AMD Targets Over 50% Server Market Share as Helios AI Rack System Gains Major Customer Commitments
Key Takeaways
- •AMD expects its data center business, including AI, to grow well above 100% in 2027 and is targeting more than 50% revenue share of a server CPU market it estimates will reach $220 billion by 2030.
- •The company secured gigawatt-scale AI infrastructure commitments from OpenAI, Meta, and Anthropic, with additional deployments planned through Oracle Cloud Infrastructure and Microsoft Azure.
- •AMD acquired ZT Systems for approximately $4.9 billion to enhance rack-scale design and manufacturing capabilities for its Helios AI rack system, which begins component shipments in September.
- •AMD reported Q2 earnings of $1.66 per share on revenue of $11.54 billion, exceeding consensus estimates and representing a 50.1% year-over-year revenue increase.
- •AMD stated that ROCm software improvements have narrowed the competitive gap with NVIDIA's CUDA to the point where software is no longer a primary concern in discussions with major AI companies.

AMD stock edged lower to $480.80 in premarket trading Thursday, down 0.44%, as investors digested a wave of data center and AI growth announcements from the company's Technology Leadership Forum 2026.
The figures presented at the event are notable. AMD's Corporate VP of Financial Strategy Matt Ramsay said the company's early 2027 server revenue outlook is roughly 20% larger than the entire server market in 2025.
Server revenue grew more than 50% in Q1 and more than 70% in Q2 2026, with both cloud and enterprise revenue rising more than 70% over the same period. AMD now expects server revenue growth to exceed 80% in the second half of 2026.
Looking further ahead, AMD expects its data center business, including AI, to grow well above 100% in 2027. The company raised its estimate for the total server CPU market to $220 billion by 2030 and is targeting more than 50% revenue share — a level that would mark a dramatic shift in the x86 server landscape, where Intel has historically held a dominant position.
Helios AI Rack System and Customer Commitments
AMD plans to begin shipping components for its Helios AI racks to manufacturing partners in September. The package includes MI450 accelerators, Venice CPUs, and Pensando networking products. The rack-scale system is AMD's most direct answer to NVIDIA's GB200 NVL72 and similar integrated platforms that have become the reference architecture for hyperscale AI deployments.
The company has secured 1-gigawatt commitments from OpenAI and Meta. Anthropic has committed to 1 gigawatt, with ambitions to reach 2 gigawatts. Deployments are also planned through Oracle Cloud Infrastructure and Microsoft Azure. These commitments represent some of the largest single-vendor AI infrastructure buildouts disclosed publicly, reflecting the scale at which major AI labs and cloud providers are now procuring compute.
A revenue ramp is expected beginning in Q4, with another increase projected in Q1. Ramsay noted that deployment speed will depend on access to land, power, infrastructure, and capital — constraints that affect the entire industry and have led major cloud providers to pursue dedicated power agreements and site selections tied to grid capacity.
AMD acquired ZT Systems in a deal valued at approximately $4.9 billion to build out its rack-scale design and manufacturing capabilities. Ramsay said ZT Systems helped sharpen the Helios design and reduce execution risk. The Helios ramp will begin with a limited number of manufacturing partners before broadening.
ROCm and Narrowing the NVIDIA Gap
AMD addressed its long-standing software disadvantage against NVIDIA's CUDA ecosystem directly. Ramsay said ROCm improvements over the past 18 months have narrowed the competitive gap to the point where software is no longer a central concern in talks with major AI companies. CUDA has been one of NVIDIA's deepest competitive moats, and the perception that ROCm had closed enough of the gap for large-scale deployment would mark a meaningful change for customers evaluating multi-vendor AI compute strategies.
ROCm AI will launch alongside Helios, with continued updates planned for future accelerator generations.
On the CPU side, Venice scales to 256 cores and 512 threads. AMD plans to follow Venice with its Florence CPU family in 2028, and Ravenna is also in development.
AMD's x86 server market share has grown from 0.4% to the high-40% range. Ramsay said supply, not demand, is the current bottleneck, and AMD is working with TSMC and others to secure sufficient capacity. Advanced packaging and leading-edge node availability at TSMC have been industry-wide constraints affecting multiple chip designers competing for the same finite wafer allocation.
Q2 Earnings and Analyst Outlook
AMD reported Q2 EPS of $1.66, beating the $1.62 consensus. Revenue came in at $11.54 billion, ahead of the $11.31 billion estimate and up 50.1% year over year.
The stock carries a consensus "Moderate Buy" rating and an average analyst price target of $546.87, with KeyCorp setting the highest target at $650.