Here's a bet: Most manufacturing plants won't write a six-figure check for a robot they've never seen work. Eden Robotics thinks they might, however, rent one.
The London-founded startup—still finding its footing after graduating from Y Combinator earlier this year—is pushing what CEO Stamatis Floratos calls "physical agents for hire." Instead of asking warehouse managers or factory supervisors to gamble on capital equipment that might gather dust in a corner, Eden proposes something closer to the gig economy: robots deployed on usage-based contracts, billed by the task or the hour, maintained and updated by Eden itself.
It's an appealing pitch in theory. In practice, the company hasn't yet shown its cards.
Floratos and his co-founder, Dr. Joey Humphreys, launched Eden on November 23, 2025. Humphreys—who earned his PhD from the University of Leeds last year studying bio-inspired locomotion and reinforcement learning for legged robots—handles the technical architecture. Floratos, who has referenced earlier bootstrapped ventures in his public posts (though these are not independently verified), runs the business side. The team, listed at four people on the Y Combinator profile (though this information may be stale), appears to have relocated from London to San Francisco during the accelerator program, based on recruiting posts from earlier this year.
What they're building, or say they're building, spans hardware and software. The hardware line includes a semi-humanoid robot dubbed "Eden I," a robotic arm called "Maradona I," and something called "The Cluster"—an autonomous work cell. Specifications? Payload capacity, degrees of freedom, cycle times, ingress protection ratings—Eden Robotics has not publicly disclosed specifications for its hardware. No safety certifications that an operations manager could review before signing off on a deployment.
The software stack, which some find conceptually interesting, includes what Eden calls Theta OS—a robotics operating system now at version 0.2.1—that blends onboard autonomy with human oversight. Two coordination layers sit on top: Interlink, described as a low-latency teleoperation system originally designed (per company materials) for satellite and lunar applications, and Fleet-Sync, which coordinates multiple robots across a facility.
The idea is "shared autonomy." Robots handle the repetitive stuff—picking, sorting, kitting, machine tending. When they hit something novel, a human operator steps in remotely. It's not a new concept; others in the space have explored similar models. But Eden is wagering that pairing this approach with a rental model—rather than asking customers to own the complexity—might lower the adoption barrier.
A Market Accelerating, Perhaps Faster Than Expected
Eden's timing coincides with what looks like an inflection point in humanoid robotics, though skeptics might counter it's been an inflection point for three years running now.
1X Technologies opened a factory in Hayward, California on April 30 to manufacture its NEO robot at scale, with consumer shipments reportedly planned for this year. Agility Robotics—now just "Agility" after a March rebrand—has inked commercial deals with Toyota Motor Manufacturing Canada. Figure AI ran a pilot at BMW's Spartanburg plant last year where its robots handled north of 90,000 parts, and BMW has since moved humanoids into German factories.
Meanwhile, Boston Dynamics announced in January 2026 that its new Atlas platform will target factory deployments by 2028. Hyundai's plans have been mentioned in industry analysis, though specific production targets have not been confirmed in detail. Siemens partnered with a UK-based outfit called Humanoid in April to deploy the HMND-01 Alpha at a logistics facility in Erlangen, Germany.
China's humanoid output, according to TrendForce, is projected to jump 94% year-over-year in 2026, with Unitree and AgiBot expected to dominate the domestic market. Goldman Sachs' 2026 report revised its base-case forecast for the global humanoid market upward to $38 billion by 2035—a significant jump from an earlier $6 billion estimate, though the range of scenarios remains wide and the barriers substantial.
Not everyone is convinced the sector is ready to scale. Gartner predicted in January that fewer than 20 companies will reach production-stage manufacturing and supply chain deployments by 2028, with the vast majority stuck in pilot purgatory. McKinsey flagged component bottlenecks as a constraint, even as it noted rising investment in "physical AI" and partnerships between startups and industrial giants.
