REVEALED 2025.02REASSESSED 2026.08.04

Foundation

Foundation Phantom

Name: Foundation Phantom
Brand: Foundation

Proprietary rolling contact gearboxes and a camera-only perception stack differentiate the Phantom at the hardware level. Teleoperation remains central to the operating model, placing it firmly at Level II. The production gap between a reported 2025 target of approximately 40 units and a 10,000-unit 2026 target is the defining open question.

PriceNot listedNot tracked yet

Autonomy Ladder

Why Level II.

One five-rung scale, every category, from acting only on command to working unsupervised anywhere.

VGeneralized AutonomyAny task, any setting. No consumer robot has reached it.
IVEnvironmental AutonomyHandles new obstacles and re-plans in real time; rare in consumer products.
IIIConditional AutonomyFinishes whole jobs in familiar spaces; struggles at the edges.
To reach Level IIIMost sessions finish without intervention. Rescue is occasional, not routine.
IIAssisted AutonomyThis robot

Foundation Phantom runs preset routines on its own, and the owner is its recovery system.

Navigation and control
Camera-only (8 cameras, no LiDAR); hybrid LLM reasoning plus physics-based action models; reinforcement learning locomotion
Why this level
- Manufacturer documentation and press coverage confirm the Phantom uses an LLM-plus-physics hybrid AI stack for locomotion and basic task execution, with teleoperation (VR headset or supervised remote control) retained as the primary intervention model for tasks the robot cannot complete autonomously.
- Per the Autonomy Ladder framework, teleoperation dependency caps classification at Level II regardless of onboard autonomy demonstrations.
- End-to-end task completion without teleoperator involvement in real industrial deployments has not been independently documented.
8 sources reviewed
IManual AutomationOne action on command; no independent operation.

How we assign a level →

Behind the classification

Sources we reviewed.

Pre-Release8 sources reviewed

Classified from pre-release material; figures will firm up once units ship and are independently tested.

Sources reviewed

Manufacturer documentation2 sourcesOngoing

Foundation company statements, product announcements, and CEO interviews (Fox Business, Humanoids Daily)

Press & video reviews2 sourcesOngoing

Press coverage of Pentagon contracts, Ukraine deployment, and company credibility questions

Longitudinal signal2 sourcesReviewed

Robovations dated log for this robot: 4 tracked entries spanning ~22 months.

Specification sources1 sourceReviewed

Pages the published specifications were read from.

foundation.botfoundation.bot

Monitoring record5 sourcesReviewed

Sources behind the tracker entries following this robot.

How we evaluate evidence →

Human Readiness

Promising Progress.

Available only via research contracts and lease to industrial/defense partners; no consumer purchase path exists. Initial automotive deployments logged in 2025, with two units sent to Ukraine for military logistics testing in February 2026. That test was reported on 31 May 2026 as unsuccessful: testers found the MK-1 unsuitable for battlefield use in its current form, citing a payload of about 20 kg, no waterproofing, and battery life insufficient for large-scale deployment. Supervised teleoperation remains integral to the operating model, capping autonomous classification.

Where this sits on the readiness scale

Ready NowProven autonomous performance, reliable, and available for purchase.
Promising ProgressThis robotReal advancement, but the ownership experience is still maturing.
OverhypedMarketing is outpacing actual capability.
WaitTechnology exists but better options are coming soon.
Falls ShortFails core autonomy or reliability standards.

How we assess readiness →

Pre-Release Assessment

How this projection was reached

41of 100

Concept Stage

SpeculativeConceptPlausiblePromisingReady

The hatched tail is unreachable by construction. A projection does not publish as a full mark, because the evidence that settles it does not exist yet: whether the thing ships and works in ordinary use. Those 10 points are what shipping has to earn.

The reading

Novel actuator approach with real industrial pilots, undercut by a 2024 GM fabrication incident and a 250x production scale gap on thin capitalization. Provisional classification pending independent verification of autonomous task performance and manufacturing viability.

