Tuhin Srivastava
Showing 1–2 of 2 transcripts.
- RAISE Summit19 min
Inference's Scaling Era | Tuhin Srivastava, Baseten & Corinne Riley, Greylock | RAISE Summit 2026
Tuhin Srivastava, Corinne Riley
Base10 CEO Thuy Hinh projects that AI inference will consume $2.5 trillion of the estimated $5 trillion global AI spend by 2030, prompting the company to deploy an inference cloud with clusters across 90 regions to lower customer infrastructure costs. Driven by open-source models closing the performance gap with frontier closed-source alternatives in just three months, enterprises are increasingly shifting spend to custom models to reduce costs by 70% and retain data sovereignty. Hinh predicts that within 12 to 18 months, a Fortune 500 company will achieve 90% open-source adoption as the market moves toward verticalized cloud solutions capable of supporting high-volume tasks like code generation and biological compound synthesis.
- Stanford Online49 min
Stanford MS&E435 Economics of the AI Supercycle | Spring 2026 | Applications, Applied AI
Tuhin Srivastava, Apoorv Agrawal, Doohan
Base10, led by CEO Toohin, provides a managed inference infrastructure that powers over 30 trillion tokens daily by aggregating 18 clouds to optimize costs for custom open-source AI models. The company differentiates itself from hyperscalers by abstracting complex hardware management and enabling customers to post-train proprietary models, a strategy driven by a thesis that GPU scarcity will remain permanent as agentic demand grows exponentially. With a projected capital expenditure of $7 billion to secure 150,000 B200 equivalents, Base10 aims to industrialize AI deployment through modular data centers and a transition from compute markup to token-based pricing.