ExplainersBrain providers & foundation models
How did DEPLOY correct the Physical Intelligence valuation conflation?
Widely cited coverage puts Physical Intelligence's valuation at $11 billion. It's actually $5.6 billion.
The $10B and $38B figures reporters keep quoting belong to Project Prometheus, a different Jeff Bezos-backed robotics lab that people confuse with Physical Intelligence because both have Bezos money and both work near robotics AI. Two companies, two valuations, one persistent mix-up.
Know the moment this changes - verified, the second it's real.
Related
The framing error: aggregator coverage cites PI at $11B
Aggregator coverage of Physical Intelligence frequently cites the company's valuation at $11B. The framing appears across multiple aggregator pieces and surfaces in secondary press references to PI's funding round + valuation tier. Per Agent A primary-source verification, the framing is wrong. Physical Intelligence is valued at $5.6B confirmed.
The $10B and $38B figures often cited in aggregator coverage belong to Project Prometheus, a separate Jeff Bezos-backed lab commonly conflated with PI due to adjacent positioning + overlapping investor coverage.
Verified state: $5.6B Physical Intelligence; $10B/$38B Project Prometheus
Per primary-source verification, two distinct entities with structurally distinct valuations. Physical Intelligence: $5.6B confirmed valuation; foundation-model-for-robotics brain provider; pi0 + pi05_base Apache-2.0 open + π0.6 + π0.7 closed model lineage; lab + limited-customer-pilots deployment maturity. Project Prometheus: separate Jeff Bezos-backed lab; robotics-adjacent AI research focus; $10B and $38B valuation figures attach to Project Prometheus per Agent A primary-source verification.
The two entities operate adjacent positioning in the robotics-AI investment space but are structurally distinct corporate entities. The aggregator framing conflating them produces inflated valuation claims that attach to the wrong company.
Why entity-distinction discipline matters: adjacent-entity conflation pressure
The PI + Project Prometheus conflation operates as a worked example of entity-distinction discipline at the funding-round verification layer. Adjacent positioning + similar investor coverage produces conflation pressure. Trade-press coverage of high-valuation AI investments frequently cites figures without per-entity verification depth. When two adjacent entities operate similar funding patterns + similar positioning + similar investor backing, the aggregator coverage tends to collapse them into a single "the company raised at $X valuation" claim.
Per-entity primary-source verification catches the conflation structurally: audit-first discipline against named-entity primary sources resolves which valuation attaches to which entity. The conflation gets caught at the entity-distinction layer.
Cap-flag transparency surfaces the conflation explicitly
Per DEPLOY's how-deploy-verifies methodology editorial, the corrected attribution doesn't just surface the verified state; it surfaces the conflation explicitly so readers can recognize the aggregator-drift pattern and apply the entity-distinction discipline themselves. "$5.6B verified (not $11B; $10B/$38B belong to Project Prometheus separate Bezos lab per Agent A correction)" is the canonical surface framing in the Physical Intelligence entity anchor.
The discipline doesn't just produce the right number. It produces the right number with the wrong-number rejection surfaced so the conflation gets caught structurally across downstream coverage.
Why this catch matters: framework operates at editorial-anchor depth
Institutional partners audit DEPLOY's framework discipline at the operational-practice layer, not just the stated-methodology layer.
This piece documents the catch at narrative-canonical depth: how the catch happened (audit-first verification against named-entity primary sources + adjacent-entity-conflation audit); what the discipline was (entity-distinction discipline at the funding-round verification layer); what the editorial outcome was (corrected $5.6B PI valuation + explicit conflation rejection); what the broader pattern is (adjacent positioning + similar investor coverage produces conflation pressure; per-entity primary-source verification catches it structurally).
The catch demonstrates the discipline operationally at editorial-anchor depth. Two distinct entities + two distinct funding profiles + one aggregator conflation; primary-source verification resolves the entity-distinction + the wrong-attribution rejection surfaces transparently.
The framing error
Widely cited coverage puts Physical Intelligence's valuation at $11B. That's wrong. Physical Intelligence is worth $5.6B as of its most recent verified funding round. The $10B and $38B figures that keep surfacing in trade-press coverage belong to a different company: Project Prometheus, a separate Jeff Bezos-backed lab.
Reporters conflate the two because both companies have Bezos money and both work near robotics AI. But they're distinct entities with distinct funding rounds and distinct valuations.
How the mix-up happens
Both companies sit in the same corner of the market:
- Both are backed by Jeff Bezos.
- Both work on AI systems adjacent to robotics.
