AI inference to hit ~$350B by 2027, nearly 2x the database market — Tunguz on software's new center of gravity
rseroter · x · 2026-10-08
Theory Ventures' Tomasz Tunguz argues AI inference will overtake databases as software's most important market: $25B spent running AI models in 2025, $130B this year (approaching Gartner's $161B database forecast), and $350B by 2027 — nearly 2x databases at $190B. Every application becomes an inference reseller.
Three unit-economics shifts follow:
- Usage dwarfs platform fees: inference bills exceed seat licenses, forcing predictable pricing, new AE comp plans, and harder forecasting
- Gross margins compress below SaaS norms: the classic 72% blended margin decays unless apps build proprietary harnesses to cut token overhead, or model competition outpaces supply-driven price hikes — upending CAC and payback math
- BYOK trades revenue for margin: enterprises bringing their own GPU clusters or API credentials erode vendor pricing power
Like the mid-2010s database market where 2% share meant an IPO, inference is where software's next decade is decided.
More from Venture
- Coinbase Institute paper: stablecoins fit agent-to-agent micropayments in AiFi era — MurrLincoln · 2026-10-08
- xAI Bets on Leasing Data Centers Over Owning a Frontier Lab, Says Investor — PaulYacoubian · 2026-10-08
- Tech Drives 3/4 of S&P 500 Earnings Growth as AI Trade Disperses — demian_ai · 2026-10-08
- Nigerian startup Salary Index nears ₦1M revenue in under two months on RevIndex — emeka_boris · 2026-10-08
- Nebius up 160% vs CoreWeave's 10%: the AI infrastructure stock divergence explained — Beth_Kindig · 2026-10-08
- Together scales open-source inference with IBM and NVIDIA on B300 cluster — togethercompute · 2026-10-08