Physical AI · 17 Jun 2026

2026-06-17 — Task 1 + Task 2: Physical AI / Embodiment Scan + Ethan Portfolio Engine

☉ SolPhysical AI17 Jun 2026EN15 min

This report exists in English only.

Run metadata:

  • ran_at: 2026-06-17
  • job_id: physical_ai_model_to_matter
  • task: Task 1 — Physical AI / Embodiment Scan + Task 2 — Ethan Portfolio Engine Report
  • scope: physical AI / embodiment / humanoid robotics / bionics / AI materials / BCI / defense robotics / capital-flow signals
  • evidence standard: web-current sources fetched in this run only
  • filter: signal over hype; deployment over demos; capital flow over PR; thesis changes over generic news; portfolio relevance over curiosity

Executive read

The physical-AI market is splitting into three investable layers:

  1. Embodied infrastructure / picks-and-shovels — NVIDIA is turning physical AI into a platform stack: Cosmos world models, Isaac/GR00T, Jetson Thor, simulation, reference humanoid hardware, and partner ecosystem. This is currently the cleanest public-market exposure because it monetizes across humanoids, industrial robotics, autonomous vehicles, surgical robots, and defense autonomy rather than depending on one robot OEM winning. NVIDIA announced Cosmos 3 and Isaac GR00T-related releases in 2026, including an open humanoid reference design using Unitree H2 Plus, Sharpa hands, Jetson Thor, and Isaac GR00T software/models. (nvidianews.nvidia.com)

  2. Robot-brain / data-flywheel companies — Figure, Skild AI, 1X, Physical Intelligence, and Apptronik are converging around the same thesis: the scarce asset is no longer just the robot body, but the data loop connecting real robots, teleoperation, simulation, human video, world models, and on-policy deployment data. Figure reports BotQ output rising to 1 Figure 03 per hour after 350+ units produced; 1X launched a dedicated World Model Lab; Skild raised $1.4B at a $14B+ valuation and acquired Zebra’s robotics arm; Apptronik closed a $520M extension bringing Series A to over $935M. (figure.ai)

  3. Defense / sovereign physical AI — defense autonomy is converting faster into funded deployments than consumer humanoids. Anduril announced a $363M CBP contract for 200+ extended-range autonomous sentry towers, a Dutch C-UAS contract, U.S. Navy/DIU XL-AUV selection, and a production agreement for at least 3,000 Barracuda-500M systems over three years; DARPA is also moving physical intelligence into materials and robotic servicing of satellites. (anduril.com)

Core thesis update: deployment evidence is now visible, but uneven. Logistics, surveillance, defense autonomy, and industrial pilots have stronger conversion signals than home humanoids. The portfolio stance should favor enablers and verified deployment rails over pure humanoid hype until reliability hours, customer economics, and production yields are disclosed.


SIGNAL_LAYER

NVIDIA is formalizing physical AI as a platform category

NVIDIA’s 2026 announcements push physical AI from “robot demo” toward a full stack: Cosmos world models, Isaac simulation, GR00T VLA models, Jetson robotics processors, Omniverse, and partner integrations across ABB, Agility, CMR Surgical, FANUC, Figure, Hexagon, KUKA, Medtronic, Skild AI, Universal Robots, World Labs, and YASKAWA. (nvidianews.nvidia.com)

NVIDIA also announced an open Isaac GR00T Reference Humanoid Robot for academic research, combining Unitree H2 Plus, Sharpa five-finger hands, Jetson Thor, and the Isaac GR00T platform. This matters because it lowers the barrier to standardized humanoid research and accelerates benchmarking around dexterity, onboard compute, and sim-to-real workflows. (nvidianews.nvidia.com)

Signal quality: High. This is infrastructure, not a one-off demo.

Figure shows a concrete production-rate signal and new commercial deployment channel

Figure says BotQ has produced over 350 Figure 03 humanoids and increased throughput from 1 robot/day to 1 robot/hour in under 120 days. Figure explicitly frames each robot as both deployable hardware and a data-collection engine for Helix. (figure.ai)

Figure also announced a commercial agreement with Catalyst Brands to deploy Figure humanoids into Catalyst’s Reno, Nevada distribution/logistics center. This is portfolio-relevant because it moves Figure from automotive pilot proof toward broader logistics deployment, though customer economics and fleet-scale productivity metrics remain undisclosed. (figure.ai)

Figure’s older but still relevant BMW disclosure reported 1,250+ runtime hours, 90,000+ parts loaded, and contribution to 30,000+ X3 vehicles during an 11-month Figure 02 deployment at BMW Spartanburg. That remains one of the better public humanoid deployment-hour datapoints. (figure.ai)

Signal quality: Medium-high. Production-rate and deployment claims are company-reported, but they are specific enough to track.

