Why Every Tech Giant Suddenly Wants a Humanoid Robot

A year ago, a humanoid robot was mostly a conference-stage demo. In 2026, it's a line item. Tesla, Boston Dynamics, Figure AI, Agility Robotics, 1X Technologies, and a fast-growing pack of Chinese manufacturers led by Unitree are all racing to put robots with two legs and two arms into factories, warehouses, and eventually homes. Billions of dollars have moved in the last twelve months alone. The harder question is how much of the robot workforce promised in press releases is actually clocking in anywhere yet — and the honest answer is: less than the hype suggests, but more than skeptics expected even two years ago.

Three different bets on the same shape

The companies chasing humanoids aren't actually chasing the same goal, and that's easy to miss because the robots look so similar.

Tesla is betting on vertical integration at automotive scale. Optimus is meant to be built the way Tesla builds cars — in-house, at volume, eventually cheaper than any competitor could match. The company has talked about a production line aiming for roughly a million units a year and a long-term ambition of 10 million units annually, with a second dedicated factory at Giga Texas expected around 2027. But Tesla has never published an actual Optimus production count, and on the company's own Q2 2026 earnings call, Elon Musk cautioned that output would be "extremely slow at first" and called this year's production rate "literally impossible to predict." Independent estimates put the real fleet at roughly 1,000–1,200 units, mostly doing light internal tasks at Tesla's own factories — a long way from the figures sometimes repeated online.

Figure AI is betting on partnerships with existing manufacturers rather than building its own factories. Its robots have been running multi-shift pilots on BMW's production line in Spartanburg, South Carolina, focused on dexterous tasks like sheet-metal handling. The strategy has been expensive to fund but easy to sell to investors: Figure's valuation reportedly reached $39 billion after a Series C round in September 2025 backed by Nvidia, Microsoft, and Jeff Bezos, and the company has since struck a multibillion-dollar compute partnership to train its models.

Boston Dynamics and Hyundai (which bought Boston Dynamics in 2021) took the slowest, most industrial route. At CES 2026 in January, they unveiled a production-ready, fully electric version of Atlas — 56 degrees of freedom, able to lift roughly 110 pounds, controllable autonomously or by a human operator. Boston Dynamics says it began manufacturing immediately, with the first fleets committed to Hyundai's Robotics Metaplant Application Center and to Google DeepMind for AI development. The target customer is Hyundai's own electric-vehicle plant near Savannah, Georgia, where Atlas is slated to help with parts sequencing — but Hyundai's own timeline doesn't expect that assembly-line role until 2028.

What's actually working today is boring, and that's the point

Strip away the keynote choreography and the most credible deployments right now are the least glamorous ones. Agility Robotics' Digit — a purpose-built logistics robot rather than a general-purpose humanoid — has a formal, multi-year deal with logistics giant GXO. At a GXO-run Spanx distribution center in Atlanta, Digit has reportedly been lifting totes off a mobile robot and placing them on a conveyor since October, work it was specifically designed for rather than a flashy showcase of general intelligence. GXO is separately piloting Apptronik's Apollo humanoid with an undisclosed client. Reports that Amazon is also testing Agility's robots in its own warehouses are harder to pin down — neither company has made a joint, detailed announcement.

That gap between claim and confirmation runs through the whole industry. Figure has been credited in some write-ups with "10,000 deployments," a figure industry trackers dispute. Chinese manufacturers like Unitree are frequently described as shipping "tens of thousands" of units, but most of those appear to be research, education, or entertainment units — Unitree's G1, at roughly $16,000, has become a favorite for universities and robotics hobbyists rather than factory floors. The honest state of the industry in late 2026: real industrial humanoid labor exists, but it's concentrated at a handful of named sites, running narrow, specific tasks under heavy supervision — not the autonomous general labor shown in sizzle reels.

The home-robot wildcard

The most consumer-facing bet comes from 1X Technologies, a Norwegian-founded, OpenAI-backed startup that opened preorders for its NEO home robot in October 2025. NEO is priced at $20,000 outright or $499 a month on subscription, with a refundable $200 deposit holding a place in line — and the company's own executives have been candid that early units still rely heavily on remote human teleoperation while the underlying AI learns, something 1X's CEO has described as a deliberate, temporary bridge rather than a bait-and-switch. First US deliveries are promised before the end of 2026, though as of this writing there's no confirmation that customer units have actually shipped.

The UAE is leaning in — carefully

The Gulf has become an enthusiastic adopter of the humanoid moment, if not yet a manufacturing hub for it. Gulf News reported on 1X's NEO pricing as part of its consumer-tech coverage, reflecting regional interest in the category. On the ground, Dubai Future Labs installed a Unitree G1 at the Museum of the Future, letting visitors interact with it as part of a deliberate effort — as the project's engineers put it — to build public familiarity with robotics and "reduce fear" of the technology, per Khaleej Times' coverage of the UAE's broader robotics push. In early 2026, an Emirati-styled humanoid nicknamed "Bu Sunaidah," running on a Unitree G1 platform, went viral after appearing at Abu Dhabi's Make it in the Emirates event. Following the BRIDGE Summit in 2025, there's also been talk of a LimX Dynamics humanoid piloting a public-facing role in Abu Dhabi's malls and restaurants in 2026, though that remains a stated plan rather than a confirmed rollout. A Dubai-based firm, Action to Action Robotics, has also announced what it bills as the region's first production-ready humanoid — an early sign that local manufacturing ambitions, not just imported demo units, are starting to form.

Why this is so hard, and what to actually watch

Humanoid robots are difficult for a reason that has nothing to do with marketing: balancing on two legs while manipulating objects with two hands demands a level of real-time motor control and object-level perception that wheeled robots and robot arms never needed to solve. That's why nearly every serious deployment today narrows the job down — one task, one station, often with a human nearby — rather than promising a robot that can do "anything a person can." The gap between that narrow reality and the general-purpose future every company's investor deck promises is where this industry will be won or lost over the next few years.

The metric worth tracking isn't funding rounds or CES demos — it's whether companies like GXO, BMW, and Hyundai renew and expand their pilots with hard numbers attached, rather than optimistic case studies. That's the difference between a robot workforce and a very expensive proof of concept.

Sources: TechRadar, SiliconANGLE, and Boston Dynamics' own announcement on the Atlas CES 2026 launch; The Robot Report and Interesting Engineering on Agility Robotics' GXO deployment; multiple funding trackers on Figure AI's September 2025 Series C; Gulf News on 1X's NEO pricing; Khaleej Times on UAE robotics adoption and the Museum of the Future's Unitree G1; and public reporting on Tesla's Q2 2026 earnings call.