Alpha invests in Generalist AI, builder of general intelligence for the physical world

We are delighted to announce our investment in Generalist AI’s $400 million Series B round. The financing was led by Radical Ventures with participation from 8VC, Union Square Ventures, Hanabi Capital, and Norwest, alongside existing investors including NVIDIA’s NVentures, Boldstart, Spark Capital, Bezos Expeditions, and NFDG. Generalist is building foundation models for the physical world, with the goal of giving robots the general-purpose intelligence they need to perform complex, dexterous tasks across industrial environments.

Robots have long been capable of doing a narrow set of things extremely well, but they’ve never been able to adapt. Current robotic systems can perform specific, pre-programmed tasks reliably, but often break down the moment conditions change or a new task is introduced. That limitation has kept automation out of enormous portions of the industrial economy that require flexibility, such as sorting, assembly, kitting, and other dexterous work that humans perform naturally.

Foundation models are beginning to change that. The same scaling dynamics that transformed language AI, where model performance improves predictably as data, compute, and model capacity increase, appear to apply to physical interaction as well. If that holds, robots trained on large-scale real-world data could become dramatically more capable over time, unlocking automation in markets that have resisted it for decades.

Building the intelligence layer for robotics

Generalist is pursuing a software-first, hardware-agnostic approach to this opportunity. Rather than building robots, the company is creating the intelligence layer that runs on top of existing robotic platforms. Its flagship model, GEN-1, is trained on more than 500,000 hours of proprietary physical interaction data collected on Generalist hardware. The model architecture, inference stack, and training pipeline are built specifically for physical intelligence, not adapted from systems designed for digital tasks.

A team built for this moment

The Generalist team consists of veteran builders who have already spent years shipping core systems at the industry’s most prominent AI companies:

• Pete Florence (CEO) co-invented Vision-Language-Action models and trained Google DeepMind’s first multimodal LLM.
• Andy Zeng (Chief Scientist) pioneered manipulation robotics at Google DeepMind and invented handheld gripper data collection.
• Andy Barry (CTO) led early development of Boston Dynamics’ Spot robot.
• Evan Morikawa (VP of Engineering) led the launch and scale-up of ChatGPT at OpenAI.

Why we’re excited

The market opportunity here is substantial. Morgan Stanley projects that embodied AI could create $430 billion of annual value for S&P 500 companies alone through labor replacement. Goldman Sachs estimates the global humanoid market will reach $38 billion by 2035. Millions of robots are operating in the world today, and that number is expected to grow by orders of magnitude across factories, warehouses, laboratories, and elsewhere. Nearly all of them will need general-purpose intelligence to truly become useful.

The combination of a differentiated data strategy, a hardware-agnostic architecture, strong early commercial validation, and one of the most credible technical teams in robotics gives us high conviction that Generalist is building something with real and lasting significance, and we are proud to be part of it.