2026-06-17 👁️ — views

Physical AI Investment Tracker — Who Is Funding the Robot Ramp (Mid-2026)

Capital flows, funding rounds, and implied valuations for the leading physical AI companies — Waymo, Figure, Physical Intelligence, and more — through mid-2026.

Why investment data is the real conviction signal

Revenue is sparse. Customer deployments are in early pilot. Public benchmarks remain thin. For physical AI companies — autonomous vehicles, humanoid robots, robot foundation models — the clearest signal of where the industry believes value will accrete is where the largest, most sophisticated investors are placing capital. This tracker compiles publicly reported funding rounds, disclosed valuations, and known lead investors for the major physical AI companies through mid-2026.

One definitional note: “Physical AI” here means systems that perceive and act in the physical world using AI — autonomous vehicles, humanoid and mobile robots, and the foundation models built specifically to control them. It excludes pure software AI and cloud inference businesses.

Section 1 — Master funding table

The table below draws from publicly reported rounds. Valuations marked (est.) are analyst estimates or reported secondary-market signals, not disclosed figures. Revenue multiples are not meaningful for most of these companies because commercial revenue is near zero — valuations represent pure option value on the technology and its eventual market.

Reading the table: Waymo and Tesla occupy a different order of magnitude from the humanoid startups. The startup cohort (Figure, π, Apptronik, Skild, 1X) ranges from roughly $100M to $675M per round, with disclosed post-money valuations in the $2B–$3B range where reported. Agility and Boston Dynamics have exited the fundraising cycle through strategic acquisition — their capital needs are now absorbed by their corporate parents.

Section 2 — Implied valuation analysis

Waymo: the benchmark pure-play AV valuation

Waymo is the only major physical AI company that is (a) commercially operational at scale and (b) partly independent from its parent in the sense that it has accepted outside capital with disclosed terms. The $45B–$50B+ range circulating in analyst reports as of 2025–2026 (est.) implies a specific per-unit-economics logic:

• Waymo operates approximately 150,000+ weekly paid rides as of mid-2026 (publicly reported by Alphabet)
• At $45B valuation and 150,000 rides/week: implied valuation per weekly ride is roughly $300,000
• At $50B and 150,000 rides/week: roughly $333,000 per weekly ride
• For context, Uber’s entire global business — millions of rides daily — trades at roughly $150B

The implied multiple only makes sense if Waymo’s weekly ride count scales by 10x–50x over the next five to seven years. That is the embedded growth assumption in the current valuation range (est.). Investors who own Waymo exposure through Alphabet stock are paying for that optionality.

Tesla: embedded option, not standalone valuation

Tesla’s market capitalization peaked above $1 trillion in 2025. Several equity research models attempt to decompose it:

• Bear case: FSD/Robotaxi/Optimus assigned near-zero value; vehicle business alone worth $200B–$400B at current margins
• Bull case (ARK Invest model, 2024–2025): Robotaxi fleet NPV assigned $500B–$700B+ in a scenario where Tesla deploys millions of Cybercabs by 2029
• Consensus range: Most sell-side analysts model $0–$300B for the autonomy/robotics segment as of mid-2026, with wide disagreement

The structural difference between Tesla and the startups: Tesla does not need to raise capital for its autonomy programs. Vehicle margin cash flow self-funds FSD compute, Dojo, Optimus production trials, and Cybercab development. This is a significant structural advantage — the startups must convince investors to continue funding pre-revenue programs, while Tesla can sustain them through an existing cash-generating business.

Physical AI startup multiples: pure option value

Most of the humanoid and robot foundation model startups have commercial revenue that is effectively zero or sub-$10M annually as of mid-2026. Their valuations — $2.4B for Physical Intelligence, $2.6B for Figure — are entirely forward-looking. The implied valuation framework is not P/E or revenue multiple; it is expected-value probability weighting:

• If humanoid robots reach meaningful deployment at scale (millions of units) in 5–8 years, the market is worth hundreds of billions
• A $2B–$3B valuation at a $400M–$675M round implies investors assign a 20%–40% probability to the startup being a major winner in that outcome

That is the honest math behind the numbers. These are not companies being valued on fundamentals — they are options on a future market that may or may not materialize at scale within the investment horizon.

Section 3 — Capital concentration: who is betting most aggressively

A handful of investors appear repeatedly across the physical AI funding landscape. The pattern reveals distinct strategic theses.

Jeff Bezos / Bezos Expeditions — largest individual physical AI position

Bezos appears as a lead or significant investor in at least four physical AI companies:

• Figure AI — $675M Series B participant
• 1X Technologies — Series B participant (alongside OpenAI)
• Physical Intelligence — $400M Series A lead
• Skild AI — $300M Series A lead

This is not passive diversification. Bezos is co-leading rounds, not just writing checks. The thesis appears to be that embodied AI — robots that act in the physical world — is the next major platform, and that multiple layers (hardware, foundation model, general-purpose brain) will each produce large independent outcomes. The portfolio spans two layers: humanoid hardware (Figure, 1X) and foundation models for robot control (π, Skild).

