Advanced Micro Devices used the first day of its Advancing AI 2026 conference in San Francisco to announce an agreement with Anthropic that significantly expands the company's position in the AI accelerator market. Anthropic will deploy up to 2 gigawatts of AMD Instinct MI450 Series GPUs in AMD Helios rack-scale solutions, with the first gigawatt beginning deployment in the first half of 2027. AMD committed to a strategic equity investment of up to $5 billion in Anthropic as part of the arrangement, and the two companies will collaborate to use Claude to optimize workloads for AMD Instinct GPUs and accelerate AMD's ROCm software development.
The Anthropic announcement adds to an already substantial demand backlog. AMD had previously disclosed a 6-gigawatt agreement with OpenAI, signed in October 2025, and a separate 6-gigawatt agreement with Meta, announced in February 2026, both centered on MI450-class accelerators. With Anthropic's 2-gigawatt commitment, AMD's total publicly disclosed committed demand now stands at 14 gigawatts. One gigawatt of AI compute capacity translates to approximately 25,000 to 50,000 high-end GPUs, depending on chip generation and power configuration, putting the combined order volume well into the hundreds of thousands of units.
AMD also used the event to formally launch and price the Helios rack-scale system at $5.25 million per unit, confirming Microsoft as the first hyperscaler to deploy Helios for frontier model AI inference. The Helios architecture delivers 31 terabytes of HBM4 memory and is built around the EPYC Venice server CPU, which AMD announced simultaneously as the first x86 server processor produced on TSMC's 2-nanometer process, offering a 1.7x performance improvement over its predecessor. ROCm, AMD's GPU software stack, reached 90 to 95 percent compatibility with major AI frameworks, addressing a longstanding competitive gap with Nvidia's CUDA ecosystem.
The scale of the Anthropic deal drew some initial disappointment on social media, with observers pointing out that 2 gigawatts is significantly smaller than the 6-gigawatt Meta and OpenAI agreements. Analysts who follow AMD closely pushed back on that framing, noting that the MI450 deployment window runs into 2027, meaning the deal represents a commitment to AMD's next product generation rather than just current-quarter revenue, and that all 2026 HBM4 production is reportedly already allocated to hyperscale customers, limiting what AMD could promise for delivery this year regardless of deal size.
Wall Street's bull case on AMD has consistently centered on two pillars: the GPU accelerator franchise anchored by the OpenAI and Meta commitments, and the server CPU business, where AMD now holds approximately 46% of server CPU revenue following a series of market share gains over Intel. KeyBanc maintains a $725 price target and UBS holds a $700 target on AMD, with both firms arguing that the transition from GPU component sales to full rack-scale systems like Helios dramatically improves AMD's revenue per unit and gross margin trajectory over the 2026 to 2028 delivery window.
AMD reports second-quarter 2026 earnings on August 4, with consensus estimates of approximately $11.2 billion in revenue and $1.67 in non-GAAP EPS. Investors will be watching those results for evidence that the multi-gigawatt commitments are beginning to translate into recognized revenue, and for any update on the Helios production timeline as hyperscale customers finalize data center buildout schedules. AMD shares entered the Advancing AI event trading near $553, up roughly 308% over the past twelve months.