- Both raised at venture-tier valuations that trade press wants to write about.
- Neither has the kind of household-name recognition that would make the confusion obvious to a reader.
When a story mentions "the Bezos-backed robotics-AI lab worth $10B," a busy reader (or a busy aggregator) assumes it's Physical Intelligence, because Physical Intelligence is the one they've heard of. The number attaches to the wrong name, and the mistake propagates across coverage.
The verified state
Physical Intelligence (entity page):
- Valuation: $5.6B (post-Series B, verified against primary sources)
- What they make: foundation models for robotics (Pi0 and pi05_base under Apache-2.0; π0.6 and π0.7 closed)
- Where they are commercially: lab work plus limited customer pilots
- Positioning: a third-party brain provider you can integrate into humanoid or adjacent platforms
Project Prometheus (separate company):
- Valuation: $10B and $38B are the figures that show up in coverage
- Backer: Jeff Bezos
- Focus: robotics-adjacent AI research (details still emerging in public reporting)
The two are structurally different companies with different funding rounds, different investors, and different product roadmaps. Any story that treats "Physical Intelligence at $11B" as a fact is compressing them.
Why this catch matters
The catch matters for three reasons.
The number is load-bearing. People use company valuations to compare who's real, who's ahead, and who's likely to survive. A $5.6B company and a $38B company aren't in the same conversation. Getting the number right for the right company is the whole point of a registry.
The pattern recurs. Physical Intelligence and Project Prometheus aren't the only pair like this. Any time two companies have similar investors, similar positioning, and adjacent focus areas, coverage tends to collapse them. The same discipline (check which entity the round is actually attached to) catches the same class of mistake elsewhere.
Being explicit about the mistake helps readers. The Physical Intelligence page carries the correction with the wrong numbers named alongside the right one, so a reader who arrives with the wrong figure in mind sees it flagged, not just replaced.
Related catches on the same pattern
- How DEPLOY corrected the Figure 03 BMW narrative
- Monogram: Doug Unis is CMO/founder, not CEO
- 1X Redwood is a captive brain, not humanoid hardware
For more on the underlying method, see How DEPLOY verifies. For the broader tier of foundation-model-for-robotics providers, see Brain providers cluster and Captive vs third-party brain providers.
Frequently asked
What is Physical Intelligence's actual valuation?
$5.6B confirmed per Agent A primary-source verification. Physical Intelligence operates as foundation-model-for-robotics brain provider at $5.6B post-Series-B valuation. Per the Physical Intelligence entity anchor, the $5.6B figure is the load-bearing primary-source-anchored funding tier. Aggregator coverage citing $11B PI valuation operates outside primary-source-anchored verification of the PI specific funding-round attribution; the $10B and $38B figures often cited belong to Project Prometheus (separate Jeff Bezos-backed lab; adjacent positioning + overlapping investor coverage produces the aggregator conflation).
What is Project Prometheus?
A separate Jeff Bezos-backed lab distinct from Physical Intelligence per Agent A primary-source verification. The two entities operate adjacent positioning in the robotics-adjacent AI research space, both with Jeff Bezos involvement at the investor or backer level. The $10B and $38B valuation figures attach to Project Prometheus per Agent A audit; specific Project Prometheus founding date + funding rounds + lead investors operate at separate primary-source verification depth pending dedicated entity-anchor coverage.
The two entities are structurally distinct corporate entities with structurally distinct funding rounds + investor structures + product focus.
Why does aggregator coverage conflate PI with Project Prometheus?
Adjacent positioning + overlapping investor coverage produces conflation pressure. Both entities operate in the robotics-adjacent AI research space; both have Jeff Bezos involvement at the investor or backer level; both operate at substantial venture-tier valuation depth.
Trade-press coverage of high-valuation AI investments frequently cites figures without per-entity verification depth; when two adjacent entities operate similar funding patterns + similar positioning + similar investor backing, the aggregator coverage tends to collapse them into a single "the company raised at $X valuation" claim. Per DEPLOY's framework discipline, audit-first verification against named-entity primary sources resolves the entity-distinction structurally.
How does DEPLOY catch funding-round conflations?
Per the audit-first verification pattern, funding-round + valuation claims operate at primary-source-verified depth across four axes. Which entity is the funding round attached to? Named-entity recipient verification against company IR + lead-investor announcements + SEC primary-source filings. What is the verified valuation at the round close? Post-money valuation verification against company IR + lead-investor confirmation + SEC primary-source disclosure.
