U.S. semiconductor giant AMD (Advanced Micro Devices) announced on July 20 a major expansion of its long-term strategic partnership with Microsoft. At the core of this expanded collaboration is the large-scale deployment of AMD’s next-generation integrated rack-scale AI system, “AMD Helios,” on Microsoft’s cloud platform, Microsoft Azure. This is expected to dramatically enhance the inference processing capabilities for cutting-edge AI models and various AI services deployed on Azure.<\/p>\n
The partnership deploys AMD’s comprehensive AI solutions\u2014spanning GPUs, CPUs, networking, and software\u2014on Azure. Microsoft CEO Satya Nadella noted, “Customers require AI infrastructure optimized for a broad range of workloads, from training and inference to data preparation, search, and reinforcement learning.” He added, “Through our collaboration with AMD, we are expanding Azure’s infrastructure portfolio with AMD Helios to deliver the performance, scalability, and choice our customers need.”<\/p>\n
Helios Ushers in a New Phase of AI Infrastructure<\/p>\n
AMD Helios is not a standalone chip but a rack-scale AI system housing 72 GPUs. It integrates the latest AI-focused GPU, the “AMD Instinct MI455X”; the 6th-generation server CPU “AMD EPYC” (codenamed Venice), manufactured on Taiwan Semiconductor Manufacturing Company’s (TSMC) 2-nanometer (nm) process; a DPU (Data Processing Unit) for network control leveraging technology from Pensando, which AMD acquired in 2022; and the open-source GPU programming foundation “ROCm.”<\/p>\n
Commenting on the announcement, AMD CEO Lisa Su stated, “AMD and Microsoft have worked together for years to build high-performance infrastructure. Microsoft’s new adoption of AMD technology marks an important milestone in delivering cutting-edge computing solutions to Azure customers and jointly scaling the next generation of AI infrastructure.”<\/p>\n
Microsoft will utilize Helios for inference processing of its own and customers’ most advanced AI models, Azure AI services, and customer applications. AMD plans to begin shipping Helios to customers, including Microsoft, in the second half of 2026. Meta, OpenAI, Oracle, and SpaceXAI (formerly xAI) have already publicly announced their adoption of AMD’s Instinct series.<\/p>\n
Three New VM Series Launch on Azure<\/p>\n
Alongside this partnership expansion, Microsoft announced three new virtual machine (VM) series on Azure powered by 6th-generation AMD EPYC processors.<\/p>\n
Designed for AI agents and data pipeline processing, “Azure HDv2” features approximately 500 CPU cores, 4 terabytes (TB) of RAM, 32TB of local NVMe storage, and 400Gb Azure Boost networking, engineered to handle the most demanding AI workloads.<\/p>\n
“Azure HXv2,” specialized for advanced engineering workloads such as semiconductor design, packs 176 CPU cores operating at clock speeds exceeding 5 gigahertz (GHz), with a 50% increase in addressable cache per core compared to the previous generation. With 2TB or 4TB RAM options, customers can optimize based on their memory requirements.<\/p>\n
Finally, the “ND MI455X v7,” which directly leverages Helios capabilities, is designed for inference, search, and agent-based workloads, expanding Azure’s infrastructure options for large-scale inference.<\/p>\n
Networking Layer and Software Evolution<\/p>\n
On the networking front, the deployment scope of Pensando DPUs on Azure is being expanded. Integration between Microsoft’s cloud architecture “Azure Boost” and AMD’s semiconductor technology will also advance, offloading virtualization, networking, and storage tasks from host servers to dedicated hardware. This frees up CPU capacity for customer workloads and enhances connectivity processing performance and energy efficiency at cloud scale.<\/p>\n
On the software side, the open-source ROCm platform serves as the core of Helios. Neil Shah, an analyst at market research firm Counterpoint Research, assessed AMD’s hardware performance as “on par with NVIDIA’s GPUs and CPUs,” but pointed out that “the key lies in software and optimization.” He indicated that the gap with the CUDA ecosystem NVIDIA has built over many years remains a critical challenge for AMD.<\/p>\n
Shifting Dynamics in the AI Semiconductor Market?<\/p>\n
This partnership symbolizes AMD’s challenge to NVIDIA (NVDA), which holds an overwhelming share of the AI semiconductor market. Currently, NVIDIA commands over 95% of the data center GPU market, while AMD holds approximately 4.5%. However, Daniel Newman, an analyst at market research firm Futurum Group, points to the possibility of AMD expanding its share to 20-25%, analyzing that “this would mean hundreds of billions of dollars in revenue.”<\/p>\n
AMD’s data center business is performing strongly. Data center segment revenue in the first quarter of 2026 grew 57% year-over-year. The company expects its data center AI business to generate tens of billions of dollars in revenue from 2027 onward, with Helios driving the majority of that growth.<\/p>\n
On the pricing front, Futurum Group estimates Helios’s selling price at $5 million to $5.5 million (approximately \u00a5900 million), exceeding the estimated $3.5 million to $4 million (approximately \u00a5700 million) price tag for NVIDIA’s second-generation rack-scale system, “Vera Rubin.” Forrest Norrod, head of AMD’s data center business, explained Helios’s competitive edge as “delivering the best total cost of ownership (TCO) and the lowest cost per token.”<\/p>\n
Meanwhile, Newman poses a core question: “Is AMD winning because of technological superiority, or simply because compute capacity supply is tight and everything they make sells?” The actual performance of Helios in its initial deployments will likely determine the answer.<\/p>\n
Stock Price and Market Reaction<\/p>\n
AMD’s stock rose following the announcement, ahead of its quarterly earnings release scheduled for August 4. In trading on July 21, AMD shares closed 1.58% higher at $503.57, briefly reaching $524.28 in after-hours trading. Strong market expectations for the company’s earnings are underpinning the stock price.<\/p>\n
On technical indicators, the Relative Strength Index (RSI) stands at 47.60, below its moving average of 53.39, suggesting near-term buying momentum is mixed. Analysts note that maintaining the $500 support level is crucial. Sustained trading above $522 could set the stage for testing recent highs around $540.<\/p>\n
AMD’s customer base is steadily expanding. Meta has adopted AMD GPUs for inference on its Llama series, planning deployments of up to 6 gigawatts. OpenAI is similarly pursuing large-scale adoption, and India’s Tata Consultancy Services has also made significant deployment decisions. According to AMD, eight of the world’s top ten AI companies are already running workloads on Instinct GPUs.<\/p>\n
For Microsoft, this partnership is also a strategic move to diversify its AI compute capacity sourcing. The company already deploys NVIDIA GPUs alongside its in-house developed AI accelerator “Maia.” With the addition of AMD Helios, Microsoft is poised to complete a broad infrastructure portfolio capable of meeting diverse customer needs.<\/p>\n","protected":false},"excerpt":{"rendered":"U.S. semiconductor giant AMD (Advanced Micro Devices) announced on July 20 a major expansion of its long-term strategic…\n","protected":false},"author":2,"featured_media":114950,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[11],"tags":[3125,57518,420,7829,58681,28342,58680,320,7828,58,57650,31792,2806],"class_list":["post-114949","post","type-post","status-publish","format-standard","has-post-thumbnail","category-microsoft","tag-amd","tag-amd-helios","tag-azure","tag-azure-ai","tag-epyc-venice-cpu","tag-lisa-su","tag-mi455x-gpu","tag-microsoft","tag-microsoft-ai","tag-nvidia","tag-pensando-dpu","tag-rocm","tag-satya-nadella"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/114949","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/comments?post=114949"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/114949\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/114950"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=114949"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=114949"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=114949"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}