Apptronik has the money, partners and factory access Apollo needs, but humanoid robots only become a real business when they survive dull, repeated work.
Apptronik’s Apollo robot has moved past the easy part of the humanoid story. The easy part is the video: a machine with arms, legs and a friendly face lifting a box or moving an object across a table. The harder part is a warehouse aisle, a production line and a supervisor who doesn’t care how impressive the demo looked. That is where Apollo now has to prove itself.
The Austin company has a serious war chest. In February 2026, Bloomberg reported that Apptronik raised a $520 million extension to its Series A, taking the round to more than $935 million and putting the company’s valuation around $5.5 billion. Existing backers included Google, Mercedes-Benz Group and B Capital, while AT&T Ventures, John Deere and the Qatar Investment Authority came in as new investors. You don’t raise that kind of money for another glossy lab clip. You raise it because customers want to know whether the robot can be built, deployed and maintained at scale.
Apptronik’s most useful detail is not the valuation. It is the customer list. Mercedes-Benz announced in March 2024 that it was exploring Apollo for logistics use cases in manufacturing, including moving parts to workers on production lines. Business Insider later reported that Jabil, the Florida-based electronics manufacturing giant, agreed in February 2025 to test Apollo in its own factories for repetitive tasks such as inspection, sorting, lineside delivery and fixture placement. Those are not glamorous jobs. That is the point. Humanoid robots don’t need glamour. They need boring competence.
The original version of this story overstated the case by saying Apollo was already working factory shifts at Mercedes-Benz and GXO Logistics. The publicly verified record is narrower. Mercedes has been testing or exploring Apollo on manufacturing logistics work, and Jabil has agreed to test and eventually produce Apollo robots if the trials work. I found no reliable current source confirming Apptronik deployments inside GXO warehouses. Frankly, that distinction matters. A pilot is not a scaled deployment, and a robot that works for a filmed task is not yet a robot you can build a staffing plan around.
The software side is where Apptronik gets its strongest outside validation. Google DeepMind introduced Gemini Robotics in March 2025 as a model built to connect language, vision and physical action. The Verge reported in June 2025 that Google’s on-device Gemini Robotics model had been adapted to Apptronik’s Apollo, after being trained on Google’s ALOHA robot, and could run locally rather than depending entirely on a cloud connection. For factories, that is not a small technical footnote. Connectivity drops, security rules are strict, and a robot that needs constant outside access is harder to place in serious industrial environments.
Apollo itself is built for that industrial pitch. Axios reported when the robot was unveiled that it stands 5-foot-8, weighs 160 pounds, can lift up to 55 pounds and uses a swappable battery intended to keep it working across shifts. Those numbers do not make it a worker. They make it plausible enough to test around workers. The difference is everything.
Also read: IBM's 0.7nm chip puts nearly 100 billion transistors on a fingernail and resets the AI silicon calculus • Walmart pays $1.4 billion for French adtech startup Vibe.co in its biggest bet yet on streaming advertising • Amazon raises its India bet to $48 billion as Big Tech turns the subcontinent into an AI proving ground
The timing is still current because the humanoid race has not slowed down. Business Insider reported this week that Agility Robotics plans to go public through a SPAC at a $2.5 billion valuation, with customers including Amazon, Toyota and GXO. Figure has tested robots with BMW. Tesla keeps pushing Optimus as part of its manufacturing future. If you run a factory or logistics operation, you are no longer watching a research curiosity. You are watching suppliers fight to become part of the labor stack.
Apptronik’s advantage is that it has lined up the right kind of partners. Google helps with the model. Mercedes gives it a real manufacturing setting. Jabil gives it a path toward production and a place to test the robot inside the kind of factory that could eventually build it. That combination is stronger than a startup trying to invent hardware, software, manufacturing and customer trust all by itself.
Still, the missing number is the one readers should care about most: sustained performance. How many hours can Apollo work without intervention? How often does it fail a pick, stop for safety, need a human reset or lose time on an object it has never seen before? Until Apptronik publishes or customers disclose that kind of operating data, the company is still selling a credible bet rather than a settled fact.
That is not a knock on Apptronik. It is the real test for the whole sector. Humanoid robotics has plenty of money now, and it has enough demos. Apollo’s next job is much less theatrical: show up, do repetitive work, avoid getting in the way and keep doing it tomorrow.
