Figure All Hands 2026

In this talk

1. 1:47

   Four chapters to general robotics

   Brett frames the work as a race to build robots that can perform useful tasks across commercial and residential settings. He divides it into four chapters: develop capable hardware, create an AI recipe for autonomous work, scale intelligence with data and compute, and scale manufacturing and deployment. He says the first two require sustained engineering, while the latter two need major capital and organizational capacity.

2. 10:41

   Hardware generations and Figure 04

   Brett reviews three robot generations, describing Figure 01 as an architecture test, Figure 02 as feature-complete, and Figure 03 as designed for Helix and higher-volume manufacturing. He says Figure built its 1,000th Figure 03 over the summer and has Figure 04 in design lock, with parts shipping. He highlights a redesigned hand intended to track human fingertips and contact points with millimeter-level precision.

3. 22:44

   From tabletop tasks to generalization

   Brett describes Helix’s move from tabletop manipulation to whole-body control, where visual input guides every joint as a robot walks up and handles a task. He says Helix 2.5 robots worked in 30 rented homes they had not seen before, with unfamiliar objects and conditions. He also cites a 200-hour logistics run and autonomous operation in new commercial environments as evidence of longer, more robust deployment.

4. 35:53

   Scaling data and compute

   Brett argues that robot capability should improve predictably as data and compute increase, and says Figure observed lower model losses and better evaluation results as pretraining data grew. Figure’s Index initiative, he reports, has received 37.5 million videos from 210,000 contributors across 112 countries, at a rate of 81 minutes of data per second. He says the company plans substantial spending on data and compute, including a $3.5 billion initial commitment with Nscale, with intent to scale it to $6 billion for computing capacity planned for 2027.

5. 50:48

   Manufacturing and the next model

   Brett says Figure is building a new supply chain and scaling manufacturing alongside intelligence, with robots already operating in three states. He reports a cash balance above $2 billion and says the company is investing in data, compute, and production together. He closes by describing a recently trained model as a major step for the team, while saying it is not ready for public release.