--- title: Vera Rubin racks are live at Microsoft, and the hype... description: Nvidia's next-generation Vera Rubin AI hardware platform has moved from roadmap slide to production, with Microsoft Azure confirming its first Vera Rubin racks... canonical: https://daily.dev/posts/vera-rubin-racks-are-live-at-microsoft-and-the-hype-train-left-the-station-xwy9tybnk twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Vera Rubin racks are live at Microsoft, and the hype train left the station | daily.dev og:description: Nvidia's next-generation Vera Rubin AI hardware platform has moved from roadmap slide to production, with Microsoft Azure confirming its first Vera Rubin racks... og:url: https://daily.dev/posts/vera-rubin-racks-are-live-at-microsoft-and-the-hype-train-left-the-station-xwy9tybnk og:image: https://api.daily.dev/og/posts/XWY9tYBnk.png og:image:alt: Vera Rubin racks are live at Microsoft, and the hype train left the station og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Vera Rubin racks are live at Microsoft, and the hype train left the station **[Trends](https://daily.dev/sources/trends)** · 2 min read · 3 upvotes · 0 comments ## Summary Nvidia's next-generation Vera Rubin AI hardware platform has moved from roadmap slide to production, with Microsoft Azure confirming its first Vera Rubin racks are live and running training workloads. CoreWeave and Google previously stood up racks earlier in the year, and OpenAI's infrastructure lead confirmed its training stack is now running on Rubin hardware, a claim amplified by Sam Altman. The commentary so far is celebratory rather than analytical, with no public benchmarks or skeptical takes yet on power draw, cooling costs, or real-world performance versus Blackwell. ## Content NVIDIA's Vera Rubin systems are now in production at AWS, Microsoft Azure, Google, and CoreWeave, and the benchmark claims coming out of Santa Clara are the kind that make you want to see the fine print. The headline number: up to 30x higher throughput per megawatt and 35x lower cost per million tokens compared to the previous-gen GB300 NVL72, based on SemiAnalysis's AgentX benchmark on agentic coding workloads. NVIDIA is quick to note these results are still pending SemiAnalysis review and don't yet include Vera CPU tool-calling performance, which is a notable caveat given that the Vera CPU is half the pitch. That CPU side is worth paying attention to. NVIDIA built Vera with 88 custom Olympus cores and up to 1.2TB/s memory bandwidth, targeting the orchestration work that sits between GPU inference steps: Python execution, tool calls, retrieval, sandboxed code. The argument is that as agents get more complex, the CPU becomes the bottleneck, and x86 wasn't designed for this workload pattern. NVIDIA claims up to 1.8x faster performance on selected agentic workloads than x86 CPUs. Independent Phoronix testing found Vera about 10% ahead of AMD EPYC 9575F on the geometric mean, which is real but considerably more modest than 1.8x. The catch: NVIDIA restricted the test suite to Vera-relevant workloads, so the comparison is doing some work to look favorable. On the GPU side, the rollout is moving fast. CoreWeave had one running in June, Google had racks by July, and AWS just received its first Vera Rubin GPU server alongside the Vera CPU. Sam Altman retweeted OpenAI's infrastructure team confirming their first Vera Rubin racks are running their training stack. AWS has also expanded its NVIDIA roadmap by 2 million GPUs across Blackwell Ultra, Rubin, and Rubin Ultra. Cisco is also in the mix, adding Vera Rubin NVL72 support to its Secure AI Factory offering through a new Supermicro partnership. The 30x efficiency claim will get scrutinized hard once SemiAnalysis publishes its full review. Until then, the real story is simpler: Vera Rubin is no longer a roadmap slide. ## Questions this post answers ### Is Nvidia's Vera Rubin AI platform actually in production yet? Yes, Microsoft Azure has confirmed its first production Vera Rubin racks are live and running training workloads. CoreWeave had racks running by June and Google by July, following Nvidia's January announcement naming Microsoft, AWS, Google, and CoreWeave as first deployment partners. OpenAI's infrastructure lead also confirmed its training stack is now running on Vera Rubin hardware. _Track how Vera Rubin's rollout reshapes AI infrastructure choices as benchmarks emerge on daily.dev._ ### Which cloud providers got early access to Nvidia Vera Rubin hardware? Nvidia named Microsoft, AWS, Google, and CoreWeave as the first companies to receive Vera Rubin deployments when it announced the platform in January. CoreWeave had systems running by June, Google had racks by July, and Microsoft's production systems went live on Azure afterward, with OpenAI confirmed as a workload running on the new hardware. _Compare hyperscaler