Still open

Whether rolling contact gearbox efficiency claims (90-95%) hold under sustained industrial loads
Battery runtime sufficient for a full factory shift without tethering
Production viability at 10,000 units in 2026 on approximately $21 million in total funding
Waterproofing and short-circuit mitigation for outdoor or wet-environment tasks
Independent verification of autonomous task completion without teleoperator intervention
MK2 actuator redesign timeline and whether payload doubles as claimed

Five pre-shipping signalsweighted, then capped41What can be judged before anyone owns one.

Track Record25% of the assessment28

What this asks

Has this maker shipped what it demoed before, on schedule, with the promised capability?

Engineering25% of the assessment61

What this asks

Are the demonstrated capabilities achievable in shipping form, or do they need breakthroughs not in evidence?

Demo Match20% of the assessment43

What this asks

What has been shown to work, against what is claimed in marketing.

Readiness15% of the assessment30

What this asks

Ship-date specificity, pricing transparency, distribution and regulatory status.

Disclosure15% of the assessment38

What this asks

How much of the product is disclosed. Higher means fewer fundamental unknowns remain.

Not a Robovations Scorethe two never convert into each other

This is a projection from pre-shipping signals, not a measurement. When the robot ships, the assessment is archived, a first Robovations Score is computed from the field record, and the change is logged on the tracker rather than appearing as a new number.

Read the pre-release method →

FIT CHECK

Will the Foundation Phantom fit your scenario?

Describe your scenario using the controls. Each choice reads the spec sheet for you, flagging where the Foundation Phantom is built for your space and where it isn’t.

Where it operates

Primary tasks

Supervision needs

Timeline

A classification of fit for the conditions you set, not a recommendation. Verdicts describe engineering design intent, not test results.

SPECIFICATION

The full spec sheet.

Manufacturer-published figures, measured against every humanoid robots we track.

Dimensional drawing

Key specificationsManufacturer-published

Weight: 175 lbs✓

Where it lands · ranked across 54 humanoid robotsThis robotMedianTypical range

Runtime120min

60 min600 min

Near median

Weight175lbs

26.46 lbs197.4+ lbs

Above median

Position shows where the figure sits across the tracked category, not a quality judgment.

RECORD

How the assessment has moved.

Every announcement, release and reassessment behind the current classification.

Coming soon

Just launched

Active updates

Settled

Discontinued

Jun 2024 · NewsCNBC reports Foundation fundraising materials contained exaggerated GM partnership claims
Apr 2025 · AnnouncementFoundation publicly reveals Phantom humanoid targeting industrial and defense markets
Feb 2026 · NewsTwo Phantom MK1 units deployed to Ukraine for frontline logistics and reconnaissance testing
Apr 2026 · NewsFoundation secures $24 million Pentagon contract portfolio with SBIR Phase III designation
Jul 2026 · AssessmentPre-Release Assessment recorded at 39 of 100
Aug 2026 · AssessmentPre-Release Assessment revised 39 → 41

Common questions

What people actually ask.

7 answered

OperationWhat tasks has the Phantom MK1 actually performed in documented deployments?

Automotive parts racking and unracking in a US factory (reported as 2025 pilot deployment) and logistics runs plus reconnaissance sweeps in Ukraine (February 2026). Both contexts involved supervised operation, not fully autonomous task completion. The Ukraine testing was reported in May 2026 as showing the platform unsuitable for battlefield use without significant upgrades.

OperationDoes the Phantom MK1 use LiDAR for navigation?

No. Foundation uses eight cameras with no LiDAR, citing a philosophy that additional sensors add noise. The perception system pairs camera feeds with the hybrid LLM-plus-physics AI stack for environmental interpretation.

OperationWhat happened with Foundation's General Motors partnership claims?

