots and cobots, and foundation models / software platforms.

Humanoids: generalization claims meet factory calendars

Humanoids still own the narrative bandwidth. October's diligence question is sharper than August's: can published generalization metrics and named factory programs survive outside company blogs?

Figure's 17 September Helix 2.5 post is the clearest company-published zero-shot evaluation this cycle. A single Index-pretrained model produced three whole-body behaviors (tidying living rooms, folding towels, making beds) and ran them in 30 Bay Area homes with zero data collected in those homes. Holding task data, architecture, and evaluation fixed, Index pretraining alone moved full-task success from 9% (trained from scratch) to 56%. Figure also reports a human-to-humanoid transfer scaling law across nested Index subsets spanning an 8× data range, with forecasting error of 0.54% of the variation across that range. That is a model claim, not a factory utilization claim.

The compute side is unchanged from early September: up to 100,000 Vera Rubin GPUs via Nscale, initial $3.5 billion commitment, intent past $6 billion, Barstow target in the second half of 2027. Figure still states Index is generating roughly 35 minutes of new human experience every second. Series C remains the last primary valuation mark used here: more than $1 billion committed at $39 billion post-money (16 September 2025).

Boston Dynamics moved from CES allocation language to a named facility. On 21 September 2026 the Robotics Metaplant Application Center opened inside Hyundai Motor Group Metaplant America. Atlas robots are training on logistics and sequencing of automotive parts. The Group restates a plan that begins with 25,000 units at Hyundai Motor and Kia global plants over the next few years, plus a U.S. production facility capable of 30,000 robots per year. Component assembly is targeted by 2030. RMAC phase one is operational; a 2027 building is described as roughly ten times larger. That is a serious industrial calendar. It is still not 2026 mass labor replacement.

Agility is the company with the longest public industrial operating clock among bipedal humanoids in this edition. Digit 4 has logged more than 65,000 hours across sites including GXO, Schaeffler, Amazon, and Toyota Motor Manufacturing Canada. At GXO Flowery Branch, Digit reached a cumulative 100,000-tote milestone with about 98% accuracy while on-task. Digit 5 (15 September) adds a cooperative safety architecture, up to 50 lb (22.7 kg) repeated lifts, a 90-minute battery with 9-minute charge (10:1 run-to-charge), and plans for EU/UK CE-marked commercial access. Early access is guided for the first half of 2027; general availability by end of 2027. As of May 2026, Agility cites more than $300 million in multi-year Digit 5 orders subject to contractual milestones. Agility is also the first launch partner for NVIDIA's Halos for Robotics platform on IGX Thor.

Unitree remains the only mainland public humanoid issuer with an audited mix Inside Deep Tech has broken down in detail: more than 5,500 humanoid shipments in 2025, and industrial application at 9.01% of humanoid revenue in the first nine months of 2025, with guided tours roughly half to seven tenths of that industrial slice (see also Unitree's H2 Plus / Isaac GR00T reference announcement). The STAR listing on 19 August 2026 still matters more for diligence than another walking reel.

Apptronik's capital mark is unchanged: $520 million Series A-X on 11 February 2026, Series A total above $935 million, Bloomberg-reported valuation above $5.5 billion (Apptronik Series A announcement). Named partners still include Mercedes-Benz, GXO, Jabil, and a Google DeepMind Gemini Robotics collaboration.

See also Inside Deep Tech's Figure AI deep dive and Unitree filing breakdown.

Warehouse AMR: where the dollars already move

Here's the problem with reading Physical AI only through humanoid demos. The mobile-robot ledger is already a multi-billion-dollar market with named research-house forecasts and visible order books.

Interact Analysis (9 January 2026) still forecasts mobile robot revenue from just under $5 billion in 2024 to $14 billion in 2030 at a 19% average annual growth rate, after a 12% downward revision on tariffs. Fixed automation's predicted CAGR in the same window is 2.4%. AGV revenue share is expected to fall from about 33% (2024) to about 20% (2030) as AMRs take share. Order fulfillment robots are forecast at about half of shipments by 2030. Mobile forklifts remain the high-ASP wedge: about one-third of revenue by 2030 on only about 14% of shipments.

Warehouse automation order intake rose 7% year on year in 2025 (Interact, 23 February 2026). Interact's global warehouse automation CAGR for 2025–2030 is about 6%, with EMEA (~7%) ahead of the Americas (~6%) and APAC (~5%). Growth stayed concentrated among mega-customers in 2025. Whether 2026 demand broadens is still a testable vendor-commentary question for the rest of this year.

