Beyond Tesla Optimus： How Figure AI Plans to Win the Robot Race

Figure AI is planning one of the biggest AI infrastructure projects ever attempted in humanoid robotics.

The $39 billion Silicon Valley robotics startup says it could deploy up to 100,000 next-generation Nvidia GPUs to train its Helix AI system, backed by a multibillion-dollar investment that could eventually reach $6 billion. The goal is to give Figure’s AI-powered humanoid robots the ability to see, reason, move, and adapt to real-world tasks with far greater autonomy.

In this video, we break down how Figure plans to use massive amounts of human-generated training data, why it believes compute and data are the keys to general-purpose humanoid intelligence, and how its strategy compares with Tesla Optimus and the growing global humanoid robotics race.

We also look at Figure’s latest robot demonstrations, its expanding data pipeline, the scale of its planned AI supercomputer, and the enormous financial challenge of funding this vision.

Could scaling AI infrastructure for robots work the same way it did for generative AI? And could this be the path toward truly useful general-purpose humanoid robots?

00:00 Figure's Massive AI Supercomputer Bet
01:25 A $3.5 Billion Commitment
01:48 100,000 Nvidia Vera Rubin GPUs
02:30 Figure’s Index Global Human Data Program
03:10 Turning Human Video into Robot Training Data
03:50 How Helix Learns from People
04:11 The Robot Training Data Gold Rush
04:47 New Breakthroughs in Physical AI
06:18 Why Figure Built Its Own Data Pipeline
06:36 Figure 03’s Growing Autonomous Capabilities
08:23 Figure’s Multibillion-Dollar Spending Problem
09:26 Figure 03 vs Tesla Optimus
11:06 A Very Different Robot Ending

#machine #Robotics #robot