The question for Eden: Can a four-person team with no disclosed pilots compete in a field where established players have named customers, published metrics, and years-long relationships with automotive and logistics OEMs?
What Factories Get (and What They Don't Know)
Eden's target customers are manufacturing floors and warehouses—environments where traditional automation has required massive upfront investment and lengthy integration. The robots, as envisioned, would slot into human teams for tasks like assembly, picking, inspection, and material handling. Not the high-precision, high-speed work of a traditional industrial arm, but the adaptable, variable tasks that have resisted full automation.
The usage-based model is the hook. Instead of a capital expenditure that might take years to justify, Eden proposes a managed service: robots that show up, do the work, get maintained and updated remotely, and disappear if the economics don't pan out. The risk, in theory, shifts from customer to vendor.
But here's where things get murky. Pricing structures? Not public. Service-level agreements? Undisclosed. Integration timelines, on-site support, minimum contract terms—similarly opaque. Is billing based on hours of operation, tasks completed, or some hybrid metric? Eden hasn't said.
This isn't unusual for a startup just months out of an accelerator. But for procurement teams evaluating automation investments, it leaves them comparing a largely theoretical offering against competitors with spec sheets, case studies, and customer references.
No pilot customers have been announced. No deployment metrics. The company's public presence consists of posts on its own website, a Y Combinator profile, and scattered updates from the founders. No independent coverage. No customer testimonials. Not even photos or videos of the hardware in action.
The Technical Question
Humphreys' academic background—optimal control, reinforcement learning, teleoperation—suggests the team has the technical chops. But a PhD thesis on legged locomotion and a robotics operating system are different beasts than a robot that can run reliably in a dusty warehouse for 16-hour shifts.
The "Polymath" model—Eden's term for its general-purpose manipulation and locomotion system—is described but not demonstrated. How it handles edge cases, how quickly it learns new tasks, how often human operators need to intervene: all unknown.
Theta OS, at version 0.2.1, is still in early iterations. Fleet-Sync and Interlink sound promising on paper, but coordinating multiple robots in a live production environment is notoriously difficult. Latency, edge cases, safety interlocks, integration with existing warehouse management systems—these are the details that make or break deployments, and Eden hasn't published data on any of them.
A Crowded Room
The humanoid robotics space is getting noisy. Agility and 1X have moved to commercial production. Figure AI has performance data from automotive pilots. Boston Dynamics carries decades of robotics credibility, even if Atlas is still a few years from factory floors. Startups like Apptronik and Sanctuary AI have secured partnerships and funding.
Eden's differentiator—the RaaS model—could be compelling if the economics work. But "if" is doing a lot of work in that sentence. Without disclosed pricing, without pilots, without technical specifications, it's hard to evaluate whether Eden is positioned to deliver on the promise or whether it's another pitch deck in a market that's started to demand proof.
The company lists headquarters in London on its website but San Francisco on the Y Combinator profile, suggesting either a dual presence or a transition that hasn't been fully clarified. The team size—four, per Y Combinator—may have changed since the profile was last updated. These small inconsistencies aren't dealbreakers, but they add to the sense of a company still figuring out its public identity.
What Comes Next
For Eden, the path forward is straightforward, if not easy: land a pilot. Get a robot onto a factory floor or into a warehouse. Publish metrics—uptime, task completion rates, cost per operation. Show that the RaaS model works not just as a sales pitch but as an operational reality.
Until then, the company remains a bet on a business model rather than a demonstrated product. Manufacturing and logistics operators evaluating humanoid automation have plenty of options with more public track records. Eden's proposition—robots by the hour, risk shifted away from the customer—is attractive. But attractive propositions need proof, and proof, for now, is what Eden hasn't yet provided.
The next inflection, as with so many startups in this space, will be the first named customer. Whether that announcement comes in weeks or quarters—or at all—will determine whether Eden's rental model was prescient or just a pitch that sounded good in a demo day presentation.