Industrial robots and cobots: the five-million-unit attach surface

Of course, most factory floors still run on classical industrial robots and a growing cobot layer. That install base is the distribution channel for simulation, virtual commissioning, and foundation-model control stacks.

IFR's World Robotics 2026 industrial report (press dated 24 September 2026) recorded about 603,000 industrial robot installations in 2025, up 11%, after four years stuck near 550,000. The IFR foreword cites 603,307 units (+11.2%). Operational stock reached 5.079 million units (+9%). China installed 354,000 units (+20%), 59% of the global total, with domestic suppliers at 195,000 units and a 55% home share. The United States installed almost 38,500 units (+12%) and became the second-largest market, ahead of Japan (36,219, −19%). Korea installed 30,000 (−1%). India installed about 10,500 (+15%). Germany fell 8% to fewer than 25,000. IFR's outlook: 655,000 installs in 2026 (+9%) and 806,000 by 2029.

On cobots, IFR's president report says new collaborative robot installations reached 66,000 units in 2025 (+7%), equal to 11% of all industrial robot installations. For the first time since IFR began tracking cobots, cobot growth sat below the overall industrial robot market. Interact's May 2026 cobot note still shows ~57,000 shipments in 2025 (+14.5%) and a 17.3% shipment CAGR to 128,918 units by 2030. Interact's June industrial report separately projects a 17.4% cobot shipment CAGR inside its industrial model. Use the IFR cobot count for the 2025 install census and Interact for product-structure forecasts.

Interact's June industrial shipment model puts 2025 at 549,555 units (+5.1%) rising to 761,303 by 2030 (6.7% CAGR), with APAC at 7.1%, Americas 6.2%, and EMEA 4.7%. EMEA contracted 4.9% in 2025 on weak automotive investment. That means… operators evaluating Physical AI software should ask which stack attaches to the 5.079 million industrial robots already in service, not only which humanoid walks in a keynote.

Foundation models and software platforms

On the flip side, the software layer is where September's private capital and safety partnerships concentrated: data throughput, GPU runway, and cooperative safety stacks.

Figure's Helix 2.5 results are the month's primary model disclosure: Index pretraining accounts for most measured zero-shot success in the company's blind home evaluations, and the company claims a measurable human-to-humanoid transfer scaling law. That pairs with the still-live Nscale Vera Rubin compute runway.

NVIDIA's 16 March 2026 GTC release remains the industrial attach baseline: FANUC, ABB Robotics, YASKAWA, and KUKA (global install base exceeding 2 million robots) integrating Omniverse libraries and Isaac simulation into virtual commissioning; GR00T N1.7 in early access with commercial licensing; GR00T N2 previewed with a year-end target. Agility's Digit 5 announcement adds a concrete safety-stack attachment: first launch partner for NVIDIA Halos for Robotics on IGX Thor and Halos Core. Unitree's H2 Plus remains the stated Isaac GR00T reference humanoid body for late 2026 academic availability (PR Newswire, 1 June 2026).

Skild AI still appears in NVIDIA's GTC partner list with ABB, Universal Robots, and Foxconn references. SoftBank-led 2025 funding was widely reported in the trade press. Later secondary marks circulate in aggregators. This edition does not treat those press or secondary marks as company-confirmed primary valuations.

Inside Deep Tech's related primers: NVIDIA's robotics platform and robot foundation models.

Named movers of the month

At first glance, October inherits a September stack: IFR's census, Figure's model paper-equivalent, Boston Dynamics' factory door opening, and Agility's product/safety refresh. Zoom out, and the month resets what "evidence" means for Physical AI diligence.

Figure: Helix 2.5 generalization, same $3.5B compute runway

Helix 2.5 (17 September) and Nscale (3 September) together say Figure is buying transfer from human video at Index scale, then buying GPU runway to train it. The Series C $39 billion mark from September 2025 is unchanged as a primary round valuation. Home zero-shot success is not BMW station economics. Keep those ledgers separate.

Boston Dynamics / Hyundai: RMAC is open

The durable change is location. A named application center is operating at HMGMA, with sequencing as the near-term task and component assembly on a 2030 horizon. The 25,000-unit Hyundai/Kia plan and 30,000 robots/year factory citation set capacity ambition. Do not confuse RMAC training with 2026 plant-wide labor replacement.

Agility: operating hours, tote counts, and conditional orders

Agility's diligence edge is time-on-task. More than 65,000 Digit 4 hours, a 100,000-tote GXO milestone, named customers, and a Digit 5 safety architecture aimed at barrier-light cooperative work are harder for peers to hand-wave. The $300 million order book is still milestone-conditioned company guidance as of May 2026. Digit 5 early access is H1 2027. The Churchill Capital Corp XI SPAC path is noted in the same PR as a proposed transaction, not treated here as a closed public valuation.

