Advancing AI 2026 event and product launches
At its Advancing AI 2026 event in San Francisco, AMD launched the Instinct MI400 Series GPUs and Helios rack-scale systems, with press releases for the products dated July 24, 2026. The MI400 family includes the MI455X for frontier AI and AI factory deployments and the MI430X for sovereign AI and HPC, combining HBM4 memory, advanced security, and scalable system architectures, according to StorageNewsletter coverage published July 27, 2026. The MI430X is described as delivering hardware-based FP64 performance for scientific computing, with the exact teraflops figure truncated in the publicly available source text. The Helios platform pairs MI455X GPUs with sixth-generation EPYC "Venice" server processors, ROCm software, and Pensando networking into a unified rack-scale architecture. AMD put its total addressable market at roughly $2 trillion by 2030, including approximately $1.4 trillion for AI accelerators and $220 billion for server CPUs, framing both estimates as driven by agentic AI workloads. The company also introduced a competitive yardstick, claiming the most AI agents per watt, per dollar, and per rack, with its own endnotes noting that those agent counts are estimated from CPU thread resources used as a proxy.
Zen 7 EPYC roadmap split into three families
AMD confirmed that the Zen 7 generation, scheduled for 2028, will launch as three separate EPYC families rather than a single stack. Tom's Hardware reported on July 27, 2026 that chair and CEO Lisa Su named the three families Florence, Ferrara, and Fidenza during the keynote. Florence will carry fresh Zen 7 cores, a new set of AI compute extensions, and support for newer memory technologies; Ferrara is positioned as the AI host node and is also planned as the CPU inside the Helios 600 rack alongside MI600 Series GPUs and Pensando "Palma" and "Levanzo" networking; Fidenza is described as the agentic sandbox product. AMD disclosed no core counts, process node, or socket details for any of the three families, stating only that Zen 7 uses leading-edge process technology. The approach reverses the single-stack launch AMD used with Zen 5 "Turin," which folded Zen 5 and Zen 5c parts into a single 27-SKU stack on one socket. Sixth-generation EPYC "Venice" launched on the new SP7 socket as the 9006 series, with Venice-X scheduled for 2027 offering 1,152MB of 3D V-Cache, 96 cores, and a 5.15 GHz boost clock. Three named Zen 7 families at announcement, two years out, is a wider spread than AMD has previously opened a generation with.
OpenAI, Anthropic, and Meta deployment commitments
OpenAI plans to deploy AMD's Helios rack-scale AI platform at scale through multiple deployment partners in the second half of 2026, according to DataMagz coverage published July 27, 2026. The deployment forms the first phase of an agreement announced in October 2025 that outlined plans for up to 6 gigawatts of AMD GPU capacity and granted OpenAI an option to acquire up to a 10 percent equity stake in AMD. OpenAI Vice President of Compute Strategy Sachin Katti said the company expects to deploy Helios at massive scale after evaluating a pre-production rack over the preceding three months. DataMagz also reported that AMD separately announced an agreement to support up to 2 gigawatts of AI infrastructure for Anthropic, alongside a strategic investment in that company. Meta, according to Global 1 News coverage dated July 26, 2026, appeared at the keynote to discuss co-designing rackscale OCP standards with AMD and described CPUs as just as important as GPUs for AI infrastructure. The Anthropic 2-gigawatt figure and Meta co-design remarks come from secondary commentary on the keynote and were not independently confirmed by a wire-service report in the source set.
South Korea AI Center of Excellence and heterogeneous computing
AMD and South Korea's Ministry of Science and ICT signed a memorandum of understanding on bilateral cooperation in the AI semiconductor ecosystem at Advancing AI 2026 in San Francisco on July 23, 2026 local time, with coverage in The Korea Times published July 27, 2026 and Seoul Economic Daily the same day. Deputy Prime Minister and Minister of Science and ICT Bae Kyung-hoon and AMD CEO Lisa Su signed the agreement. Under the terms, AMD will prepare to establish an "AI Center of Excellence" in Korea that will serve as a collaboration hub for Korean companies, universities, and research institutes, supporting AI semiconductor software development, computing technology validation, and joint research. The two sides also agreed to pursue development and demonstration of heterogeneous AI computing infrastructure combining AMD CPUs and GPUs with domestically developed Korean neural processing units. Bae Kyung-hoon described the global AI competition as shifting toward heterogeneous computing and open ecosystems, while Su called Korea a key strategic market for AMD with world-class competitiveness in semiconductors, AI, advanced research, and physical AI. Seoul Economic Daily reported that cooperation will also extend into memory, networks, servers, and software, with joint research centered on the National AI Research Center.
Taiwan robotics and physical AI expansion
AMD is deepening its robot and physical AI ambitions through a broader partner network in Taiwan's industrial PC sector, according to DigiTimes reporting published July 27, 2026. The company aims to challenge Nvidia as global developers and system integrators seek edge AI and robotics alternatives. The specific partner identities, contract terms, and roadmap details were not disclosed in the publicly available DigiTimes summary, which sits behind a paid subscription. The Taiwan robotics push forms part of the same set of Advancing AI 2026 announcements that framed AMD's expansion across CPUs, GPUs, rack-scale systems, and AI infrastructure partnerships, and it extends the company's stated TAM into physical AI in addition to agentic AI and frontier model training.
Share this article
