Yaroslav Bulatov's bear case for local models: data centers win on intelligence per dollar
yaroslavvb · x · 2026-10-22
Yaroslav Bulatov lays out a bear case for on-device models: intelligence per dollar is better in a data center; networking isn't the bottleneck — Battlefield streams fully rendered remote frames at 60 fps over GeForce; and on security, quant funds and TSMC already keep secrets in the cloud. Counterpoint he notes: SF cell networks are worse than Rwanda's (tested during ICLR 2024), so on-device intelligence may still be needed.
Related event: Why local models lose to the cloud: Yaroslav Bulatov's three-part bear case(2 posts)→
More from Infra
- DeepSeek has open-sourced every layer of an AI lab, from weights to inference engine — examachine · 2026-10-22
- Redditor converts missing models to OpenVINO IR and takes requests — Sudden_Topic5154 · 2026-10-22
- Meta plans up to $145B in capex for 2026, more than double 2025 — Polymarket · 2026-10-22
- pi ships rec inference engine, runs locally on M5 Max in two lines — _rockt · 2026-10-22
- Researcher warns GitLab flaw is 'really bad,' urges users to patch now — basedjensen · 2026-10-22
- PyTorch Con Demo Theater: 13 talks including Wan 2.2 video generation sped up 53.6x — PyTorch · 2026-10-22