The $2.3 Billion Race to Teach AI Using Video Games

General Intuition has raised a staggering $320 million Series A funding round at a $2.3 billion valuation, defying the broader venture capital slowdown. Led by Khosla Ventures, the massive capital injection underscores a growing appetite for high-risk, high-reward AI infrastructure plays. This outlier transaction proves that early-stage deep tech ventures can still command premium valuations if they promise a fundamental breakthrough. While average Series A rounds hover in the tens of millions, General Intuition has managed to secure late-stage scale capital before delivering a commercial product.

The core of General Intuition's thesis lies in how it trains its models. Instead of scraping the open internet or relying on expensive real-world robotics data, the startup is leveraging video game environments to teach machines how to navigate the physical world. This approach targets the scaling bottleneck currently crippling traditional large language models. By utilizing both gameplay video and player inputs, the company plans to build advanced world models and large action models faster and cheaper than competitors.

The investor cap table reads like a who's who of deep tech backing, featuring General Catalyst, Hillspire, Bezos Expeditions, and former Formula One champion Nico Rosberg alongside leading academic researchers. This collective backing signal-boosts a strategy that Goldman Sachs analysts identify as the next phase of AI investment, which is moving rapidly into the physical economy. The funding will immediately be deployed to acquire high-fidelity gaming data and scale up compute infrastructure. By bypassing physical world trials in the early stages, the startup aims to slash model training costs by a whole order of magnitude.

For the broader startup ecosystem, this round resets the ceiling for what is possible in early-stage fundraising. It signals to founders that contrarian data collection strategies, such as using simulated gaming environments, are highly investable. However, it also widens the gap between foundational deep tech plays and standard software-as-a-service startups, which continue to face compressed valuations. VCs are increasingly concentrating their capital into a select few high-conviction bets rather than spreading it across a wide portfolio.

Over the next twelve months, the pressure will be on General Intuition to prove that virtual gaming skills can successfully translate into real-world robotic dexterity. The tech industry will closely watch whether these simulation-trained models can handle physical environments without failing. If successful, we will see a gold rush of AI startups partnering with major gaming studios to license gameplay data. This transaction could mark the beginning of a paradigm shift where simulated worlds become the primary training ground for physical automation.