The lab that had to become a company
For most of its existence Boston Dynamics was the most famous robotics company that sold almost nothing. Founded in 1992 as a spin-off from MIT's Leg Laboratory, it spent two decades building astonishing research machines — BigDog, the hydraulic Atlas, the parkour-and-backflip viral videos — funded largely by DARPA contracts and, later, by deep-pocketed corporate owners willing to bankroll a science project. The robots were the best in the world at dynamic locomotion; the business underneath them barely existed.
The company's defining challenge, and the reason it is interesting to track now, is the pivot from spectacle to product. Spot's 2020 commercial launch was the first real attempt to sell a robot rather than a demo. The all-electric Atlas — a ground-up redesign that replaced the theatrical hydraulic version — pushed that logic to its conclusion: a humanoid engineered not to impress the internet but to be manufacturable, serviceable, and profitable on a factory floor. In January 2026 that machine entered production. The question that will define the next few years is whether the world's best robotics lab can also be a good manufacturer.
A company passed between giants
Few technology companies have changed hands as often, or been valued as inconsistently, as Boston Dynamics. Alphabet's Google X acquired it in 2013, tired of the defense associations and the lack of a business model, and sold it to SoftBank in 2017. SoftBank, in turn, sold an 80% controlling stake to Hyundai Motor Group in 2021 in a deal that valued the company at roughly $1.1 billion. Each owner inherited the same asset — world-class locomotion IP with no proven path to profit — and each eventually decided that path was someone else's problem.
Hyundai has broken that pattern by doubling down rather than moving on. In June 2026 it bought out SoftBank's remaining ~9.65% stake for about $325 million, making Boston Dynamics a wholly owned subsidiary — a transaction that implied a valuation near $3.4 billion, more than triple the 2021 mark. The buyout, which exercised a put option baked into the original deal, matters less for its price than for what it signals: Hyundai is not a passive financial owner waiting to flip the asset, but an industrial parent integrating Boston Dynamics into its own manufacturing future.
Where the revenue actually comes from
The Atlas headlines obscure the fact that Boston Dynamics is already a revenue-generating company — and that its revenue today comes from quadrupeds and warehouse arms, not humanoids. Spot, the agile inspection quadruped commercially available since 2020, and Stretch, the box-unloading warehouse robot, together account for more than 2,000 units deployed across hundreds of customer locations globally. Orbit, the fleet-management software layer, ties them together with dashboards, alerts, and increasingly AI-driven inspection analytics.
This installed base is the company's most underrated asset. Stretch in particular has real logistics traction: DHL signed a 2025 memorandum of understanding for more than 1,000 additional units by 2030, and Lidl is folding Stretch into regular import-warehouse operations across the Netherlands, Belgium, Austria, and Spain. Spot, meanwhile, keeps finding high-visibility deployments — four units helped secure the 2026 FIFA World Cup, patrolling MetLife Stadium and a broadcast center. These are not science projects; they are recurring commercial relationships that give Boston Dynamics something most humanoid startups lack: paying customers and a support organization before the humanoid even ships in volume.
The Atlas bet and the AI question
The all-electric Atlas is Boston Dynamics' wager that the same locomotion mastery that produced the backflips can be redirected toward dull, valuable factory work. The production robot carries 56 degrees of freedom with fully rotational joints, a 2.3-meter reach, and the strength to lift up to 50 kg — with internal tests handling 100+ lb loads. In May 2026 the company demonstrated whole-body manipulation of a ~50 lb mini-fridge with shifting internal contents, trained through GPU-parallel reinforcement learning in simulation and transferred zero-shot to the real robot, without visual identification of the object.
Hardware, though, was never the hard part for Boston Dynamics; general-purpose intelligence is. That is why the CES 2026 partnership with Google DeepMind is strategically central. DeepMind's Gemini Robotics foundation models already power Spot's AIVI-Learning inspection reasoning as of April 2026, and the roadmap extends those models into Atlas for broader task learning beyond hand-authored RL policies. Boston Dynamics is effectively betting that its best-in-class bodies plus DeepMind's frontier robot-learning models can leapfrog the vertically integrated approaches of rivals like Tesla and Figure — while conceding that it does not intend to build the brain alone.
The Hyundai machine and the financial reality
Boston Dynamics' economics only make sense in the context of its parent. Hyundai has committed a $26 billion U.S. investment plan that includes a dedicated robotics factory targeting 30,000 Atlas units per year by 2028, and has signaled demand for tens of thousands of robots across its own plants — an anchor customer relationship no humanoid competitor can currently match. The entire 2026 Atlas production run is already spoken for, split between Hyundai's Robotics Metaplant Application Center and Google DeepMind, with outside customers slated to begin in early 2027.
The flip side is that this remains a deeply unprofitable business. Public disclosures via Hyundai affiliates put 2025 revenue around ₩150 billion (~$109 million) against years of accumulated losses, with cumulative Hyundai capital injections running into the billions of dollars since 2021. The Atlas price target — below two years of U.S. manufacturing payroll, roughly $320,000 — hints at the unit economics Boston Dynamics needs to hit before humanoids pay for themselves. For now, the company is a bet being funded by an automaker willing to absorb losses to own the robotics stack that could reshape its factories.
What to watch next
Three threads will determine how the next 18 months read. The first is the long-anticipated Nasdaq IPO: with Hyundai now sole owner and SoftBank's put option resolved, Boston Dynamics is widely reported to be selecting underwriters and preparing a preliminary review ahead of a listing that could come as early as the first half of 2027, at implied valuations that have ranged from roughly $21 billion toward far higher marks as the humanoid sector re-rates. Whether public-market investors reward the science or punish the losses will be a referendum on the entire humanoid thesis.
The second is leadership. Founder and long-time CEO Robert Playter retired in February 2026 after roughly three decades, leaving CFO Amanda McMaster as interim chief while the board runs a search. The permanent appointment will say a lot about whether Hyundai wants a roboticist or an operator to scale manufacturing. The third is execution against a crowded field: Tesla's Optimus, Figure's Series C-funded fleet, and low-cost entrants like Unitree are all racing for the same factory and logistics work. Boston Dynamics enters with the best hardware pedigree and the strongest industrial backer in the field — but for the first time, its success will be measured in units shipped and margins earned, not videos watched.
Sources
- Boston Dynamics — Unveiling the production all-electric Atlas (CES 2026)
- Hyundai Motor Group — Completes acquisition of Boston Dynamics from SoftBank
- Boston Dynamics — Training a humanoid robot for hard work (heavy-object manipulation)
- Boston Dynamics — AIVI-Learning now powered by Google Gemini Robotics
- DHL Group — MOU with Boston Dynamics for 1,000+ Stretch units by 2030
- Seoul Economic Daily — Boston Dynamics valuation and 2025 financials