Descriptions:
Gradient, the AI-focused seed fund that originally launched in 2017 with Google as its sole LP, has closed a $220 million fund and is now diversifying its investor base for the first time. In a Bloomberg Technology interview, Gradient’s managing partner explains that the nine-month fundraise was challenging given limited liquidity in the institutional LP network, but argues the firm’s decade-long track record — backing over 500 AI founders before the ChatGPT moment made AI investing fashionable — made the case to new partners.
Central to Gradient’s pitch is a pointed thesis about where the bubble is and where it isn’t. The firm sees significant overvaluation in foundation model companies competing directly with OpenAI, Google, xAI, and Anthropic, arguing that the moat at the model layer is simply capital — whoever raises the most buys the most data and compute. Instead, Gradient writes average $3 million checks into AI infrastructure and application companies, targeting 10–15% ownership at seed, where valuations remain attractive even as headline AI funding numbers surge.
The fund sees roughly 2,000 companies per year pitching AI technology, up from about 100 annually before the transformer era, and selects just 10–15 to back. The interview also touches on the competitive dynamics of so-called “avocado seeds” — the billion-dollar seed rounds being raised by some AI contenders — and why Gradient believes those bets are structurally different from the early-stage infrastructure and application companies it targets.
📺 Source: Bloomberg Technology · Published March 18, 2026
🏷️ Format: Interview