In June 2024, CNBC reported that Foundation fundraising materials claimed GM as a committed customer with a targeted $300 million purchase order. GM denied any such commitment. Foundation co-founder LeBlanc confirmed the denial publicly.

OperationWhy is the Phantom MK1 classified at Level II if it has an onboard AI stack?

The Autonomy Ladder requires end-to-end task completion without teleoperator involvement. Manufacturer documentation and press coverage confirm teleoperation via VR headset or supervised remote control remains part of the operating model, which caps classification at Level II regardless of onboard capabilities.

OperationIs the Phantom MK1 available to purchase?

No retail or direct-purchase channel exists. The robot is available via research contracts and a lease model (approximately $100,000 per year) to qualified industrial and defense partners only.

OperationHow much can the Phantom MK1 carry?

About 20 kg, roughly 44 lb. Independent coverage and Ukraine’s February 2026 field testing agree on that figure, and the testers named it as a limitation for resupply work.

OperationWhat is the Phantom MK2 and what changes from the MK1?

Foundation describes the Phantom 2 as doubling the MK-1’s payload, which would take it from about 20 kg (44 lb) to about 40 kg (88 lb). The reported change list runs wider than the actuators: consolidated electronics to reduce short-circuit risk, added waterproofing, larger battery packs, and cast-moulded bodywork to cut manufacturing cost. Waterproofing and battery capacity are two of the shortfalls Ukrainian testers named in the MK-1. The MK2 is dated to April 2026.

Cite / Embed

Reference the Foundation Phantom classification.

Embed the Autonomy Ladder™ mark or copy the citation. The mark is served from this record rather than saved as a file, so it always shows the level we currently hold. If we reassess, it updates on your page by itself.

<a href="https://robovations.com/robots/foundation-phantom/?ref=badge" target="_blank" rel="noopener"><img src="https://robovations.com/badge/r/foundation-phantom/seal-light" width="268" height="124" loading="lazy" alt="Autonomy Ladder classification for Foundation Phantom, assessed by Robovations." /></a>

The mark must link to this record. An unlinked mark is not authorised.
No recolouring, cropping, rotating, restyling, or re-typesetting. Scale proportionally only, up to 2×: the seal to 536px wide, the standard to 224px.
Leave clear space equal to half the numeral tile on all sides: 32px on the seal, 22px on the standard.
The mark states a classification, not an endorsement. It may not appear beside language implying Robovations recommends, rates, or approves the product.
Hot-link the image from robovations.com. Do not download, self-host, proxy or re-upload it: a saved copy stops updating, and a mark that has stopped updating is a claim we no longer stand behind.
The mark is live. It reflects the classification on this record as it stands. If we reassess, it changes on your page automatically. If we withdraw the classification, it stops rendering. Nothing is required from you either way.

Citation

Robovations. "Foundation Phantom: Autonomy Level II (Assisted Autonomy)." Autonomy Ladder™, assessed August 2026. https://robovations.com/robots/foundation-phantom/

Where to next

Three ways to keep going.

For evaluators

Compare with a peer.

See the Foundation Phantom and MagicLab MagicBot Z1 side by side: autonomy level, Robovations Score, readiness, and price. Classification, not a verdict.

See the full comparison

Or compare with Galaxea R1, PNDbotics Adam, Noetix Bumi.

For browsers

See every Level II.

Every robot we’ve classified at assisted autonomy. Sorted by readiness, score, and price. Filter by category, manufacturer, or release date.

Browse the database

Updated weekly with new entries and reclassifications.

Foundation Phantom

Will the Foundation Phantom fit your scenario?

The full spec sheet.

How the assessment has moved.

What people actually ask.

Other humanoid robots we have classified.

MagicLab MagicBot Z1

Galaxea R1

PNDbotics Adam

Noetix Bumi

Reference the Foundation Phantom classification.

Three ways to keep going.

Get the next reassessment.

Compare with a peer.

See every Level II.