Agility Robotics continues to lead on boring-but-real logistics validation

Agility announced a Robots-as-a-Service commercial agreement with Toyota Motor Manufacturing Canada after a successful pilot, with Digit intended to support manufacturing, supply-chain, and logistics operations. (agilityrobotics.com)

Agility previously reported Digit moving over 100,000 totes in a GXO live commercial deployment, emphasizing throughput, safety, and integration into existing logistics workflows. (agilityrobotics.com)

Signal quality: High for deployment credibility. Agility is less flashy than Figure/Tesla but has stronger evidence of repetitive logistics execution.

Boston Dynamics / Hyundai moves Atlas from research into committed 2026 industrial deployments

Boston Dynamics says the production-ready electric Atlas was unveiled in January 2026 and that all 2026 Atlas deployments are already committed, with fleets scheduled for Hyundai’s Robotics Metaplant Application Center and Google DeepMind. (bostondynamics.com)

Hyundai Mobis says it will supply actuators for Atlas and that Hyundai Motor Group plans to build and deploy tens of thousands of robots over the next few years as part of a $26B U.S. investment. (bostondynamics.com)

Signal quality: Medium-high. Committed fleets and actuator supply-chain integration matter, but production volumes and customer ROI are not yet public.

1X is leaning into home humanoids, but the stronger signal is its factory + data-loop architecture

1X says its Hayward NEO Factory is online, vertically integrated, and has capacity for up to 10,000 NEOs/year, with planned automation increases toward 100,000+ annual capacity by end-2027. It also says first U.S. deliveries start in 2026 and early-access ownership is priced at $20,000, with a $499/month subscription option later. (1x.tech)

The stronger portfolio signal is 1X’s World Model Lab: 1X explicitly identifies embodied data — web-scale video, egocentric human video, simulation, teleoperated robot data, and on-policy NEO data — as the durable moat for humanoid autonomy. (1x.tech)

Signal quality: Medium. Factory/data-loop claims are concrete; home autonomy remains unproven.

Skild AI validates “robot brain as horizontal platform” and uses M&A to acquire deployment surface

Skild AI announced a $1.4B Series C in January 2026, led by SoftBank with NVentures, Macquarie, Bezos Expeditions, LG, Schneider Electric, Salesforce Ventures, IQT, and others participating, at a valuation above $14B. (skild.ai)

In April 2026, Skild acquired Zebra Technologies’ robotics division, formerly Fetch Robotics, to bring its omni-bodied robot brain into warehouse environments and accelerate the data flywheel. (skild.ai)

Signal quality: High for capital flow and strategic direction; medium for product economics.

Apptronik capital raise shows corporate strategic demand for humanoid deployment

Apptronik announced a $520M Series A-X extension in February 2026, bringing its Series A to over $935M and total capital raised to nearly $1B. Investors include B Capital, Google, Mercedes-Benz, PEAK6, AT&T Ventures, John Deere, and QIA; Apptronik also cites partnerships with Mercedes-Benz, GXO, Jabil, and Google DeepMind. (apptronik.com)

Signal quality: High for capital flow and strategic investor interest; medium for deployment evidence.

BCI remains medically early, but the human-machine boundary is moving into robotics control

Neuralink’s current clinical-trials page lists active trials for thought-based computer/robotic-arm control and speech decoding, plus an upcoming vision-restoration trial. (neuralink.com)

Neuralink’s January 2026 “Two Years of Telepathy” update says Telepathy aims to enable people with paralysis to control computers, phones, and robotic limbs using thoughts alone, and shows trial participants using implants for digital and physical device control. (neuralink.com)

Paradromics received FDA IDE approval for its Connect-One early feasibility study with the Connexus BCI, targeting high-bandwidth BCI applications that may eventually include advanced prosthetics and intelligent systems. (paradromics.com)

Signal quality: Medium. Clinically important, but long commercialization path.

