Moonshots EP 156: A Humanoid Robot in Every Home? It’s Closer Than You Think w/ Brett Adcock

Summary

Peter Diamandis interviews Brett Adcock, founder/CEO of Figure AI, at the Abundance 2025 Summit. This is a deep technical and business dive into Figure’s trajectory and the Helix AI system. Adcock frames the humanoid robot as “the ultimate deployment vector for AGI” — a single platform that with no hardware changes can do everything a human can. Figure went from C Corp filing to walking robot in under 12 months, and from cold start to shipping first commercial robot in 31 months.

The Helix reveal is the centerpiece: Figure’s internally-built vision-language-action model (replacing their earlier reliance on OpenAI systems) that runs a single neural net per robot, enabling generalization — robots handling objects they’ve never seen in training, understanding semantic relationships (picking up a cactus toy when told to “pick up the desert item”), and doing multi-robot handovers with emergent coordination behaviors (robots look at each other as a release signal, which emerged from training rather than being programmed). Helix was trained on only 500 hours of data. Adcock calls this “probably the most important AI update for robotics in human history.”

The business case is staggering: human labor is roughly half of global GDP ($50-60T TAM). Figure has two commercial customers (BMW in Spartanburg, SC, and a major logistics company). The second customer was onboarded in under 30 days with Helix (vs. a year for BMW). Adcock believes they could now replicate it in 48 hours. If Figure had 100,000 robots today, they’d sell all of them immediately, with 50 more Fortune 100 customers ready to sign. Production pricing target: $20-30K per unit at scale, which at a $300/month lease is $10/day or $0.40/hour.

Figure 3 is design-complete: 90% cheaper than Figure 2, smaller, less mass, better sensors, with hands/head/feet designed specifically for neural nets. Production manufacturing starts this year. Home deployment enters alpha testing (engineers’ homes) this year, though Adcock acknowledges the home is “the Wild West” — vastly harder than commercial workforce due to semantic intelligence, safety (not falling on humans, not knocking over candles), and infinite environmental variation.

Key Segments

• [00:00-04:00] Why Figure exists: AGI needs a body, humanoid as single platform for all physical tasks
• [04:00-08:00] Iteration speed (new hardware every 12-18 months), vertical integration, BMW deployment
• [08:00-14:00] Pricing ($20-30K at scale, $0.40/hour), team building and culture (5-7 days/week, anti-Silicon Valley)
• [14:00-20:00] Helix AI system: single neural net, semantic grounding, emergent robot-to-robot coordination, trained on 500 hours
• [20:00-26:00] Figure 3 (90% cheaper, design-complete), home alpha testing this year, data-bound not capability-bound

Notable Claims

• Figure went from C Corp filing to walking robot in under 12 months
• Helix was trained on only 500 hours of data
• Second commercial customer onboarded in under 30 days (vs. 1 year for first); could now do it in 48 hours
• If Figure had 100K robots today, all would be sold; 50 Fortune 100 companies ready to sign
• Figure 3 is 90% cheaper than Figure 2 with better sensors and neural-net-optimized extremities
• Home alpha testing (engineers’ homes) planned for this year
• Human labor is ~50% of global GDP ($50-60T), making this potentially the largest market in history

Bias/Sponsor Notes

Diamandis’s Bold Capital is an early investor in Figure — disclosed on stage. This is a celebratory founder interview at a paid event with no critical pushback on timelines or claims. The claim of “most important AI update for robotics in human history” goes unchallenged.