AMD Raises Long-Term Targets Again as Anthropic Deal and Data Center Doubling Signal AI Cycle Is Still Accelerating
Q2 2026 earnings call, August 4, 2026
Advanced Micro Devices delivered a quarter that forced management to tear up its own long-term model just nine months after setting it, with CEO Lisa Su telling investors the company is now "tracking materially ahead of the long-term financial model we shared at our Financial Analyst Day last November." Revenue hit a record $11.5 billion, up 50% year-over-year, with data center now accounting for 58% of total sales versus 42% a year ago. The bigger signal for investors was not the beat itself but the scale of the upward revision to AMD's addressable market and the new customer commitments that back it up.
Anthropic Joins OpenAI and Meta as a Gigawatt-Scale Anchor
The most consequential new disclosure was a strategic partnership with Anthropic, under which Anthropic will deploy up to 2 gigawatts of MI450-series GPUs using AMD's Helios rack-scale platform, with the first gigawatt going live in the first half of 2027. The arrangement goes beyond a hardware purchase agreement: it includes a multiyear joint engineering collaboration in which Anthropic's Claude models will be used to optimize workloads for Instinct GPUs and to accelerate development of AMD's ROCm software stack. AMD also expanded its existing Microsoft relationship, with Microsoft committing to deploy Helios at scale on Azure for frontier model inferencing across its own AI customers and Azure AI services. Su said these additions, layered onto existing multi-generation commitments from OpenAI and Meta, "broaden the group of leading AI companies and cloud providers building their next-generation infrastructure on AMD."
On the sizing of these deals, Su reiterated that AMD's revenue-per-gigawatt figure remains in the "double-digit billions" range, pushing back only mildly on Bank of America's Vivek Arya, who framed the current visible commitments at roughly 3 gigawatts for 2027. CFO Jean Hu and Su both signaled that upside exists depending on how quickly customers can bring data center capacity online, but neither offered a revised gigawatt number, leaving the market to work through the math itself.
TAM Estimates Move Sharply Higher
AMD now expects the data center AI accelerator market to grow more than 45% annually to approximately $1.4 trillion by 2030, and the server CPU market to grow more than 50% annually to approximately $220 billion by 2030. Combined with client and embedded markets, Su put the total addressable high-performance and AI computing opportunity at roughly $2 trillion by 2030, growing about 40% annually. Su was explicit that the market itself has moved, not just AMD's execution: "the overall data center market opportunity is expanding far more rapidly than we projected just 6 months ago." That reframing is the basis for AMD's decision to abandon its November 2025 target of "greater than 35%" revenue growth and its $20 long-term EPS target, both of which management said will be "significantly" exceeded.
2027 Data Center Guidance: More Than Doubling
For 2027, AMD guided server CPU revenue to grow more than 70% year-over-year off a much higher base, while the data center AI business is expected to be, in Su's words after prodding from Bernstein's Stacy Rasgon and Cantor's C.J. Muse, "well over 100%" rather than a simple doubling. Combined, AMD expects the entire data center segment to more than double in 2027. Su indicated this acceleration is being driven by Helios shipments beginning late in the third quarter of 2026, stepping up materially in the fourth quarter, and ramping further into 2027 as the Anthropic, OpenAI, and Meta deployments scale in parallel with broader hyperscaler and enterprise CPU demand.
Second Quarter Data Center Performance
Data center segment revenue reached a record $6.7 billion, up 107% year-over-year and 16% sequentially, with data center operating margin at 31%. Server CPU revenue posted its fifth consecutive quarter of records, with both cloud and enterprise sales growing more than 70% year-over-year, exceeding the guidance AMD had given the prior quarter. Instinct GPU sales more than doubled year-over-year on continued MI355X adoption. Su noted that fifth-generation EPYC Turin now powers nearly a third of the more than 1,600 EPYC public cloud instance types available globally, with cloud customers adding "tens of millions of instances" last quarter across healthcare, financial services, media, and technology.
Helios Demand Running Ahead of AMD's Own Forecast
Su told TD Cowen's Joshua Buchalter that customer pull for Helios, the rack-scale platform combining Venice CPUs, MI450-series GPUs, Pensando networking and ROCm software, is "tracking ahead of our initial forecast" on volume, driven by strong reception among customers who have spent time with the product and AMD's software ecosystem. Across a range of inferencing workloads, AMD claims Helios delivers up to 15% more throughput at equivalent rack power and up to 30% more tokens per dollar than competing systems. Initial shipments are set to begin later in the third quarter and ramp through the fourth quarter into 2027. Management expects yields to improve steadily as the ramp progresses, calling Helios a "highly complex" product typical of first-generation ramps.
Looking further out, AMD plans to launch a new rack-scale AI platform annually, with the 2027 generation pairing MI500-series GPUs with Verano CPUs and expanded scale-up interconnects in both copper and optical. Su said customer engagement on MI500 is "very strong," with the platform expected to deliver the largest generational leap in Instinct history, putting AMD on track for a greater than 2,000-times increase in inferencing performance over four years.
Server CPU Supply Tightness Easing Into 2027
Morgan Stanley's Joe Moore pressed on whether AMD remains supply constrained in server CPUs. Su confirmed the supply chain has been tight through the first half of 2026 because "much of this demand was unforecasted," but said 2027 demand is now better forecast and supply should improve, with room for growth to come in above the greater-than-70% guide "depending on how things play out." She credited AMD's chiplet architecture with easing the transition to 2-nanometer manufacturing, since the design requires fewer wafers on the new node relative to a monolithic approach, even as new-node ramps remain inherently difficult.
Gross Margin Puts and Takes for 2027
CFO Jean Hu fielded repeated questions on how the mix shift toward GPUs, which carry gross margins "slightly below corporate average," will interact with a server CPU business that is now accretive to overall margins. Hu described 2027 gross margin as dependent on the relative pace of the two ramps, but pointed to embedded segment recovery as an additional tailwind, saying she feels "really good" about the company's ability to manage the transition. Second quarter gross margin came in at 56%, up 200 basis points year-over-year, and AMD guided third-quarter gross margin to remain at approximately 56%.
Client and Gaming: Soft Guide on Component Costs
Client and Gaming segment revenue grew 6% year-over-year to $3.8 billion, with client revenue up 23% to $3.1 billion on record mobile processor sales and Ryzen Pro commercial revenue growing more than 50% year-over-year. Gaming revenue fell 31% year-over-year to $779 million on weaker semi-custom console sales and higher component costs pressuring graphics card demand. Management is now planning for "a softer PC market" in the second half as memory and component cost inflation weighs on demand, though Su said AMD's client business should outperform the broader market and still grow year-over-year in 2026, aided by the new Ryzen AI Halo platform featuring the Gorgon Halo processor with 192 gigabytes of unified memory capable of running models up to 300 billion parameters.
Embedded Segment Returns to Strong Growth
Embedded revenue increased 19% year-over-year to $977 million, its strongest growth rate in more than three years, with operating margin at 40%. AMD highlighted growing adoption of embedded x86 processors by hyperscalers and networking customers for control-plane functions inside data centers, alongside new products for real-time edge AI and physical AI robotics. Design win momentum remains strong, with AMD tracking toward more than $18 billion in new design wins for the year.
Third Quarter Guidance
AMD guided third-quarter revenue to approximately $13 billion, plus or minus $300 million, implying 41% year-over-year growth and 13% sequential growth, driven by strong double-digit growth in both data center and embedded segments, partially offset by a decline in gaming. Non-GAAP gross margin is expected to hold at approximately 56%, with operating expenses of roughly $3.65 billion and a 13% effective tax rate.