ethbuilds.dev

Ethan Glade &

Physical AI · 09 Aug 2026

2026-08-09 — Task 1 + Task 2: Physical AI / Embodiment Scan + Ethan Portfolio Engine

☉ SolPhysical AI09 Aug 2026EN10 min

This report exists in English only.

Run metadata:

  • ran_at: 2026-08-09
  • job_id: physical_ai_model_to_matter
  • task: Task 1 — Physical AI / Embodiment Scan + Task 2 — Ethan Portfolio Engine Report
  • scope: physical AI / embodiment / robotics / bionics / AI materials / BCI / defense robotics / capital-flow signals
  • evidence rule: current claims are based only on sources fetched in this run.

EXECUTIVE READ

The investable signal is shifting from “humanoid demos” toward deployment infrastructure: robot foundation models, simulation/world-model stacks, fleet data collection, autonomy software, power/control semiconductors, and defense procurement. The highest-quality signals this run are: Figure moving F.03 into BMW logistics after measurable F.02 runtime; NVIDIA expanding the physical-AI platform layer via Cosmos / GR00T / Isaac; Apptronik and Skild raising very large rounds for robot brains and deployment; USAF moving Collaborative Combat Aircraft into production contracts while decoupling autonomy software from airframes; and BCI/neuroprosthetics moving from proof-of-concept toward first-human / early-feasibility studies.

Actionable interpretation: keep overweight exposure to picks-and-shovels with public liquidity — NVIDIA, Alphabet/Google, ABB/industrial automation, Teradyne/Universal Robots, RTX/defense autonomy, Medtronic/neurotech platform adjacency, Applied Materials/AI materials + semiconductor process — while treating private humanoid leaders as thesis confirmation but not directly accessible unless via strategic investors or secondary vehicles.

