CES 2026: Google DeepMind and Boston Dynamics announced a partnership. Gemini Robotics will be integrated into Atlas humanoids. Testing begins at Hyundai facilities.
This isn’t new. Figure AI already did this in 2024/2025 – with OpenAI, at BMW. And not just as a demo: 11 months of deployment, 10-hour shifts, over 90,000 parts loaded, 1,250+ runtime hours.
The question isn’t “does it work?” anymore. The question is: “Who scales better?”
Figure Proved It Works
February 2024: Figure AI + OpenAI partnership. $675M raised, $2.6B valuation.
2025: 11 months operational at BMW Spartanburg. Over 90,000 parts handled, contributing to 30,000+ X3 vehicles. This wasn’t a pilot. This was real deployment.
The problem: Figure ended the OpenAI partnership in 2025. LLMs were becoming “commoditized,” so they’re now building their own AI. As of October 2025: No Figure robots in operation at BMW.
Figure proved it works. But not that it scales.
What Boston Dynamics Brings
Hardware: Atlas has decades of advantage in mechanics and control. Figure 02 is good. Atlas is arguably the best humanoid in the world.
Operational Experience: 2,000+ Spot robots deployed. Not demos – operational scale over years. Boston Dynamics knows how to deploy and support robots at scale.
AI Stack: Gemini Robotics built on Gemini, one of the most advanced multimodal models. DeepMind hired Aaron Saunders (former Boston Dynamics CTO) as VP of Hardware Engineering in November 2025. Full-stack integration, not licensed APIs.
Go-to-Market: Boston Dynamics is owned by Hyundai. Testing starts in Hyundai’s own factories. No customer acquisition problem – direct access to global automotive supply chain.
Robert Playter (CEO, Boston Dynamics): “It takes two to three years for our customers to get used to having robots in their facility where they’re ready to start expanding.”
That’s knowledge from 2,000+ deployments. Figure has 11 months with one customer. Different level.
The Critical Difference
Figure is a startup. Must convince each customer. Must develop proprietary AI. Must scale manufacturing (BotQ facility: 12,000 units/year planned). Must do everything simultaneously.
Boston Dynamics has: Superior hardware. DeepMind’s AI. Deployment experience. Hyundai’s captive market. They only need to do one thing: Scale.
The classic pattern: First mover proves the concept. Second mover with better stack dominates the market.
What This Means
Figure did the important work in 2024/2025: Proved that AI-powered humanoids work in real manufacturing.
Boston Dynamics must now show: We scale faster.
Figure has momentum and operational proof. Boston Dynamics has tech stack and distribution.
The answer to “Who wins?” isn’t “Who was first?” but “Who scales better?”
2026 will show which approach works: Startup with first-mover advantage, or incumbent with superior stack and captive market.
That’s the real story.
Sources:
- Boston Dynamics (Jan 5, 2026), Figure AI BMW Report (Nov 2025), Figure + OpenAI (Feb 2024)