Bessemer's Robotics Predictions #1: The GPT-2.5 Moment

Why It Matters

Scaling progress will decide which capital‑rich firms can commercialize intelligent robots, reshaping supply chains and consumer services.

Key Takeaways

• Robotics entering a "GPT‑2.5" scaling phase with larger models.
• Demo shows kitchen robot, but real‑world deployment remains distant.
• Scaling laws emerge from more egocentric and tele‑operated data.
• Massive data volume and compute needed; unlikely by 2026.
• Capital‑rich firms will dominate by harvesting diverse robot data.

Summary

Bessemer analysts say robotics is at a "GPT‑2.5 moment," where larger models begin to show outsized gains similar to early large‑language‑model scaling.

They point to recent demos—Physical Intelligence’s kitchen robot—and research from Nvidia and PI that demonstrate emerging scaling laws when models are trained on more egocentric and tele‑operated data, indicating performance improves with data and compute.

Despite the hype, the speakers caution that moving from lab prototypes to household products will require massive data volumes and compute resources, a hurdle unlikely to be cleared before 2026. They quote that “the absolute volume of data … is still probably not going to happen in 2026.”

The analysis concludes that firms with deep pockets and the ability to collect diverse robot data will capture the biggest gains, positioning them to lead the next wave of commercial robotics.

Original Description

Today we're kicking off a new series where we share our 2026 robotics predictions.

Alexandra Sukin, Investor at Bessemer Venture Partners, breaks down where robotics stands today: scaling laws are emerging, the demos are incredible, and the companies with the capital to collect data at scale are going to pull ahead.

Follow Bessemer Venture Partners so you don't miss the rest of the series.

To learn more → https://www.bvp.com/atlas/bessemer-predicts-robotics-and-physical-ai

#robotics #ai #technology #physicalai