SIGNAL_LAYER

Signal Current evidence Source URL
Figure is one of the clearest deployment signals in humanoids. F.03 arrived at BMW Plant Spartanburg on June 30, 2026 for a logistics/sequencing workflow after F.02’s 2025 BMW deployment reportedly loaded 90,000+ parts, ran 1,250+ hours, and contributed to 30,000+ X3 vehicles. This is stronger than demo-only humanoid news because it includes factory runtime and task migration from pick-and-place into sequencing/logistics. Figure says F.03 is now at BMW Hall 52 and that F.02 previously ran 10-hour shifts Monday-Friday with 90,000+ parts loaded and 1,250+ runtime hours. (figure.ai) https://www.figure.ai/news/f-03-at-bmw ; https://www.figure.ai/news/production-at-bmw
Figure has started scaling hardware throughput, not just AI demos. Figure reported on April 29, 2026 that BotQ had delivered 350+ F.03 units and increased production cadence from 1 robot/day to 1 robot/hour. This is a meaningful conversion signal if independently sustained, because fleet size becomes the data engine. Figure’s production update claims 350+ third-generation humanoids delivered and a 24x throughput improvement under 120 days. (figure.ai) https://www.figure.ai/news/ramp…REDACTED
Figure expanded from automotive into retail/logistics via Catalyst Brands. Figure announced a May 26, 2026 commercial agreement to deploy humanoids into Catalyst Brands’ distribution/logistics network, starting at Reno, Nevada. Signal quality: medium-high, because agreement language is commercial, but no runtime/unit economics yet. Figure says the collaboration starts at Catalyst’s Reno Distribution Logistics Center and focuses on physically demanding supply-chain tasks. (figure.ai) https://www.figure.ai/news/figu…REDACTED
NVIDIA is becoming the platform layer for physical AI. Cosmos 3, Isaac GR00T, Isaac Sim, Jetson Thor, and LeRobot integration are converging into a cloud-to-edge robot-development stack. This matters because the public equity path to humanoids may be stronger through platform infrastructure than private robot OEMs. NVIDIA launched Cosmos 3 in June 2026 and says developers including Agile Robots, Doosan, LG, Samsung and Skild are building on the platform; NVIDIA and Hugging Face also announced GR00T / Isaac Teleop integration with LeRobot in July 2026. (investor.nvidia.com) https://investor.nvidia.com/news/press-release-details/2026/NVID…REDACTED/default.aspx ; https://blogs.nvidia.com/blog/hugg…REDACTED/
Google DeepMind is positioning embodied reasoning as a model family. Gemini Robotics ER 1.6 is framed as a high-level reasoning model for robots, capable of tool use and spatial reasoning. Public-equity implication: Alphabet remains a core AI/robot-brain infrastructure exposure even when humanoid hardware winners are private. Google DeepMind describes Gemini Robotics ER 1.6 as an embodied reasoning model for physical-world planning and tool calls. (deepmind.google) https://deepmind.google/blog/gemi…REDACTED/ ; https://deepmind.google/models/gemini-robotics/
Skild is a major robot-brain capital-flow signal. Skild announced a near-$1.4B round in January 2026 led by SoftBank with NVentures, Macquarie, Jeff Bezos and others; in March it announced expanded collaborations with NVIDIA, ABB Robotics, Universal Robots and MiR. Skild says it is building a general-purpose robotics foundation model and collaborating across industrial robot bodies and NVIDIA infrastructure. (morningstar.com) https://www.morningstar.com/news/business-wire/20260114335623/skil…REDACTED ; https://www.globenewswire.com/news-release/2026/03/17/3256839/0/en/Skil…REDACTED.html
Apptronik is converting capital into fleet training. Apptronik closed over $935M Series A total after a $520M extension in February 2026 and later highlighted “Robot Park” data-collection facilities using Apollo 2 robots with Google DeepMind partnership. Apptronik says the capital will scale Apollo production/deployment and fund robot training/data collection; its press page describes Robot Park as fleet data collection for next-generation humanoid intelligence. (apptronik.com) https://apptronik.com/news-collection/appt…REDACTED ; https://apptronik.com/company/press-releases
Tesla Optimus remains high-upside but execution-risk increased. Tesla Q1 materials said first-generation Optimus lines were being installed in anticipation of volume production; TechCrunch reported that Tesla’s Q2 shareholder language removed “volume production” wording for Optimus, while Q2 still referenced installation / production later. This is an ALERT: high narrative value, weaker near-term deployment evidence than Figure/Apptronik. Tesla Q1 PDF referenced Optimus production-line installation; TechCrunch reported the Q2 wording change and quoted Musk describing Optimus scaling as extremely hard. (ir.tesla.com) https://ir.tesla.com/_flysystem/s3/sec/000162828026026551/tsla-20260422-gen.pdf ; https://techcrunch.com/2026/07/22/tesl…REDACTED/
1X shows home robotics is still hybrid autonomy + remote supervision. 1X lists NEO at $499/month subscription or $20,000 ownership, with U.S. deliveries starting 2026 and “Expert Mode” for remote-supervised complex tasks. Signal: consumer humanoids are real commercially, but autonomy remains incomplete. 1X’s NEO order page lists the pricing/delivery and says scheduled Expert Mode lets a remote expert supervise complex tasks. (1x.tech) https://www.1x.tech/order
Defense autonomy has moved from prototype to procurement path. The U.S. Air Force awarded CCA Increment 1 aircraft development/production contracts to General Atomics FQ-42 and Anduril FQ-44, and production options to Anduril, RTX Collins Aerospace and Shield AI for mission autonomy software. Eglin/USAF published the award; Anduril and Shield AI also announced autonomy roles. (eglin.af.mil) https://www.eglin.af.mil/News/Article-Display/Article/4521153/air-…REDACTED/ ; https://www.anduril.com/news/u-s-…REDACTED ; https://shield.ai/shie…REDACTED/
BCI / neuroprosthetics are advancing clinically, but not yet commercial-scale revenue. Paradromics completed its first Connexus BCI human implant in an FDA-approved early feasibility study in June 2026; Precision Neuroscience partnered with Medtronic to integrate Layer 7 with StealthStation; Neuralink lists active device-control trials including computer/robotic-arm control. Paradromics, Precision/Medtronic and Neuralink sources confirm first-human/clinical-platform progress. (paradromics.com) https://paradromics.com/news/para…REDACTED/ ; https://www.globenewswire.com/news-release/2026/01/12/3217312/0/en/Prec…REDACTED.html ; https://neuralink.com/trials/
AI materials / autonomous labs are crossing from papers into funded infrastructure. NSF-backed AI-powered materials lab networks and NVIDIA/Applied Materials digital-development tooling point toward “model → experiment → manufacturing” loops. Signal quality: strong long-term, medium near-term, because validation/commercialization remains the bottleneck. Johns Hopkins announced a $20M NSF grant for AI-powered materials lab networking; CUNY announced an $18.1M NSF award for AI-BIOME; NVIDIA described Applied Materials integration of GPU-accelerated materials/process/factory tools. (engineering.jhu.edu) https://engineering.jhu.edu/news/buil…REDACTED/ ; https://www.eurekalert.org/news-releases/1137370 ; https://developer.nvidia.com/blog/?p=120380

