Venture capitalists are now spending AI tokens each day to stay ahead of the curve, a practice exemplified by Activate AI co‑founder Pratyush Choudhury.
Why deep tech knowledge has become a must‑have
Traditional VC deals often hinged on evaluating founding teams, product‑market fit, and growth metrics. In the AI era, investors argue that understanding the underlying models is essential to spotting sustainable opportunities. “Unless you fundamentally understand what the technology can and cannot do, I don’t think you can build a truly great AI company,” Choudhury told Rest of World.
He adds that the era of “business‑only” founders launching AI firms without technical fluency is likely over, at least for the next few years. The focus, he says, is on founders who can anticipate technical advances and apply them to lasting customer problems, rather than chasing fleeting capability gaps that a new model release could erase.
How the fund burns through tokens
Choudhury’s daily routine involves consuming between 300 and 500 million tokens to test, prototype, and evaluate AI tools. This translates into a spend of a few hundred to a few thousand dollars per day, a cost he says would quickly deplete the fund without subsidized access from partners.
Most of the usage comes from frontier models such as OpenAI’s Codex and Anthropic’s Claude, which together account for roughly 90 % to 95 % of his consumption. He also taps into Google’s Gemini and Grok extensively.
“I use frontier models for almost all of my work, from brainstorming event formats and guest lists to researching startups to implementing research papers on GitHub,” he explained. The expense is steep, but Choudhury believes it is a necessary investment to maintain a competitive edge.
Related: Communities Seek Control Over Personal Data
Building a global AI network
To keep pace with rapid advances, Choudhury reads research papers and relies on platforms like X to filter which studies merit deeper review. He maintains regular conversations with researchers and applied‑AI builders across the United States, Europe, and China, seeking to understand second‑order effects that could reshape markets.
For example, a foundation model emerging from that country might influence a legal‑tech startup in the United States, while a semiconductor breakthrough in India could benefit a company in London. By mapping these cross‑regional trends, he hopes to anticipate shifts before they become obvious to competitors.
Activate AI, the $75 million fund launched in December with co‑founder Aakrit Vaish, already participated in the round that lifted Indian AI startup Sarvam to unicorn status, surpassing a $1 billion valuation. Vaish, who founded one of India’s first conversational‑AI firms Haptik and advised the national IndiaAI mission, brings experience that complements Choudhury’s technical focus.
In many ways, this approach mirrors the early days of internet venture investing, when capital flowed to companies that grasped the underlying protocols. The difference now is the speed at which models evolve, demanding daily hands‑on experimentation rather than periodic due diligence.
As the AI field continues to mature, the line between investor and technologist blurs. Whether this token‑burning strategy will consistently produce the next unicorn remains to be seen, but for now, it reflects a new reality where deep technical immersion is as important as traditional financial analysis.