Unitree: still the audited industrial-mix benchmark

Public listing converts marketing claims into prospectus liability. Unitree's 9.01% industrial mix and tour-heavy industrial slice remain the best audited humanoid revenue breakout available (covered with filing citations in Inside Deep Tech's Unitree post; company platform context: Unitree H2 Plus announcement). ASP compression with retained gross margins above 60% in 1H and Q3 2025 is still the filing detail operators should not ignore.

IFR + Interact: census vs shipment models

The durable market-structure update this month is IFR's five-million stock and 603k install year, not another startup round. Interact's June industrial shipment model (549,555 in 2025 → 761,303 in 2030) and still-current mobile/cobot notes remain the best public product-structure forecasts in this series.

NVIDIA: industrial attach plus Halos safety path

The durable NVIDIA claim for Physical AI diligence is still the stated integration path into FANUC, ABB, YASKAWA, and KUKA environments covering more than 2 million installed robots, plus commercial licensing on GR00T N1.7. Agility's Halos launch-partner status adds a cooperative-safety product path on IGX Thor.

Honest limits: what this data does not show

• Unitree's industrial percentage is still the best audited humanoid revenue mix available. It is one issuer in one jurisdiction. It does not describe Figure, Apptronik, Agility, or Tesla.
• Figure's Helix 2.5 home zero-shot metrics are company evaluations. They do not disclose factory utilization, cost per station, or BMW revenue.
• Figure's BMW "30,000 cars" claim remains a company statement about contribution, not a filed P&L.
• Hyundai's 25,000 Atlas plan and 30,000 robots/year factory are stated intentions in the RMAC release, not 2026 delivery tables.
• Agility's >$300M Digit 5 orders are milestone-conditioned as of May 2026. Digit 5 GA is end-2027 guidance.
• Tesla Optimus production and deployment counts remain management commentary without an audited unit table in this edition. They are omitted rather than guessed.
• IFR industrial totals are the authoritative install census for classical robots. They are not a humanoid shipment census.
• Interact Analysis dollars and CAGRs are market research forecasts, not company filings. IFR and Interact industrial unit counts are not forced into a single reconciled number here.
• Warehouse automation's +7% order intake in 2025 includes price inflation from steel and labor. Nominal order growth is not identical to unit demand growth.
• Secondary quotes and proposed SPACs are not new primary valuations.

This report is written for operators, investors, and technical leaders who need citable numbers. It is not a buy list, a valuation model, or a substitute for primary diligence.

Inside Deep Tech's take

The team at Inside Deep Tech will keep this monthly report boring on purpose: tables first, opinions labeled, and every metric cut if it cannot be sourced.

Methodology and sources

Primary preference for this edition: IFR World Robotics 2026 press materials and president report (24 September 2026), company newsrooms (Figure 3 and 17 September; Boston Dynamics 21 September; Agility 15 September), Interact Analysis press notes (9 January, 23 February, 29 May, and 17 June 2026), NVIDIA newsroom (16 March and 31 May 2026), and Unitree listing/filing materials covered in Inside Deep Tech's Unitree post. September 2026 sources are prioritized. Older figures appear only when they remain the best public authoritative number and are date-stamped inline.

Every quantitative claim in the tables and body above maps to a named source and date in Inside Deep Tech's research log for this edition. Unsourced metrics were cut. Contested or unaudited competitor shipment comparisons are not restated as fact.

Internal context used for cross-links (HTTP 200 as of 1 October 2026): September Physical AI Market Report, Unitree filing breakdown, Figure deep dive, NVIDIA robotics platform primer, robot foundation models primer, 25 fast-growing Physical AI companies, and Physical AI startups to watch in 2026.

FAQ

What to watch in November

November is for evidence that survives outside a keynote reel.

• Any customer-published KPIs from BMW Spartanburg or Hyundai RMAC that survive outside company blogs.
• Follow-through on IFR service/mobile/humanoid statistics after the 30 October webinar window.
• Digit 5 early-access readiness signals and any update to the milestone-conditioned order book.
• Ship dates and licensing terms for NVIDIA GR00T N2 after the March preview.
• Whether Interact's hoped-for broadening of warehouse automation demand beyond mega-customers shows up in vendor commentary.
• Follow-through on Figure Index throughput and the Nscale/Vera Rubin build timeline into 2027.
• Any second public issuer forced to break out humanoid industrial revenue the way Unitree did.

For the wider startup map, see Physical AI startups to watch in 2026. The next Physical AI Market Report lands in early November.