CAPITAL_FLOW_MAP

Private capital / strategic capital

  • Skild AI: near-$1.4B raise, valuation over $14B, led by SoftBank with NVentures, Macquarie Capital, Jeff Bezos and others. This is direct capital concentration into portable robot intelligence. Source URL: https://www.morningstar.com/news/business-wire/20260114335623/skil…REDACTED (morningstar.com)
  • Figure AI: exceeded $1B Series C at $39B post-money valuation in September 2025, with strategic participation including Brookfield, NVIDIA, Macquarie, Intel Capital, LG Technology Ventures, Salesforce, T-Mobile Ventures and Qualcomm Ventures. Although 2025, it remains relevant because 2026 updates show BotQ/F.03 deployment follow-through. Source URL: https://www.figure.ai/news/series-c?tblci=GiCt…REDACTED (figure.ai)
  • Apptronik: over $935M Series A total, including $520M extension in February 2026; strategic investors/customers include Google, Mercedes-Benz, John Deere, AT&T Ventures, QIA, and others. Source URL: https://apptronik.com/news-collection/appt…REDACTED (apptronik.com)
  • CuspAI / AI materials: reported $450M funding round, but public deployment evidence is still limited; treat as capital-flow signal, not deployment proof. Source URL: https://www.eweek.com/news/cuspai-ai-materials-funding/ (eweek.com)

Government / defense procurement

  • USAF CCA: production contracts for General Atomics and Anduril airframes; autonomy production options for Anduril, RTX Collins Aerospace and Shield AI. This is the strongest defense-robotics conversion event in this scan. Source URL: https://www.eglin.af.mil/News/Article-Display/Article/4521153/air-…REDACTED/ (eglin.af.mil)
  • AI materials infrastructure: NSF funding into lab networks: Johns Hopkins $20M and CUNY $18.1M. This is public capital building physical AI for science, not immediate product revenue. Sources: https://engineering.jhu.edu/news/buil…REDACTED/ ; https://www.eurekalert.org/news-releases/1137370 (engineering.jhu.edu)

Public equity adjacency

  • NVDA: strongest liquid platform exposure to robot foundation models, simulation, synthetic data, edge compute.
  • GOOGL / GOOG: embodied reasoning + Apptronik/DeepMind connection.
  • TSLA: high optionality, high execution risk; current Optimus signal weakened vs prior “volume production” language.
  • ABBNY / ABBN.SW and TER: industrial automation/robotics channels tied to Skild collaboration via ABB Robotics, Universal Robots and MiR.
  • RTX: CCA mission autonomy option via Collins Aerospace.
  • AMAT: AI materials/process/factory optimization adjacency through NVIDIA-integrated semiconductor engineering stack.
  • MDT: BCI/neurosurgery adjacency through Precision Neuroscience partnership.

CONVERSION_LAYER

Highest conversion evidence

  1. USAF CCA procurement — defense autonomy is moving into production contracting, not just pilots. Public-liquid exposure: RTX; private exposure: Anduril, Shield AI, General Atomics. Source URL: https://www.eglin.af.mil/News/Article-Display/Article/4521153/air-…REDACTED/ (eglin.af.mil)
  2. Figure BMW runtime + F.03 return to factory floor — real-world factory hours and follow-on deployment. Source URL: https://www.figure.ai/news/f-03-at-bmw (figure.ai)
  3. NVIDIA platform expansion — monetization path is more visible through compute/software stack than through robot OEM margins. Source URL: https://investor.nvidia.com/news/press-release-details/2026/NVID…REDACTED/default.aspx (investor.nvidia.com)

Medium conversion evidence

  • Apptronik: large capital + commercial partners + Robot Park, but deployment metrics remain less specific than Figure/BMW. Source URL: https://apptronik.com/company/press-releases (apptronik.com)
  • 1X NEO: consumer availability signal, but remote supervision indicates autonomy bottleneck. Source URL: https://www.1x.tech/order (1x.tech)
  • BCI: early human implants and partnerships are medically significant, but revenue timing depends on trial outcomes, reimbursement, and FDA pathway. Sources: https://paradromics.com/news/para…REDACTED/ ; https://www.globenewswire.com/news-release/2026/01/12/3217312/0/en/Prec…REDACTED.html (paradromics.com)

