Zip Robotic

Category: Humanoids

Humanoid robot news: Tesla Optimus, Figure, Unitree, Agility, Boston Dynamics, 1X and the race to general-purpose robots.

  • Figure Lines Up 100,000 Nvidia GPUs in $3.5B Nscale Compute Deal

    Figure Lines Up 100,000 Nvidia GPUs in $3.5B Nscale Compute Deal

    Figure has signed a strategic partnership with Nscale for up to 100,000 GPUs on Nvidia’s Vera Rubin platform to train Helix, the AI model that drives its humanoid robots. The initial commitment is $3.5 billion, with stated intent to scale past $6 billion. First deployment is targeted for the second half of 2027 in Barstow, Texas.

    Why a robot company needs a frontier-lab GPU budget

    Figure’s argument is that humanoid intelligence is now a data-and-compute problem, not a hardware problem. The company says its Index dataset is generating 35 minutes of robot data every second from deployed machines, and that training on it at scale is the bottleneck. A 100,000-GPU cluster puts Figure’s compute in the same league as major AI labs.

    Nscale is also making a strategic investment in Figure, and the two companies say they’ll explore putting Figure humanoids to work in Nscale’s own data-center supply chain — the kind of closed-loop deal that has become a pattern in physical AI.

    “The physical AI flywheel that will accelerate the path from models to robots.”

    Jensen Huang, Nvidia CEO, on the partnership

    Sources

  • Beijing’s Robot Olympics Triples in Size With 2,056 Humanoids

    Beijing’s Robot Olympics Triples in Size With 2,056 Humanoids

    BEIJING — The second World Humanoid Robot Games wrapped up this week at the National Speed Skating Oval — the “Ice Ribbon” built for the 2022 Winter Olympics — and the numbers tell the story of how fast this field is moving. Organizers reported 666 teams from 16 countries fielding 2,056 humanoid robots across 51 events, up from roughly 280 teams and 500 robots at the inaugural games in August 2025. That’s a 138% jump in participation in a single year.

    Less spectacle, more work

    The 100m dash, football and boxing bouts that went viral last year were all back. But the bigger change was the addition of 21 scenario-based events designed to test skills companies will actually pay for: industrial assembly, hotel service, emergency rescue and warehouse logistics. Thirty events remained pure competition; in total the games ran 1,301 individual matches.

    Of the 666 teams, 641 were Chinese and 25 came from overseas — a reminder that while the event bills itself as a world championship, it is first and foremost a showcase for China’s domestic humanoid industry, which now counts Unitree, UBTech, AgiBot and dozens of smaller players.

    Open benchmarks to follow

    Organizers say full performance datasets from the games will be released as open benchmarks, which would give researchers outside China a rare, standardized look at how these machines perform under identical conditions.

    One claim we’re not repeating: a secondary site reported a sub-Bolt 9.32-second 100m. We couldn’t corroborate it with a primary source, so treat it as unconfirmed.

    Sources

  • Tesla Tells JPMorgan Optimus Sales Could Start in Late 2027

    Tesla Tells JPMorgan Optimus Sales Could Start in Late 2027

    FREMONT, Calif. — At a JPMorgan-hosted tour of its Fremont factory, Tesla told analysts that the design of Optimus Gen 3 is finalized, its supply chain is “essentially locked in,” and a production line is being installed in the facility that until earlier this year built the Model S and Model X.

    The headline: external commercial sales of Optimus could begin “as early as” the second half of 2027. Tesla repeated its long-term capacity goals of 1 million units a year at Fremont and 10 million in Texas.

    The context

    Electrek, which covered the tour, notes that the Gen 3 reveal has slipped repeatedly and that Elon Musk has acknowledged no Optimus units are yet doing useful work inside Tesla. The Model S/X line was torn down in April to make room — a real, expensive commitment — but “as early as” is doing a lot of work in that sentence.

    For comparison: Unitree shipped more than 5,500 humanoids in 2025 and Figure has deployed units with paying customers. Tesla’s bet is that when it does ship, it ships at automotive scale. That has never been done in robotics.

    Sources

  • Unitree Soars Nearly Sixfold in Blockbuster Shanghai IPO Debut

    Unitree Soars Nearly Sixfold in Blockbuster Shanghai IPO Debut

    SHANGHAI — Unitree Robotics, the Hangzhou company whose dancing and backflipping robots have become the public face of Chinese humanoids, listed on Shanghai’s STAR Market today under ticker 688836 — and promptly delivered one of the wildest debuts in the exchange’s history.

    Priced at 150.8 yuan, the stock opened at 1,100 yuan and closed at 883.87 yuan, up roughly 486%. The offering raised about $900 million for a 10% stake. Depending on whose math you use, that values the company somewhere between $50 billion (Reuters) and $66 billion (SCMP and Fortune) — far above any US humanoid rival.

    A real business underneath the hype

    Unlike most robotics IPO candidates, Unitree is profitable. Filings show 2025 revenue of 1.699 billion yuan and net profit of 590 million yuan, with humanoids now making up 51.78% of sales. The company shipped more than 5,500 humanoid units in 2025 — a volume no Western company has publicly come close to.

    The catch

    Three weeks before the listing, the US FCC added foreign-made mobile robots over 2 kg to its Covered List, effectively banning imports of new Unitree models into the United States on national-security grounds. The market clearly didn’t care today. Whether it matters over the next five years is the real question for shareholders.

    Sources

  • DeepMind’s Gemini Robotics 2 Brings Whole-Body Control to Humanoids

    DeepMind’s Gemini Robotics 2 Brings Whole-Body Control to Humanoids

    Google DeepMind has released Gemini Robotics 2, a family of three models that gives robots whole-body control, multi-minute task planning and the ability to coordinate with other robots. It’s the most significant update to Google’s robot “brain” since the original Gemini Robotics launched, and it lands as every humanoid company on earth is racing to build or buy exactly this kind of model.

    The three models

    • Gemini Robotics 2 (VLA) — the vision-language-action model that turns instructions and camera input into motor commands. Early access.
    • Gemini Robotics-ER 2 — the embodied-reasoning model for planning and spatial understanding. Available now in Google AI Studio.
    • On-device variant — runs locally on the robot for low-latency control. Early access.

    Demos ran on Apptronik’s Apollo 2 humanoid, Franka’s Duo gripper and Sharpa Wave dexterous hands, and DeepMind says the models can adapt to a new robot body in hours rather than weeks. Reported success rates included 76.3% on a shelf-pickup task, 89.6% on insertion and 92% on unscrewing light bulbs.

    Safety as a benchmark

    Alongside the models, DeepMind introduced ASIMOV-Agentic, a safety benchmark for agentic robots. In demonstrations, robots refused commands judged unsafe and asked a human for help when uncertain — behavior that will matter enormously once these systems are operating around people rather than in labs.

    Sources