AI materials / autonomous labs are becoming a sovereign-capex category

The DOE launched the Genesis Mission Consortium in February 2026 to use AI for scientific discovery, energy, emerging technologies, and national security. (energy.gov)

The DOE Genesis Mission page frames the effort as an integrated platform connecting supercomputers, experimental facilities, AI systems, and datasets, including autonomous labs and materials discovery. (energy.gov)

On June 4, 2026, DOE and Japan announced a planned $1B five-year partnership, with joint teams across DOE labs and Japanese institutions to advance quantum information, fusion, biotechnology, advanced materials, particle physics, and autonomous laboratory systems. (energy.gov)

Signal quality: High for sovereign intent; medium for investable private/public exposure.

Defense autonomy is converting from prototype to procurement faster than humanoids

Anduril announced a $363M CBP contract for more than 200 Extended Range Sentry Towers, expanding autonomous situational awareness on the Southwest Border. (anduril.com)

Anduril also announced a Dutch Ministry of Defence C-UAS contract using Lattice to link sensors and effectors into a coordinated network that autonomously helps operators defeat threats. (anduril.com)

Anduril and the U.S. Navy/DIU are moving XL-AUVs through the Combat Autonomous Maritime Platform Project after Anduril demonstrated long-duration XL-AUV performance; Anduril reports 42,355 km and 6,752 hours of autonomous undersea mission time. (anduril.com)

Anduril also announced a production agreement for at least 3,000 Barracuda-500M systems over three years, with first tranche deliveries expected in H1 2027 and optional integration with Lattice for autonomous/collaborative behaviors. (anduril.com)

Signal quality: High. These are procurement/production signals, not consumer demos.

CAPITAL_FLOW_MAP

Large private humanoid / robot-brain capital

  • Skild AI — $1.4B Series C, valuation over $14B; investors include SoftBank, NVentures, Macquarie, Bezos Expeditions, LG, Schneider Electric, Salesforce Ventures, IQT, and others. (skild.ai)
  • Apptronik — $520M Series A-X extension; Series A total over $935M; nearly $1B total raised; strategic investors include Google, Mercedes-Benz, John Deere, AT&T Ventures, QIA. (apptronik.com)
  • Figure AI — more than $1B committed Series C at $39B post-money valuation; investors include Parkway, Brookfield, NVIDIA, Macquarie, Intel Capital, LG Technology Ventures, Salesforce, T-Mobile Ventures, Qualcomm Ventures. (figure.ai)
  • 1X — no new financing found in this run, but factory capacity and World Model Lab expansion indicate internal capital deployment toward manufacturing and model/data infrastructure. (1x.tech)

Defense/autonomy capital and procurement

  • Anduril — $5B Series H announced in May 2026, plus multiple defense/public-sector contract signals in C-UAS, border surveillance, XL-AUV, and Barracuda production. (anduril.com)
  • Anduril / CBP — $363M for 200+ autonomous sentry towers. (anduril.com)
  • Anduril / Barracuda — minimum 3,000 SLB-500M systems over three years; first tranche expected H1 2027. (anduril.com)
  • U.S. Army JTAARS — production contract awarded to SURVICE Engineering for Joint Autonomous Aerial Resupply Systems. (army.mil)

Sovereign AI science / autonomous-lab capital

  • DOE Genesis Mission — AI-driven scientific discovery initiative tying together DOE labs, supercomputers, AI systems, experimental facilities, and unique datasets. (energy.gov)
  • U.S.–Japan Genesis partnership — planned combined $1B over five years to advance AI science, autonomous labs, advanced materials, fusion, biotech, quantum, and computing infrastructure, subject to future appropriations. (energy.gov)
  • DOE Office of Science awards — over $320M for 217 university/industry projects across physical sciences, including materials science and advanced materials. (energy.gov)

CONVERSION_LAYER

Converted / higher-confidence signals

  • Agility → Toyota RaaS agreement: successful pilot converted into commercial Robots-as-a-Service agreement. (agilityrobotics.com)
  • Agility → GXO tote milestone: 100,000+ totes moved in live deployment; good evidence of repetitive logistics execution. (agilityrobotics.com)
  • Figure → BMW deployment hours: 1,250+ runtime hours and 90,000+ parts loaded in BMW Spartanburg deployment. (figure.ai)
  • Figure → Catalyst Brands logistics agreement: new commercial channel beyond automotive pilots. (figure.ai)
  • Anduril → contracts/procurement: CBP, Dutch C-UAS, XL-AUV, Barracuda all indicate autonomous systems are already monetizing. (anduril.com)
  • Boston Dynamics → committed 2026 Atlas fleets: all 2026 deployments committed to Hyundai RMAC and Google DeepMind; actuator supply chain tied to Hyundai Mobis. (bostondynamics.com)