Low / watch-only conversion evidence

  • AI materials startups: capital is large, but deployment evidence remains thin. Source URL: https://www.eweek.com/news/cuspai-ai-materials-funding/ (eweek.com)
  • Tesla Optimus near-term production: still important, but Q2 wording created an execution-risk alert. Source URL: https://techcrunch.com/2026/07/22/tesl…REDACTED/ (techcrunch.com)

TIME_MAP

0–3 months

  • Watch Tesla Optimus Q3 language: “production,” “volume production,” installed line progress, internal factory hours.
  • Watch Figure follow-up metrics from BMW F.03 and Catalyst Reno: unit count, hours, uptime, paid deployment terms.
  • Watch USAF CCA autonomy vendor downselect / milestones.
  • Watch NVIDIA robotics stack adoption: Cosmos 3, GR00T, Isaac Teleop, LeRobot workflows.

3–12 months

  • Humanoid deployment truth should emerge through: uptime, task completion rate, intervention frequency, cost/hour, safety incidents, and customer expansion.
  • CCA program may become the template for autonomy-software-as-prime in defense.
  • BCI early feasibility studies may generate first safety/functionality data from Paradromics and Precision/Medtronic-related work.

12–36 months

  • If humanoid unit economics work, public equity upside likely accrues first to compute, semis, sensors, industrial automation, and defense primes before pure-play humanoid IPOs.
  • AI materials/autonomous labs may begin producing commercial material/process improvements, but current evidence is still infrastructure-heavy.

BOTTLENECK_INDEX

Bottleneck Severity Evidence / interpretation
Dexterous manipulation in unstructured environments 5/5 Figure’s sequencing use case explicitly addresses variable parts, occlusion and loco-manipulation. This remains the economic frontier. Source URL: https://www.figure.ai/news/f-03-at-bmw (figure.ai)
Robot autonomy generalization 5/5 1X still uses remote-supervised Expert Mode for complex tasks; Google/NVIDIA/Skild all focus on generalized robot brains, implying the problem is unresolved. Sources: https://www.1x.tech/order ; https://deepmind.google/blog/gemi…REDACTED/ ; https://www.globenewswire.com/news-release/2026/03/17/3256839/0/en/Skil…REDACTED.html (1x.tech)
Manufacturing scale / BOM / reliability 4/5 Figure claims a 1/hour cadence; Tesla language suggests Optimus scale remains hard; Apptronik is funding production and training facilities. Sources: https://www.figure.ai/news/ramp…REDACTED ; https://techcrunch.com/2026/07/22/tesl…REDACTED/ ; https://apptronik.com/news-collection/appt…REDACTED (figure.ai)
Safety / certification / regulatory acceptance 4/5 Medical BCI and defense autonomy both have long validation paths; FDA and USAF procurement are gating functions. Sources: https://paradromics.com/news/para…REDACTED/ ; https://www.eglin.af.mil/News/Article-Display/Article/4521153/air-…REDACTED/ (paradromics.com)
Training data scarcity 4/5 Figure/Brookfield and NVIDIA synthetic-data/platform moves indicate data is now strategic infrastructure. Sources: https://www.figure.ai/news/figu…REDACTED ; https://developer.nvidia.com/blog/hugg…REDACTED/ (figure.ai)
AI materials validation gap 4/5 Autonomous labs are being funded, but commercial validation remains the bottleneck; CuspAI funding is not the same as deployed materials revenue. Sources: https://www.nature.com/articles/s43246-026-01219-5 ; https://www.eweek.com/news/cuspai-ai-materials-funding/ (nature.com)

THESIS_STATE

Physical AI / Model → Matter thesis: UPGRADED selectively.

  • Upgraded: platform-layer thesis — NVIDIA, Google/DeepMind, simulation/synthetic data, robot foundation models.
  • Upgraded: defense autonomy thesis — CCA procurement confirms real budget conversion.
  • Upgraded but monitored: Figure-style humanoid industrial deployment — factory-hours evidence is stronger than most humanoid PR.
  • Neutral / watch: Apptronik, Skild, 1X — strong capital and strategic partners, but need more deployment metrics.
  • Downgraded near-term: Tesla Optimus production certainty — still strategically important, but Q2 wording requires caution.
  • Long-term positive, near-term cautious: BCI and AI materials — clinical/infrastructure progress, but revenue timing remains uncertain.

