Frontier lab for physical AI / worldview AI

TAM TOP-DOWN

The broadest Physical AI market estimate (Grand View Research) is $81.6B for 2025 but explicitly bundles mature industrial robotics. The mature industrial robotics base alone is $33.9B (Grand View Research), leaving roughly $47.7B attributable to newer AI-driven systems. The embodied/worldview AI subset is more honestly sized at $4.44B in 2025 (MarketsandMarkets) growing at 39% CAGR, implying roughly $23B by 2028, which is the three-year forward anchor used here. A frontier lab sells into this forward window, making $23B the working top-down TAM.

$4.44B embodied AI (2025) x (1.39^3) = $23.7B, rounded to $23B (2028 forward TAM)

TAM BOTTOM-UP

ABI Research counts 7.5M factories globally; realistically only the top decile by revenue and automation maturity are buyers of frontier AI research or advanced physical AI systems. That yields 750,000 candidate facilities. Of those, perhaps 10% have both budget and technical readiness in the near term, giving 75,000 active prospects. At a blended contract or system price of $90,000 (Unitree G1 mid-range anchor as proxy for AI system contract value), bottom-up TAM is $6.75B. This is 3.

7.5M factories x 10% top decile x 10% near-term ready x $90,000 = $6.75B bottom-up floor

SAM FILTERS

Geography filter: China, US, and Japan represent roughly 56% of global manufacturing output (18% + 9.5% + 4.6% proxied from $867B Japan vs $4.66T China base). Segment filter: frontier lab customers are large OEMs and research-licensing partners, not the long tail of SME factories; this narrows to an estimated 15% of the geographic slice. Buyer readiness filter: ISO 25785-1 still in Working Draft means many would-be buyers are paused on large commitments; a 35% readiness discount applied. Combined: $23B x 56% x 15% x 65% = $1.

$23B x 56% geography x 15% segment x 65% readiness = $1.26B top-down; $338M bottom-up; midpoint = $690M

SOM BUILD

Year 1: a new frontier lab realistically signs 3 to 5 anchor research or licensing contracts; at $1.5M average (below enterprise hardware cost, reflecting software and model licensing), year-1 revenue is roughly $6M, under 1% of SAM. Year 2: reputation and first deployments enable 8 to 10 contracts at $1.5M average, reaching $14M. Year 3: 23 contracts at $1.5M average, reaching $34M, representing approximately 5% of SAM. Competitor pressure from five funded incumbents (Figure AI at $39.5B valuation, Skild AI at $1.

23 contracts x $1.5M average contract value = $34M year-3 SOM (4.9% of $690M SAM)