Sanctuary AI

Sanctuary AI is a Canadian humanoid-robotics and AI company headquartered in Vancouver, British Columbia, founded in 2018 by Geordie Rose (co-founder of D-Wave Systems and Kindred), Suzanne Gildert (also a D-Wave and Kindred alumna), Olivia Norton, and Ajay Agrawal (University of Toronto economist and Creative Destruction Lab founder). Sanctuary AI develops the Phoenix general-purpose humanoid robot and the proprietary Carbon cognitive architecture that powers it, with an explicit positioning around "human-equivalent intelligence in humanoid form" that distinguishes the company from the broader humanoid-robotics cohort. As of 2026, Sanctuary AI is the principal Canadian humanoid-robotics insurgent, with a strategic manufacturing-and-investment partnership with Magna International and federal-government innovation funding through the Canadian Strategic Innovation Fund.

At a glance

• Founded: 2018 in Vancouver, British Columbia by Geordie Rose, Suzanne Gildert, Olivia Norton, and Ajay Agrawal.
• Status: Private.
• Funding: Approximately $130 million USD cumulative through early 2026, including $76 million USD Series A in November 2022 plus follow-on rounds and a Canadian Strategic Innovation Fund grant of approximately CAD $30 million.
• CEO: Geordie Rose, co-founder.
• Other notable leadership: Suzanne Gildert (Chief Product Officer, co-founder), Olivia Norton (Chief Technology Officer, co-founder).
• Open weights: None. Sanctuary's Carbon cognitive architecture and Phoenix hardware are proprietary; no open-source model releases.
• Flagship products: Phoenix general-purpose humanoid robot (Generation 7 as of 2024, with subsequent iterations through 2025 and 2026). The Carbon cognitive architecture (the AI software stack), branded as "human-equivalent intelligence" by the company.

Origins

Sanctuary AI was founded in 2018 by a team with an unusually long shared history. Geordie Rose co-founded D-Wave Systems, the quantum-computing pioneer headquartered in Burnaby, British Columbia, in 1999. Rose subsequently co-founded Kindred AI in 2014 (a robotics company that built warehouse-task automation, acquired by Ocado in 2020 for approximately $260 million). Suzanne Gildert, an experimental physicist with a doctorate from the University of Birmingham, joined D-Wave in 2010 as a quantum-computing scientist and subsequently moved with Rose to Kindred and then to Sanctuary. Olivia Norton, an engineer with extensive industrial-automation experience, joined as a co-founder from the same Vancouver-robotics community. Ajay Agrawal, a University of Toronto Rotman School of Management professor and the founder of the Creative Destruction Lab, joined as a co-founder bringing the economic-and-strategy perspective on the long-run market for general-purpose humanoid intelligence.

The founding thesis is more ambitious than most humanoid-robotics startups of the same vintage. Sanctuary's explicit framing has been that the goal is not warehouse-and-manufacturing automation but human-equivalent general intelligence delivered in a humanoid form factor. The company has consistently used "AGI" (artificial general intelligence) and "human-equivalent intelligence" framings in its public communications, positioning itself less as a hardware company building robots and more as an AGI research company that happens to have built a humanoid embodiment for that intelligence. The framing is structurally distinct from the Figure AI / Apptronik / Agility Robotics commercial-task-automation framing and from the Tesla AI labour-substitution framing.

The early hardware focus was on completing a fully-actuated humanoid platform that could perform a broad range of tasks under tele-operation, with the cognitive architecture growing over time to take over autonomous control across the task surface. The Phoenix line of humanoids has gone through six published generations from 2020 through 2024, with Generation 7 unveiled in April 2024 representing the company's most-current commercial-ready platform. The Carbon cognitive architecture is the company's AI substrate, branded as a distinct approach from the vision-language-action (VLA) models that competitors deploy.

The Magna International partnership announced in April 2023 is the company's strategic-and-investment relationship with the global automotive-and-manufacturing supplier. Magna invested in the Series A round, joined the company's board, and entered into a manufacturing-and-deployment agreement for Phoenix in automotive-related applications. The partnership is structurally significant for two reasons: it provides manufacturing expertise that Sanctuary does not have in-house, and it provides a customer-and-deployment pipeline through Magna's existing automotive-OEM relationships.

Mission and strategy

Sanctuary AI's stated mission is to "create the world's first human-like intelligence in general-purpose robots." The strategic premise is that the most valuable long-run outcome of humanoid robotics is not labour-substitution in specific industrial verticals but the development of a general-purpose intelligence embodied in a form factor that integrates with the human-designed physical world. The framing makes Sanctuary the most-AGI-oriented company in the humanoid-robotics cohort.