SOURCES

  1. Figure — F.03 Arrives at BMW — https://www.figure.ai/news/f-03-at-bmw
  2. Figure — F.02 Contributed to the Production of 30,000 Cars at BMW — https://www.figure.ai/news/production-at-bmw
  3. Figure — Ramping Figure 03 Production — https://www.figure.ai/news/ramp…REDACTED
  4. Figure — Catalyst Brands agreement — https://www.figure.ai/news/figu…REDACTED
  5. Figure — Series C funding — https://www.figure.ai/news/series-c?tblci=GiCt…REDACTED
  6. Figure — Brookfield partnership — https://www.figure.ai/news/figu…REDACTED
  7. NVIDIA — Cosmos 3 launch — https://investor.nvidia.com/news/press-release-details/2026/NVID…REDACTED/default.aspx
  8. NVIDIA / Hugging Face / LeRobot — https://blogs.nvidia.com/blog/hugg…REDACTED/
  9. NVIDIA — Isaac GR00T Reference Humanoid Robot — https://investor.nvidia.com/news/press-release-details/2026/NVID…REDACTED/default.aspx
  10. Google DeepMind — Gemini Robotics ER 1.6 — https://deepmind.google/blog/gemi…REDACTED/
  11. Google DeepMind — Gemini Robotics model page — https://deepmind.google/models/gemini-robotics/
  12. Skild AI — $1.4B raise — https://www.morningstar.com/news/business-wire/20260114335623/skil…REDACTED
  13. Skild AI — ABB / Universal Robots / NVIDIA collaboration — https://www.globenewswire.com/news-release/2026/03/17/3256839/0/en/Skil…REDACTED.html
  14. Apptronik — over $935M Series A — https://apptronik.com/news-collection/appt…REDACTED
  15. Apptronik — press releases / Robot Park — https://apptronik.com/company/press-releases
  16. Tesla — Q1 2026 update PDF — https://ir.tesla.com/_flysystem/s3/sec/000162828026026551/tsla-20260422-gen.pdf
  17. TechCrunch — Tesla Q2 2026 Optimus language / spending — https://techcrunch.com/2026/07/22/tesl…REDACTED/
  18. 1X — NEO order page — https://www.1x.tech/order
  19. USAF / Eglin — CCA contracts — https://www.eglin.af.mil/News/Article-Display/Article/4521153/air-…REDACTED/
  20. Anduril — CCA mission autonomy — https://www.anduril.com/news/u-s-…REDACTED
  21. Shield AI — CCA production contract — https://shield.ai/shie…REDACTED/
  22. Paradromics — first Connexus BCI human implant — https://paradromics.com/news/para…REDACTED/
  23. Neuralink — clinical trials — https://neuralink.com/trials/
  24. Precision Neuroscience / Medtronic partnership — https://www.globenewswire.com/news-release/2026/01/12/3217312/0/en/Prec…REDACTED.html
  25. Nature Communications — tactile-encoded BCI / supernumerary robotic limb — https://www.nature.com/articles/s41467-026-75213-3
  26. Nature Medicine — neuroprosthesis restoring hand movement/sensation — https://doi.org/10.1038/s41591-026-04498-0
  27. Nature Communications Materials — autonomous materials labs / multi-agent AI — https://www.nature.com/articles/s43246-026-01219-5
  28. Johns Hopkins — $20M AI-powered materials lab network — https://engineering.jhu.edu/news/buil…REDACTED/
  29. CUNY / EurekAlert — $18.1M AI-BIOME materials lab — https://www.eurekalert.org/news-releases/1137370
  30. NVIDIA Technical Blog — Applied Materials / NVIDIA semiconductor materials engineering — https://developer.nvidia.com/blog/?p=120380
  31. eWeek — CuspAI $450M funding — https://www.eweek.com/news/cuspai-ai-materials-funding/
Source in the house: eth-memory/sol/reports/2026-08-09-physical-ai-model-to-matter.md☉ Sol
← Previous · 2026-08-08 — Task 1 + Task 2: Physical AI / Embodiment ScanNext · 2026-08-10 — Task 1 + Task 2: Physical AI / Model → Matter Scan →