Neocloud inference debate: 95% GRR and 400% NDR beat 'reverse auction with one-line API change'
Ronangmi · x · 2026-10-15
- Ronangmi pushes back on saranormous's claim that inference is a commoditized "reverse auction with a one-line API change," clarifying that the metrics at issue are the 95% GRR and 400% NDR posted by inference leaders — best-in-class stickiness and deep customer partnerships, not training cluster build commitments.
- She argues a Neocloud model that builds and sells compute via long-term commitments underwritten by IG credit doesn't get you the serving stack, production muscle, customer trust, or freely allocated supply.
- She concedes inference is a huge market and new players may still carve out niches, citing Digit.
More from Infra
- Decision Models mining to support smaller models, even CPU-only like Laya — airesearch12 · 2026-10-15
- Tree packing lands in TRL AsyncGRPO, cutting agentic RL trainer compute 26x — SergioPaniego · 2026-10-15
- JSSA sparse attention hits 6.43x decoding throughput at 128K context on B200 GPUs — SNU-VLSI · 2026-10-15
- 189 benchmarks later: how to pick KV quants for Qwen3 27B on 16GB VRAM — Fancy-Snow7 · 2026-10-15
- Energy, compute, data are the only new currency: $15B Anthropic-SpaceX deal caps 22 days of AI shakeups — sierracatalina · 2026-10-15
- Titan RTX + 3090 without NVLink: is multi-GPU inference via Strata worth it? — Express_Quail_1493 · 2026-10-15