What is the lead investor + round structure? Lead-investor identification + round structure verification against primary-source-anchored sources. What adjacent entities might be conflated? Adjacent entities operating with similar funding patterns + similar investor coverage + similar positioning that might surface in aggregator conflations.
Does cap-flag transparency just surface the right number?
No. The cap-flag discipline produces the right number with the wrong-number rejection surfaced so the conflation gets caught structurally across downstream coverage. Per how-deploy-verifies methodology editorial, the corrected attribution doesn't just produce $5.6B as the verified PI valuation. It produces $5.6B with explicit rejection of the $11B aggregator framing + attribution of the $10B/$38B figures to Project Prometheus.
Readers can recognize the aggregator-drift pattern + apply the entity-distinction discipline themselves. "$5.6B verified (not $11B; $10B/$38B belong to Project Prometheus separate Bezos lab per Agent A correction)" is the canonical surface framing.
Why document this correction as a worked example?
Institutional partners audit DEPLOY's framework discipline at the operational-practice layer, not just the stated-methodology layer. The verification-posture statement at /verified-vs-claimed describes the framework abstractly. The corrections journal at /corrections lists corrections that shipped publicly.
This piece operates at narrative-canonical depth: how the catch happened (audit-first verification against named-entity primary sources + adjacent-entity-conflation audit); what the discipline was (entity-distinction discipline at funding-round verification layer); what the editorial outcome was (corrected $5.6B PI valuation + explicit conflation rejection); what the broader pattern is (adjacent positioning + similar investor coverage produces conflation pressure). The catch demonstrates the discipline operationally at editorial-anchor depth.
Sources: Source: Agent A primary-source verification + Physical Intelligence company communications + adjacent-entity-conflation audit. Entity-distinction discipline at funding-round verification layer.
Explore the deployment map
View the global mapSee where these robots are verified operating, by place and type.
The Physical Intelligence valuation correction-as-worked-example documents the entity-distinction discipline operating at the funding-round verification layer. Aggregator framing: Physical Intelligence valued at $11B (often $10B-$38B framings). Per Agent A primary-source verification: Physical Intelligence at $5.6B confirmed; the $10B and $38B figures belong to Project Prometheus (separate Jeff Bezos-backed lab). The two entities operate adjacent positioning in the robotics-adjacent AI research space but are structurally distinct corporate entities with distinct funding rounds + valuations + investor structures + product focus. Adjacent positioning + overlapping investor coverage produces conflation pressure across aggregator coverage; per-entity primary-source verification catches the conflation structurally; cap-flag transparency surfaces the conflation explicitly so the wrong-attribution rejection propagates through downstream coverage. The Physical Intelligence entity anchor surfaces the corrected $5.6B valuation + explicit conflation rejection against $11B/$10B/$38B framings. Two distinct entities + two distinct funding profiles + one aggregator conflation; primary-source verification resolves the entity-distinction + the wrong-attribution rejection surfaces transparently. How DEPLOY verifies →
Continue reading
What is Physical Intelligence?
Canonical Physical Intelligence entity anchor; $5.6B confirmed valuation; foundation-model-for-robotics tier within brain-providers cluster.
Read article →
How DEPLOY corrected the Figure 03 BMW narrative
Parallel framework-in-action correction-narrative; three-layer aggregator-drift rejection at deployment-attribution layer.
Read article →
1X Redwood is a captive brain, not humanoid hardware
Parallel framework-in-action correction-narrative; brain-entity vs hardware-model entity-type discipline at editorial-anchor depth.
Read article →
Captive vs third-party brain providers
Brain-provider integration model gradient; Physical Intelligence anchors third-party-foundation-model positioning at foundation-model-for-robotics tier.
Read article →
Compare alternatives
Physical Intelligence entity anchorBrain-providers clusterHow DEPLOY verifies (methodology canonical)Captive vs third-party brain providers
More questions people ask
- How did DEPLOY catch the 1X Redwood brain-vs-hardware framing error?
- How did DEPLOY correct the Covariant corporate-state narrative?
- How does DEPLOY apply verified-vs-claimed at within-entity granularity?
- How does DEPLOY classify autonomy boundaries across physical AI?
- How does DEPLOY verify physical AI claims?
- How does DEPLOY's 9-tier source-quality rubric classify verification evidence?
Every robot coming to market, tracked and verified.
DEPLOY is the registry of physical AI: every company, model, deployment, and safety record, with the sources behind each one. Browse the humanoids, robotaxis, drones, and delivery robots being deployed right now.
Explore the DEPLOY registry →Most-tracked robots
In the registry
Keep reading
Ask the registry