Copper Interconnects May Limit Scale-Up Expansion
tengyanAI · x · 2026-07-14
Copper interconnects have a very short range, and the author believes they face a hard physical limit, with signals significantly degrading after just a few meters.
If accurate, this means that before NPO/CPO truly scales, scale-up interconnects will remain restricted to a rack height of one or two layers. Cited opinions suggest that mass production for CPO might be delayed until 2029 due to gaps in yield and design. Consequently, copper interconnects will delay this tipping point even further, meaning $NVDA and $TSM will remain exposed to this CPO bottleneck for an extended period.
More from Infra
- Is inference latency becoming the biggest bottleneck for production AI agents? — Euphoric_Sea632 · 2026-09-11
- LLM Serving Metrics Thread: Why TPOT and Uptime Make or Break User Experience — abhijithneil · 2026-09-11
- PlanetScale launches sharded Postgres: 768 servers acting as one, 1PB scale — dhruv2038 · 2026-09-11
- Can a 7900 XTX 24GB run Qwen locally? Reddit seeks ROCm tok/s benchmarks — thenomadexplorerlife · 2026-09-11
- RTK Terminal Compression Cuts Tokens but Leaves Your AI Coding Bill Unchanged — Bartaseth · 2026-09-11
- SF Compute founder: buying compute is 'an absolutely awful experience' right now — IgorCarron · 2026-09-11