The Carbon cognitive architecture is the strategic differentiation against the broader humanoid-robotics field. Where Figure AI's Helix and Physical Intelligence's π₀ are vision-language-action foundation models (text-and-image input, motor-action output trained at large scale), Carbon is described publicly as a more-cognitive-architecture approach with explicit reasoning, memory, and planning components rather than a single end-to-end neural network. The technical merits of the two approaches are contested in the academic-robotics-research community: end-to-end approaches have been winning on standard benchmarks since 2023, but cognitive-architecture approaches have advantages on interpretability, on long-horizon task decomposition, and on plausibly approaching the kind of generalisation that AGI would require.

The capital-intensity-versus-time trade-off positions Sanctuary differently than the well-funded US humanoid cohort. With approximately $130 million USD raised (versus Figure AI's $1.5 billion), Sanctuary is operating with substantially lower runway. The strategic implication is that Sanctuary cannot fund a multi-year fully-vertically-integrated build-out the way Figure AI can, and the Magna partnership is part of the structural response: Sanctuary delivers the AI and the platform design, Magna handles manufacturing and customer-deployment relationships through the existing automotive supply chain.

Models and products

• Phoenix (general-purpose bipedal humanoid robot). Generation 7 (April 2024) is the company's most-current public platform. Five-foot-seven-inch bipedal humanoid weighing approximately 155 pounds, with hydraulic-and-electric hybrid actuation in earlier generations and fully-electric actuation in Generation 7. Distinctive five-finger end-effectors designed for high-precision human-task manipulation rather than the gripper-style end-effectors of warehouse-focused competitors.
• Carbon cognitive architecture. The proprietary AI substrate that powers Phoenix. Described publicly as a cognitive-architecture approach with explicit memory, planning, perception, and motor-control modules rather than a single end-to-end neural network. The architectural choice is the company's most-visible technical-differentiation claim.
• Tele-operation platform. Phoenix supports human tele-operation as a fallback control mode and as a primary data-collection mechanism for training the autonomous control stack. The tele-operation surface has been used in customer pilot deployments where the autonomous-task coverage is still being expanded.

Benchmarks and standing

Sanctuary has not published standard humanoid-robotics benchmark numbers; the field lacks widely-adopted standard benchmarks at the level that the language-model field has. Standing is measured through customer pilot deployments, demonstrated autonomous-task coverage, and capital raised. By those measures Sanctuary is in the second tier of the global humanoid cohort: meaningful research-grade engineering, a credible customer-deployment relationship through Magna, but smaller scale and less commercial-deployment evidence than the US leading group.

The company has published demonstration videos showing Phoenix performing a range of tasks across automotive-manufacturing pilots, but has not disclosed deployment-fleet sizes, per-unit cost, or per-hour-of-operation cost. The framing matches the broader humanoid-robotics category's current disclosure norms, which are substantially less transparent than the language-model category's benchmark publishing.

Leadership

• Geordie Rose (Chief Executive Officer, co-founder). Theoretical-physicist background with a doctorate from the University of British Columbia. Co-founded D-Wave Systems in 1999 (the world's first commercial quantum-computing company), co-founded Kindred in 2014 (acquired by Ocado in 2020). One of the most-tenured Vancouver-robotics-and-deep-tech operators.
• Suzanne Gildert (Chief Product Officer, co-founder). Experimental physicist with a doctorate from the University of Birmingham. Joined D-Wave in 2010 as a quantum-computing scientist, moved with Geordie Rose to Kindred and then to Sanctuary. Leads the Carbon cognitive-architecture work.
• Olivia Norton (Chief Technology Officer, co-founder). Industrial-automation engineer with prior experience in the Vancouver robotics-and-software community. Leads the hardware-and-engineering organisation.
• Ajay Agrawal (Co-founder, board member). Rotman School of Management professor at the University of Toronto and founder of the Creative Destruction Lab. Co-author of "Prediction Machines" and other influential AI-economics books. Provides the economics-and-strategy perspective on the long-run humanoid-AGI market.

The senior engineering team includes named hires from Vancouver-area robotics companies (Kindred, Inkblot, others), with the company maintaining a smaller and more concentrated senior-engineering organisation than the US-cohort competitors.

Funding and backers

Sanctuary AI has raised approximately $130 million USD in cumulative private capital plus federal-government innovation funding through early 2026.

• Series A (November 2022): $76 million USD led by Bell Canada and including Magna International, Workday Ventures, BDC Capital, InBC, and a Canadian Strategic Innovation Fund grant. The round's syndicate structure reflects a deliberate Canadian-strategic-capital orientation rather than the standard Silicon Valley venture-capital syndicate that most US humanoid-robotics rounds have featured.
• Canadian Strategic Innovation Fund: Approximately CAD $30 million in non-dilutive federal-government grant funding, the largest single Strategic Innovation Fund grant to a Canadian AI company.
• Subsequent rounds (2023 to 2025): Smaller follow-on rounds reported in the $20 million to $40 million range, with limited public disclosure of round structure and lead investors.

