Top 10% of AI Customers Drive 99.5% of Model-Serving Spend as Hyperscaler Capex Jumps $750B
rohanpaul_ai · x · 2026-09-26
Apollo chief economist Torsten Slok's data shows AI adoption is spreading but spending is extremely concentrated: the top 10% of customers account for 99.5% of model-serving spend and 99% of neocloud spend, with the bottom 90% at just 0.5% and 1%. The average top-group firm spends 1,791x more on inference (vs. 101x for non-AI SaaS), because AI scales with machine work rather than seats. Rohan Paul adds that combined 2026-2027 hyperscaler capex outlooks (Alphabet, Microsoft, Amazon, Meta, Oracle) are now 66% above the start of 2026 — up $750B.
Related event: Apollo: Top 10% of AI Customers Account for 99.5% of Model Spending(2 posts)→
More from Venture
- Akamai beats neo-clouds on profitability, Anthropic deal and buybacks, argues investor — pdamodaran · 2026-09-26
- Mediabunny Loses First and Longest Gold Sponsor Gling Amid Restructuring — Vjeux · 2026-09-26
- OpenAI's Internal Codex Spend Leaked: $700 Median, $7,000 for Top Users — scaling01 · 2026-09-26
- Fal in talks to raise at $15B valuation; Fireworks weighs $30B round — steph_palazzolo · 2026-09-26
- OpenRouter at 10T+ tokens/day: from dismissed wrapper to critical AI infrastructure — Latent Space · 2026-09-26
- OpenAI Makes 7% of Meta's Revenue, Loses $38B, Yet Worth Half of Meta — Wrong_Swimming_9158 · 2026-09-26