Still mostly promise / needs more proof

  • Home humanoids: 1X has factory/preorder signals, but home autonomy, safety, privacy, service costs, and customer retention remain unproven. (1x.tech)
  • Tesla Optimus: this run found investor-update references to Optimus construction, but not enough current official deployment-hour evidence to rank it above Agility/Figure/Boston on operational proof. Tesla remains a watch item rather than a confirmed deployment leader from this evidence set. (static.seekingalpha.com)
  • Robot-brain foundation models: Skild, Physical Intelligence, Figure Helix, and 1X world models are directionally important, but broad generalization remains unproven across uncontrolled customer environments. (skild.ai)
  • BCI-to-robotics control: active trials are real, but commercialization is still clinical/regulatory, not broad consumer robotics. (neuralink.com)

TIME_MAP

0–6 months

  • Track NVIDIA Cosmos 3 / GR00T adoption by robotics labs and industrial partners; watch whether reference designs become de facto standard research platforms. (nvidianews.nvidia.com)
  • Track Figure’s Catalyst deployment: fleet size, tasks, uptime, safety incidents, and measurable productivity. (figure.ai)
  • Track Agility Toyota deployment and GXO expansion: RaaS unit economics and facility count are the key conversion metrics. (agilityrobotics.com)
  • Track Anduril delivery timing and contract expansion in C-UAS, border surveillance, maritime autonomy, and Barracuda production. (anduril.com)

6–18 months

  • Figure BotQ: verify whether 1 robot/hour becomes sustained shipments and revenue-producing deployments, not internal fleet growth only. (figure.ai)
  • Boston Dynamics Atlas: monitor Hyundai/Google DeepMind deployment outcomes and whether 2027 customer availability opens. (bostondynamics.com)
  • 1X NEO: first home deliveries, teleoperation dependence, safety/privacy incidents, subscription economics, and actual household task autonomy. (1x.tech)
  • DOE/Japan Genesis: watch procurement/vendor awards for autonomous lab platforms, instruments, compute, AI infrastructure, and materials workflows. (energy.gov)

18–36 months

  • Robot-brain consolidation: acquisitions like Skild/Zebra may repeat as foundation-model labs buy deployed robot fleets and orchestration platforms. (skild.ai)
  • Defense autonomy scaling: Barracuda deliveries in H1 2027 and follow-on production options will indicate whether low-cost autonomous munitions become a durable procurement category. (anduril.com)
  • BCI/prosthetics: Neuralink, Paradromics, and others may produce meaningful assistive-control milestones, but portfolio impact likely remains medtech-specialized before becoming broad human-machine-interface exposure. (neuralink.com)

BOTTLENECK_INDEX

Dexterity and contact-rich manipulation

Walking and navigation are increasingly less scarce than useful hands, tactile feedback, grasp reliability, force control, and long-horizon manipulation. NVIDIA’s GR00T reference humanoid explicitly combines Unitree H2 Plus with Sharpa five-finger hands, reinforcing that dexterity hardware is now a central bottleneck. (nvidianews.nvidia.com)

Data flywheel quality

Figure, Skild, and 1X all explicitly frame deployment data as core to model improvement. Figure calls each robot a data-collection engine; Skild describes deployment as feeding its data flywheel; 1X says embodied data is the durable moat. (figure.ai)

Production yield and supply chain

Boston Dynamics/Hyundai Mobis actuator collaboration shows actuator supply is strategic; 1X’s vertical integration claims show motors, batteries, structures, transmissions, soft goods, and sensors are supply-chain control points. (bostondynamics.com)

Safety and certification

Agility emphasizes live deployment safety as a prerequisite for scaling in logistics; safety remains a gating factor for 3PL adoption, factory integration, and home humanoid deployment. (agilityrobotics.com)

Power, batteries, and uptime

Agility’s 2025 Digit update cited battery improvements and autonomous docking as deployment features; uptime still matters more than demo skill because customer ROI depends on continuous repetitive operation. (agilityrobotics.com)

Sim-to-real and world-model validation

NVIDIA Isaac/Cosmos and 1X World Model Lab indicate the industry is attacking sim-to-real through physics simulation, synthetic data, and world models, but real-world reliability remains the benchmark. (nvidianews.nvidia.com)

Regulatory / clinical path for BCI

Neuralink and Paradromics are real clinical signals, but BCI is gated by FDA/clinical trials, safety, surgical adoption, and reimbursement pathways. (neuralink.com)


