Nvidia says new AI server boosts Chinese mixture‑of‑experts models tenfold
Summary
Nvidia released benchmark data showing its latest AI server, which packs 72 of its top chips into a single system, can deliver roughly a 10x performance gain when serving large mixture‑of‑experts models such as Moonshot AI’s Kimi K2 Thinking and DeepSeek’s models. The results aim to show that even as some new models train more efficiently, Nvidia’s high‑end servers remain critical for large‑scale inference, reinforcing its dominance against rivals like AMD and Cerebras in the AI deployment market.
Companies Mentioned
Related Deals
OpenAI and NEXTDC entered a multi-year agreement under which OpenAI will anchor a hyperscale AI campus and GPU supercluster at NEXTDC’s S7 facility in Sydney to support large-scale AI inference and enterprise workloads.
OpenAI agreed to acquire neptune.ai to integrate its experiment‑tracking and training‑monitoring tools into OpenAI’s frontier model training stack.
OpenAI provides custom holiday tools built on ChatGPT to enhance NORAD’s annual Santa‑tracking program as a public‑facing AI partnership.
OpenAI acquires an equity stake in Thrive Holdings in exchange for dedicated AI research and integration support to modernize Thrive’s accounting and IT services portfolio.
OpenAI is acquiring an equity stake in Thrive Holdings and embedding its teams to deploy AI across Thrive’s accounting and IT services portfolio companies.
Related News
Greater Bay Area firms use AI-powered devices to power a ‘first-launch economy’ at AIE expo
TodayEnthusiasm for OpenAI’s Sora fades after early creative burst
TodayLi Auto launches Livis AI glasses as its first standalone AI hardware
Today