Brex Summer List: AI Infrastructure Beats Apps
Brex, a corporate payments company, published its summer list of the 25 fastest-growing software vendors based on card and bill-pay data from tens of thousands of customers. Its core conclusion: infrastructure won — hype at the AI application layer is cooling while underlying infrastructure has become the real growth engine, outpacing consumer-facing app companies.
Confirmed
- 14 of the top 25 are not end-user AI products but infrastructure for building them (compute, databases, GPU, sandboxes); another 6 are AI products built on that infrastructure.
- Together AI, an open-model provider, ranks first in growth; Upstash is in the top ten; Daytona made the list.
- Databases are the second-fastest-growing infrastructure: Neon says over 80% of new databases are created by AI agents, and Supabase reports over 60%. Agents generating thousands of apps a day demand databases that provision in seconds, cost near zero when idle, and clone instantly.
- Supabase's share of startups' first independent database purchase rose from about 15% in 2023 to 38% in 2026, while Pinecone fell from 22% to under 2%; vector search has not disappeared.
- Open and closed models run in parallel: closed APIs handle complex reasoning and customer-facing chat, while classification, extraction, transcription and generation shift to open models 5–20x cheaper; 98% of open-compute buyers still pay OpenAI.
- The median time from first paying OpenAI/Anthropic APIs to first buying open-model compute shrank from 23 months in 2022 to 5 months in 2025.
- A whole stack separates open weights from production use: GPU clouds rent machines; Modal and RunPod handle containers and autoscaling; Together AI and Deepinfra wrap models as APIs; Tinker offers managed fine-tuning.
Why it matters
- Commenters like @nutlope and @stuffyokodraws read the data as evidence that agent-economy infrastructure is booming and the base layer is the AI boom's real growth engine.
- @FinanceYF5's series argues that agents mass-creating databases and startups rapidly adopting open-model compute show value in the AI stack shifting toward infrastructure and middle layers — a useful signal for understanding AI economics.
- Independent reposts by @zhyncs42, @mattturck and @AccBalanced point to the same conclusion, indicating broad resonance among practitioners.
2026-08-18 ~ 2026-08-19 · 15 related posts
Primary sources
- Brex Data: 14 of Top 25 Fastest-Growing Vendors Are AI Infrastructure — AccBalanced · 2026-08-18
- Brex benchmark: Over half of top-growing startups build AI infra — mattturck · 2026-08-18
- Brex report: AI hype is in infrastructure, not apps — stuffyokodraws · 2026-08-18
- [source] Brex Summer List: Together AI ranks #1 fastest-growing vendor — nutlope · 2026-08-18
- Brex list: 14 of 25 fastest-growing software vendors serve AI teams, Together AI ranks No.1 — zhyncs42 · 2026-08-19
- Brex: 14 of the 25 fastest-growing software companies sell AI infrastructure — FinanceYF5 · 2026-08-19
- The open-model stack: every layer from GPU clouds to hosted fine-tuning sells 'less to worry about' — FinanceYF5 · 2026-08-19
- Time from closed API to open-model compute shrank from 23 months to 5 — FinanceYF5 · 2026-08-19
- Startups split workloads: closed models for reasoning, open models 5-20x cheaper for bulk tasks — FinanceYF5 · 2026-08-19
- Over 80% of new databases are created by AI agents, Neon says — FinanceYF5 · 2026-08-19
- Supabase now 38% of startups' first database purchase; Pinecone under 2% — FinanceYF5 · 2026-08-19
- [source] Brex data: over half of the fastest-growing startups sell AI infra, not AI products — FinanceYF5 · 2026-08-19
- Brex Report: Infrastructure Wins Over Apps in AI Hype — simonguozirui · 2026-08-19
- Brex Report: Infrastructure Vendors Top Fastest-Growing List — marcbhargava · 2026-08-19
1 near-duplicate retellings: FinanceYF5