Pomegra
Berkeley robot-data startup XDOF is in late-stage talks for a $1.2 billion Series B led by 8VC, backed by annualized revenue approaching $50 million just months after emerging from stealth.
- 8VC is leading a Series B that would value XDOF at $1.2 billion, fewer than three months after the company exited stealth.
- Annualized revenue is approaching $50 million, a pace that prompted investors to re-engage before XDOF sought new capital.
- The company's $70 million Series A closed in June 2026 with Thrive Capital, Andreessen Horowitz, Lux, and Spark Capital.
The Deal
XDOF, a Berkeley-born startup that supplies training data for general-purpose robots, is in late-stage discussions to raise a Series B at a valuation of approximately $1.2 billion, led by 8VC. The talks surfaced in early September 2026, fewer than three months after XDOF exited stealth and roughly the same interval after a $70 million Series A backed by Thrive Capital, Andreessen Horowitz, Lux, and Spark Capital. Annualized revenue is approaching $50 million - a trajectory that put the company back in front of venture firms before it was looking for new capital.
What Does XDOF Actually Build?
XDOF does not train robots. It trains the humans and systems that generate robot training data. The company builds data pipelines, teleoperation hardware, and annotation systems that frontier AI labs and humanoid robotics companies can't easily construct in-house - an outsourced data supply chain for physical AI.
The core product traces back to GELLO, a low-cost teleoperation controller developed by XDOF's founders at UC Berkeley. Operators use GELLO to steer robotic arms remotely, generating demonstrations that robots later imitate. A parallel track deploys egocentric operators - workers who wear body sensors to record mundane tasks like folding laundry or flattening cardboard boxes. Both streams feed a growing library of embodied AI training data.
The company was founded in October 2024 by Philipp Wu (CEO), Yide (Fred) Shentu (CTO), and Nemo Jin, all researchers from UC Berkeley's robotics program. It currently employs roughly 60 people and serves about 20 customers.
Why Is the Valuation Moving This Fast?
At roughly 24x annualized revenue, a $1.2 billion valuation demands justification. That justification is competitive scarcity. Physical AI labs are racing to source embodied data at scale, and suppliers capable of delivering high-quality, task-relevant demonstrations number in the single digits globally.
The revenue ramp - from near-zero at stealth exit to a $50 million annualized run rate in under three months - suggests demand was already queued ahead of the company's public debut, not generated by it. XDOF's 20 customers include several frontier AI labs, which limits concentration risk while anchoring the company inside the highest-value part of the market.
The Series A, which closed in June at $70 million, implicitly valued the business at a fraction of today's figures. The change is mainly evidentiary: actual revenue, a manageable customer base, and a category that has shifted from speculative to operationally necessary for every serious robotics program.
What Does 8VC's Lead Position Signal?
8VC has existing exposure to the robotics and defense-tech stack, making XDOF a logical adjacency. The firm's decision to lead signals a thesis that the training-data bottleneck for physical AI is at least as durable as the equivalent constraint was for large language models, and that XDOF's head start compounds into a defensible position. Whether Series A investors exercise pro-rata rights to participate has not been confirmed, though the earlier syndicate composition suggests competition for allocation.
The Physical AI Context
The demand underlying XDOF's growth is not company-specific. Humanoid robot developers and embodied AI labs face the same data problem that language model labs faced in 2020: more demonstrations, more diversity, faster annotation pipelines than internal teams can supply. Outsourcing to specialized vendors is the expected path.
XDOF benefits from high switching costs. Customers who integrate its pipelines into training workflows don't migrate trivially. That structural stickiness, more than any individual product feature, is what justifies the valuation multiple at this stage.
Outlook
XDOF is closing on unicorn status at a pace that will force a re-examination of what "late-stage" means when the underlying market moves this quickly. The Series B is not yet closed, and terms can shift. If it finalizes near $1.2 billion, the deal sets a pricing benchmark for robot-training-data infrastructure that every subsequent entrant will be measured against.
