AI startups rush to build their own models to cut inference costs
nmasc_ · x · 2026-09-22
The top topic at Silicon Valley dinners lately: a serious rush among AI startups to build their own models, aiming to cut costs, reduce dependence on frontier model providers, and regain control — driven by the rise of cheap, effective open-source models.
Examples: Cursor was an early mover building its own models to manage the cost of routing to expensive frontier models; Moonshot is expanding its forward-deployed engineer program; and Harvey's margin pressure is part of the same story.
Related event: Silicon Valley AI Startups Embrace Building Their Own Models(2 posts)→
More from Companies & People
- SPAR expects to spend $1-2M on compute by year-end for AI safety research — austinc3301 · 2026-09-22
- Report: OpenAI has automated much of its experimental model training pipeline — Dr_Singularity · 2026-09-22
- Model reviewers reportedly get unlimited API access during 'short review periods' — shaunralston · 2026-09-22
- Stripe's Emily Sands on agentic commerce: compute is the new cash, fraud is now full-funnel — every · 2026-09-22
- Five Princeton CS grad students named 2027 Siebel Scholars, each getting $35,000 — liuzhuang1234 · 2026-09-22
- Why Meta hasn't shipped Muse in WhatsApp: not enough hardware for 2B users, says user — zephyr_z9 · 2026-09-22