NVIDIA DSX integrates compute, power and cooling for energy-efficient AI factories
NVIDIA Developer · youtube · 2026-11-21
NVIDIA's Developer channel showcases DSX, a reference design that unifies compute, power, and cooling to maximize AI factory energy efficiency. The video features Nscale CTO David Power and Firebird CTO Harut Martirosyan discussing how to scale AI infrastructure with energy efficiency as the foundation: an integrated design approach ties infrastructure performance to power optimization, aiming to deliver more useful AI output per megawatt. Relevant for anyone tracking datacenter energy economics.
More from Infra
- Anthropic goes public at $83/share: $130B cash, $431B compute commitments, $77B supplier financing — LexSokolin · 2026-11-21
- From OSS to startups: how vLLM, SGLang and FlashAttention teams monetize inference research — AccBalanced · 2026-11-21
- SpaceX's GPU financing grows to $44B, largest ever, to fund NVIDIA Vera Rubin chips — XFreeze · 2026-11-21
- Unverified: Big Tech building ~20GW of data centers overseas, NVDA cancellations loom — SumitGup · 2026-11-21
- Ex-Super Micro exec took AI cloud CEO job days before $2.5B Nvidia smuggling case broke — jeremyakahn · 2026-11-21
- Siri co-creator launches decentralized inference network to counter AI compute centralization — film_girl · 2026-11-21