THESIS_STATE

Thesis strengthened

  • Physical AI is becoming a platform market: NVIDIA’s stack, Skild’s robot brain, Figure/1X data flywheels, and DOE autonomous labs all support the model-to-matter thesis. (nvidianews.nvidia.com)
  • Defense autonomy has the clearest revenue path: procurement and production agreements are visible now. (anduril.com)
  • Humanoid deployment is moving from demo to early customer sites: Agility/Toyota, Agility/GXO, Figure/BMW, Figure/Catalyst, and Boston/Hyundai/DeepMind show multiple conversion points. (agilityrobotics.com)

Thesis unchanged

  • Dexterity remains the frontier: hands, tactile sensing, contact-rich manipulation, and long-horizon autonomy remain harder than locomotion. NVIDIA’s reference humanoid explicitly includes five-finger dexterous hands, reinforcing the bottleneck. (nvidianews.nvidia.com)
  • Home humanoids remain high-upside but under-proven: 1X’s NEO has factory/preorder signals, but consumer household robotics still lacks broad verified task autonomy and service economics. (1x.tech)

Thesis weakened / caution

  • Pure humanoid OEM exposure is not yet clean public-market exposure: the best companies are private, valuations are stretched, and public read-through is indirect.
  • Demos are increasingly less useful as signals: the market should prioritize runtime hours, paid deployments, production yields, and customer ROI over videos.

Strategic interpretation

The strongest physical-AI thesis is no longer “humanoids will be everywhere.” It is: AI is building a control stack for matter, and the first durable money will flow to compute, simulation, sensors, actuators, fleet orchestration, defense autonomy, industrial logistics, and high-value scientific/medical interfaces.

Humanoids matter because they are the most general embodiment and the best data-collection narrative. But portfolio allocation should not confuse narrative centrality with near-term monetization. In this run, defense autonomy and industrial/logistics robots show stronger conversion than consumer-home humanoids.

The market is moving toward a stack:

  1. World/simulation layer — Cosmos, Isaac, Omniverse, 1X world models, DOE AI science platforms.
  2. Robot-brain layer — Skild, Figure Helix, Physical Intelligence π0, Apptronik/Google DeepMind, 1X.
  3. Embodiment layer — Figure 03, Digit, Atlas, NEO, Apollo, Unitree, Boston Dynamics.
  4. Deployment layer — BMW, Toyota, GXO, Catalyst, Hyundai, CBP, Dutch MoD, U.S. Navy/DIU.
  5. Feedback layer — runtime hours, teleoperation, synthetic data, on-policy data, safety data, customer productivity.

The investable edge is to own the layers that win regardless of which robot body wins.