OpenAI — robot plus AI model convergence

OpenAI appears in three physical AI rounds:

• Figure AI — Series B investor (alongside Microsoft, NVIDIA, Bezos)
• 1X Technologies — Series B lead
• Physical Intelligence — Series A investor

OpenAI’s thesis is visible: the same large language and vision model capabilities that power chat and coding can be extended to physical robot control. Physical Intelligence’s π0 model is built explicitly on this premise — a foundation model trained on diverse robot demonstration data, designed to control multiple robot hardware platforms. OpenAI wants either to own the robotics AI layer itself or to be closely coupled with the company that does.

NVIDIA — silicon and software ecosystem play

NVIDIA appears in Figure AI’s Series B. This is less a financial bet and more a platform stake: Figure uses NVIDIA Jetson and Isaac Sim for robot compute and simulation. NVIDIA’s investment signal is that Figure’s deployment scale will drive material demand for Jetson hardware and the Isaac robotics platform. At the company’s GPU margins, even moderate robot deployment volumes represent meaningful incremental revenue. It is an ecosystem investment, not a financial return play.

Amazon — vertical integration through acquisition

Amazon did not invest in Agility Robotics — it acquired it (2023, approximately $150M reported). Amazon’s strategy is different from the investor cohort: it wants to own the robot deployment directly to reduce fulfillment labor costs, not to profit from a robot company’s eventual IPO. Agility’s Digit is being deployed inside Amazon warehouses. The return model is operational efficiency, not exit multiple.

Section 4 — What the money is actually buying

The intended use of funds tells you more about a company’s current constraints than any other metric.

Waymo: manufacturing and city-by-city operations

Waymo’s capital needs are unusual among this cohort because Waymo already has a product that generates revenue. Its constraints are manufacturing volume and operational geography. The $5.6B 2020 external round, continued Alphabet capital allocation, and a subsequent external round in 2023 funded:

• Development and production of the sixth-generation Waymo One vehicle (purpose-built, no steering wheel, sensor suite integrated into body panels)
• Operations infrastructure in San Francisco, Phoenix, Los Angeles, and Austin
• Remote monitoring center staffing and vehicle maintenance
• City expansion groundwork for Miami and other markets

Waymo is not in a “build the product” phase — it is in a “scale a working product” phase. That changes the capital efficiency math entirely.

Humanoid startups: hardware iteration and foundation model training

For Figure, 1X, and Apptronik, the money buys two things in parallel:

1. Hardware iteration cycles — each generation of a humanoid robot requires tooling, materials testing, actuator development, and manufacturing process design. Figure moved from Figure 01 to Figure 02 during the post-Series-B period. Each generation improves dexterity, battery life, and cost-per-unit. These are expensive, time-consuming engineering cycles.

2. Foundation model training — training a robot to perform diverse manipulation tasks requires massive amounts of demonstration data (teleoperated or human-guided robot motion) plus GPU compute for model training. Physical Intelligence’s entire business is this layer — selling a trained robot brain, not hardware. Figure trains its own models for Figure hardware. Both require sustained compute spend.

Tesla: self-funded structural advantage

Tesla allocates capital to Optimus from its automotive cash flows without external fundraising. This means:

• No dilution from venture rounds
• No investor pressure on timeline
• No requirement to hit milestones to unlock the next tranche

The trade-off is that Tesla’s Optimus program competes internally with FSD, Dojo, and Cybercab for engineering and capital resources. The self-funding model is structurally advantageous but not frictionless.

Section 5 — H2 2026 fundraising watch list

Based on publicly reported timelines, company stage, and capital consumption rates, the following companies are candidates for new fundraising activity in the second half of 2026.

What could accelerate the timeline: A credible factory deployment — robots performing repeatable tasks in an industrial environment at 500+ units — would dramatically compress the fundraising cycle by converting “option value” into “demonstrated capability.” No humanoid startup has publicly achieved this milestone at that scale as of mid-2026. The first company to do so will likely trigger a repricing across the entire cohort.

What could compress valuations: A high-profile safety incident involving an autonomous vehicle or humanoid robot in a commercial deployment — similar to Cruise’s 2023 suspension — could cool investor sentiment across the category. The current valuations price in smooth regulatory and technical progress. A setback reprices that assumption.

Benchmark context: this is the fourth article in the physical AI series

This tracker is the fourth in a series covering physical AI from multiple angles:

1. Operational ramp metrics — production counts, deployment scale, miles driven
2. Humanoid robot technology — hardware generations, dexterity benchmarks, foundation model capabilities
3. AV safety and regulation — California DMV data, NHTSA crash reporting, state permit maps
4. Investment and valuation — this article

The investment picture is inseparable from the operational one. The companies attracting the largest rounds are those that have either demonstrated operational traction (Waymo) or assembled the most credible technical teams and investor networks (Figure, Physical Intelligence). Capital follows conviction — and conviction, in physical AI, still rests primarily on technical credibility rather than commercial proof points.

Sources

• Figure AI $675M Series B — TechCrunch ↗
• Physical Intelligence $400M Series A — Bloomberg ↗
• Waymo standalone valuation signals — Alphabet earnings ↗
• Skild AI $300M — Reuters ↗
• Apptronik $350M — Google/Samsung ↗

Tags