The capital-side trajectory differs structurally from the US cohort. Sanctuary's lower total capital raised, the substantial non-dilutive Canadian-government-grant component, and the Magna strategic-investor-and-customer relationship together produce a different operating-model trajectory than the venture-capital-heavy US cohort. The trade-off is slower scale but higher capital-efficiency-per-dollar.

Industry position

Sanctuary's industry position is as the principal Canadian humanoid-robotics company and as the most-AGI-oriented player in the global humanoid cohort. The AGI-framing differentiation matters because it positions Sanctuary differently in the investor-and-customer narrative than the labour-substitution and task-automation framings the US cohort uses. The framing has been a recruitment advantage (the company has attracted researchers interested in cognitive-architecture-as-AGI approaches who might not have joined a warehouse-task-automation company) and a fundraising challenge (the AGI-framing-with-current-capabilities gap has been more visible at Sanctuary than at competitors whose framings are tighter to demonstrated short-term task coverage).

The Magna manufacturing-and-deployment partnership provides a structural advantage that several competitors lack: a path to volume manufacturing without building a dedicated humanoid factory (like Agility Robotics's Salem RoboFab) and a customer pipeline through Magna's existing automotive-OEM relationships. The trade-off is dependency on Magna's strategic priorities, which can shift independently of Sanctuary's roadmap.

Competitive landscape

• Figure AI (US): Vertically-integrated humanoid + Helix VLA model. Substantially larger funding (~$1.5 billion).
• Tesla AI Optimus (US): Internal Tesla factory deployment; highest potential scale through Tesla manufacturing.
• 1X (Norway / US): Consumer-and-household humanoid (NEO, EVE).
• Apptronik (US): Industrial-manufacturing humanoid (Apollo) with Mercedes-Benz partnership.
• Agility Robotics (US): Warehouse-and-logistics humanoid (Digit) with Amazon deployment.
• Boston Dynamics (US, owned by Hyundai): Atlas + Spot + Stretch.
• Chinese humanoid cluster (Unitree, Agibot, UBTECH, XPeng Robotics, Xiaomi Robotics, Fourier, EngineAI). Cost-and-volume competition.
• Physical Intelligence and Skild AI (US): Robotics-AI foundation-model labs, not hardware companies; potential AI-stack partners or competitors depending on industry structure evolution.

Outlook

Open questions and watchable signals over the next 6 to 18 months:

• Magna deployment scale. The partnership has been the most-visible commercial relationship since 2023. Whether it scales to multiple automotive-supplier deployment sites, multiple task domains, or named-customer-OEM relationships through Magna's distribution channel will be the most informative signal on Sanctuary's commercial trajectory.
• Carbon cognitive-architecture demonstrations. The technical-differentiation claim for Carbon has been on architectural rather than empirical grounds. A first-public-benchmark or first-published-comparison against vision-language-action competitors would meaningfully advance the technical-credibility narrative.
• Capital trajectory. Sanctuary's total capital raised is substantially smaller than the US-cohort competitors'. A Series B or Series C round in 2026 or 2027 will reveal whether the institutional-capital market is pricing the Magna-strategic-relationship and the AGI-framing-differentiation as sufficient to fund the next development stage, or whether the company will need to seek a different operating model (a larger Magna-strategic-investment, a public listing on the Canadian or US exchanges, or an acquisition).
• Geordie Rose's strategic-direction calls. The CEO's prior trajectory (D-Wave Systems, Kindred) has been distinguished by long-horizon technical bets that produced both substantial commercial outcomes (Kindred's Ocado acquisition) and substantial technical-credibility events (D-Wave's quantum-computing position). The Sanctuary trajectory will largely follow the strategic choices Rose makes through 2026 and 2027, particularly around the Carbon-versus-VLA technical positioning and the standalone-versus-Magna-deeper-integration commercial posture.

Sources

• Sanctuary AI Phoenix product page. The official platform description.
• Sanctuary AI announces Magna International partnership. The April 2023 partnership announcement.
• Sanctuary AI Series A raises USD $76 million. Sanctuary's announcement of the November 2022 round.
• Canadian Strategic Innovation Fund award to Sanctuary AI. Federal-government program announcing the SIF grant.
• Geordie Rose Wikipedia entry. Background on the co-founder's D-Wave and Kindred history.
• Companion profiles: Figure AI, Tesla AI, 1X, Apptronik, Agility Robotics, Boston Dynamics, Physical Intelligence, Skild AI for the broader humanoid-and-robotics-AI cohort context.