Sources

  1. NVIDIA — “NVIDIA and Global Robotics Leaders Take Physical AI to the Real World” — https://nvidianews.nvidia.com/news/nvid…REDACTED (nvidianews.nvidia.com)
  2. NVIDIA — “NVIDIA Announces NVIDIA Isaac GR00T Reference Humanoid Robot for Academic Research” — https://nvidianews.nvidia.com/news/nvid…REDACTED (nvidianews.nvidia.com)
  3. NVIDIA — “NVIDIA Expands Open Model Families to Power the Next Wave of Agentic, Physical and Healthcare AI” — https://nvidianews.nvidia.com/news/nvid…REDACTED (nvidianews.nvidia.com)
  4. Figure — “Ramping Figure 03 Production” — https://www.figure.ai/news/ramp…REDACTED (figure.ai)
  5. Figure — “Figure Signs Agreement with Catalyst Brands to Scale Humanoid Operations” — https://www.figure.ai/news/figu…REDACTED (figure.ai)
  6. Figure — “F.02 Contributed to the Production of 30,000 Cars at BMW” — https://www.figure.ai/news/production-at-bmw (figure.ai)
  7. Figure — “Figure Exceeds $1B in Series C Funding at $39B Post-Money Valuation” — https://www.figure.ai/news/series-c (figure.ai)
  8. Agility Robotics — “Agility Robotics Announces Commercial Agreement with Toyota Motor Manufacturing Canada” — https://www.agilityrobotics.com/content/agil…REDACTED (agilityrobotics.com)
  9. Agility Robotics — “Digit Moves Over 100,000 Totes in Commercial Deployment” — https://www.agilityrobotics.com/content/digi…REDACTED (agilityrobotics.com)
  10. Boston Dynamics — “Boston Dynamics Unveils New Atlas Robot to Revolutionize Industry” — https://bostondynamics.com/blog/bost…REDACTED/ (bostondynamics.com)
  11. Boston Dynamics — “Atlas’ Evolution From Research Robot to Industrial Humanoid” — https://bostondynamics.com/blog/atla…REDACTED/ (bostondynamics.com)
  12. Boston Dynamics / Hyundai Mobis — “Hyundai Mobis Forms Strategic Collaboration Framework with Boston Dynamics” — https://bostondynamics.com/news/hyun…REDACTED/ (bostondynamics.com)
  13. 1X — “NEO Factory | Building Your NEO” — https://www.1x.tech/discover/neo-factory (1x.tech)
  14. 1X — “The 1X World Model Lab” — https://www.1x.tech/discover/1x-world-model-lab (1x.tech)
  15. 1X — “NEO Home Robot | Order Today” — https://www.1x.tech/ja_jp/discover/neo-home-robot (1x.tech)
  16. 1X — “Order NEO” — https://www.1x.tech/order (1x.tech)
  17. Skild AI — “Announcing Series C” — https://www.skild.ai/blogs/series-c (skild.ai)
  18. Skild AI — “Skild AI Acquires Zebra Technologies’ Robotics Arm” — https://www.skild.ai/blogs/skild-zebra (skild.ai)
  19. Apptronik — “Apptronik Closes Over $935 Million Series A” — https://apptronik.com/news-collection/appt…REDACTED (apptronik.com)
  20. Physical Intelligence — “Open Sourcing π0” — https://www.physicalintelligence.company/blog/openpi (physicalintelligence.company)
  21. Neuralink — “Clinical Trials / Patient Registry” — https://neuralink.com/patient-registry/ (neuralink.com)
  22. Neuralink — “Device Control” — https://neuralink.com/trials/device-control/ (neuralink.com)
  23. Neuralink — “Two Years of Telepathy” — https://neuralink.com/updates/two-years-of-telepathy/ (neuralink.com)
  24. Paradromics — “FDA Approval for Connect-One Clinical Study” — https://paradromics.com/news/para…REDACTED/ (paradromics.com)
  25. Anduril — “CBP Expand Partnership With 200 Additional Extended Range Sentry Towers” — https://www.anduril.com/news/andu…REDACTED (anduril.com)
  26. Anduril — “Dutch Ministry of Defence C-UAS Contract” — https://www.anduril.com/news/andu…REDACTED (anduril.com)
  27. Anduril — “DIU and U.S. Navy select Anduril for XL-AUV program” — https://www.anduril.com/news/diu-…REDACTED (anduril.com)
  28. Anduril — “Surface-Launched Barracuda-500M Production Agreement” — https://www.anduril.com/news/andu…REDACTED (anduril.com)
  29. Anduril — “Anduril Announces $5B Series H Raise” — https://www.anduril.com/news/andu…REDACTED (anduril.com)
  30. U.S. Army — “PM UAS awards production contract for Joint Autonomous Aerial Resupply Systems” — https://www.army.mil/article-amp/292147/pm_u…REDACTED (army.mil)
  31. DARPA — “Robotic Servicing of Geosynchronous Satellites technology to launch in 2026” — https://www.darpa.mil/news/2026/robo…REDACTED (darpa.mil)
  32. DARPA — “Rethinking robotics with physical intelligence” — https://www.darpa.mil/news/2026/rethinking-robotics (darpa.mil)
  33. DOE — “Energy Department Launches Genesis Mission Consortium” — https://www.energy.gov/articles/ener…REDACTED (energy.gov)
  34. DOE — “The Genesis Mission” — https://www.energy.gov/genesis (energy.gov)
  35. Genesis Mission — https://genesis.energy.gov/ (genesis.energy.gov)
  36. DOE — “U.S. and Japan Announce $1B Partnership Under Genesis Mission” — https://www.energy.gov/articles/unit…REDACTED (energy.gov)
  37. DOE — “U.S. Department of Energy Invests $320 Million in Pioneering Scientific Research” — https://www.energy.gov/science/articles/us-d…REDACTED (energy.gov)
  38. Tesla Q1 2026 Update mirror — https://static.seekingalpha.com/uploads/sa_presentations/967/125967/original.pdf (static.seekingalpha.com)
  39. Axios — “Elon Musk: Tesla will sell humanoid robots by end of 2027” — https://www.axios.com/2026/01/22/elon…REDACTED (axios.com)
Source in the house: eth-memory/sol/reports/2026-06-17-physical-ai-model-to-matter.md☉ Sol