Compute Allocation Limits: The Root Cause of Missing Architecture Innovation in European LLMs
IgorCarron · x · 2026-07-21
In response to criticism that recently released European LLMs (like Soofie) lack architectural innovation, an industry insider pointed out the root cause: Europe's compute allocation system is severely limited.
Researchers typically get access to something like 800 H100 hours, which must be spent within a short 60-day window. This highly restricted budget only allows for training one or two models over a fixed amount of tokens, leaving zero room for trial and error to explore new architectures. The commentary also suggests listening to pragmatic researchers rather than chasing grand projects, noting that LightOn successfully built the first French LLM in 2021 using this approach.
More from Infra
- Oracle Q1: Triple-digit cloud infrastructure growth as AI demand booms — DavidLinthicum · 2026-09-11
- SF Compute founder: buying compute is 'an absolutely awful experience' right now — IgorCarron · 2026-09-11
- SmolVM open-sources persistent computer infrastructure for agents that outlive chat sessions — aniketmaurya · 2026-09-11
- PyTorch Day Korea 2026 launches first offline conf, CFP closes Sept 13 — PyTorch · 2026-09-11
- Local LLM server dilemma: 4x CMP-170HX (price up 53% in 20 days) vs Mac Studio M5 Ultra — rumboll · 2026-09-11
- llama.cpp lands Flash Attention tuning for RDNA4, big prefill gains on AMD — pmttyji · 2026-09-11