AI infrastructure bets like this one on daily.dev before picking a training platform._ ## Community take How the wider developer community reacted, aggregated from 1 discussion and 2 comments across x (as of 2026-08-28). **TL;DR:** Limited reaction focuses on NVIDIA framing agent workloads as a CPU/orchestration problem rather than pure GPU throughput, with one reply noting even a modest 10% gain could matter in production. **Sentiment:** 40% positive · 30% mixed · 30% skeptical **The case for** - A small CPU-side orchestration improvement could matter more than raw benchmark percentages suggest once deployed at production scale. **The pushback** - Agent workloads shifting bottlenecks to CPUs is seen with some wry skepticism about where the real constraints now lie. **By community** - x (mixed): Only a couple of substantive replies, one cautiously optimistic about orchestration gains and one wry about CPUs becoming the new bottleneck. **Highlights** > @rohanpaul_ai the part i like here is that nvidia is treating agent workloads as more than a GPU problem. all those little tool calls and code executions add up fast. if Vera can keep that orchestration moving, 10% on a benchmark could end up mattering quite a bit in production! > — [EdgeCaseBriefs on x](https://x.com/EdgeCaseBriefs/status/2093015580345606547) > @rohanpaul_ai So agents are now making CPUs the bottleneck. Great! 😅 > — [kokowaters\_ on x](https://x.com/kokowaters_/status/2093024467723133406) **Source threads** - [x](https://x.com/rohanpaul_ai/status/2093013526088675608) · 0 points · 2 comments --- Tags: [#azure](https://daily.dev/tags/azure), [#openai](https://daily.dev/tags/openai), [#nvidia](https://daily.dev/tags/nvidia), [#ai-infrastructure](https://daily.dev/tags/ai-infrastructure) [View this post on daily.dev](https://daily.dev/posts/vera-rubin-racks-are-live-at-microsoft-and-the-hype-train-left-the-station-xwy9tybnk) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Vera Rubin racks are live at Microsoft, and the hype train left the station","url":"https://daily.dev/posts/vera-rubin-racks-are-live-at-microsoft-and-the-hype-train-left-the-station-xwy9tybnk","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/vera-rubin-racks-are-live-at-microsoft-and-the-hype-train-left-the-station-xwy9tybnk"},"datePublished":"2026-08-22T11:38:49.178Z","dateModified":"2026-08-28T03:41:41.922Z","description":"Nvidia's next-generation Vera Rubin AI hardware platform has moved from roadmap slide to production, with Microsoft Azure confirming its first Vera Rubin racks...","image":"https://pbs.twimg.com/media/HQUttXDW0AApalv.jpg","thumbnailUrl":"https://pbs.twimg.com/media/HQUttXDW0AApalv.jpg","isAccessibleForFree":true,"articleSection":"Trends","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Trends","logo":"https://media.daily.dev/image/upload/s--ZfSp3asX--/f_auto,q_auto/v1780996004/logos/trends?_a=BAMAMiWQ0","url":"https://daily.dev/sources/trends"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/vera-rubin-racks-are-live-at-microsoft-and-the-hype-train-left-the-station-xwy9tybnk","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":3},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"azure,openai,nvidia,ai-infrastructure","timeRequired":"PT2M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Trends","item":"https://daily.dev/sources/trends"},{"@type":"ListItem","position":3,"name":"Vera Rubin racks are live at Microsoft, and the hype train left the station"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/vera-rubin-racks-are-live-at-microsoft-and-the-hype-train-left-the-station-xwy9tybnk#faq","mainEntity":[{"@type":"Question","name":"Is Nvidia's Vera Rubin AI platform actually in production yet?","acceptedAnswer":{"@type":"Answer","text":"Yes, Microsoft Azure has confirmed its first production Vera Rubin racks are live and running training workloads. CoreWeave had racks running by June and Google by July, following Nvidia's January announcement naming Microsoft, AWS, Google, and CoreWeave as first deployment partners. OpenAI's infrastructure lead also confirmed its training stack is now running on Vera Rubin hardware. Track how Vera Rubin's rollout reshapes AI infrastructure choices as benchmarks emerge on daily.dev."}},{"@type":"Question","name":"Which cloud providers got early access to Nvidia Vera Rubin hardware?","acceptedAnswer":{"@type":"Answer","text":"Nvidia named Microsoft, AWS, Google, and CoreWeave as the first companies to receive Vera Rubin deployments when it announced the platform in January. CoreWeave had systems running by June, Google had racks by July, and Microsoft's production systems went live on Azure afterward, with OpenAI confirmed as a workload running on the new hardware. Compare hyperscaler AI infrastructure bets like this one on daily.dev before picking a training platform."}}]} ```