On Aug. 26, Intel published the full specifications of its next-generation Xeon 7 server processor, code-named Diamond Rapids, a chip with up to 256 performance cores, 16 compute tiles, and 1.28 gigabytes of last-level cache, built on the company’s 18A-P manufacturing process. The specifications, which Intel said deliver a 30 percent frequency gain at 0.5 volts, represent the company’s most aggressive answer yet to AMD’s advance in the server market.
The numbers are designed to change the conversation. Intel has spent the past several years losing server market share to AMD, whose EPYC processors have won on core counts, power efficiency, and price. Diamond Rapids does not merely match the competition; it exceeds it on the headline metrics, with more cores, more cache, and a process node that Intel says will be the industry’s most advanced when the chip ships. The specifications extend the roadmap Intel outlined at the Hot Chips conference, where the company laid out its vision for “agentic AI” workloads, servers built for software agents that reason, plan, and act autonomously.
The chip’s architecture is a bet on chiplets. Diamond Rapids is assembled from 16 compute tiles, each containing 16 of the new performance cores, connected by the company’s embedded interconnect. The tile-based design lets Intel scale the chip up or down for different market segments and, crucially, mix tiles built on different processes, a flexibility that AMD pioneered and that Intel has now fully adopted. The 1.28GB of last-level cache, the largest in any x86 server processor, is aimed at the data-heavy workloads that AI agents generate.
The 18A-P process is the heart of the claim. Intel has staked its turnaround on its ability to catch and pass TSMC in manufacturing, and 18A is the node where it intends to do so. The “-P” variant is tuned for performance, and Intel’s assertion of a 30 percent frequency gain at low voltage is aimed directly at the data center market, where power consumption is often a larger cost than the servers themselves. If the process delivers as claimed, Diamond Rapids would give Intel a cost and performance advantage it has not enjoyed in a decade.
The timing aligns with a shift in the server market. The center of gravity in data center spending has moved from general-purpose computing to AI accelerators, and the role of the server CPU has changed with it. The CPU in an AI server coordinates the accelerators, feeds them data, and runs the software stack around them; its performance matters more than in the past, and the market for AI servers is growing faster than the market for traditional servers. Intel’s positioning of Diamond Rapids as an “agentic AI” processor is an attempt to claim that growth.
AMD will not concede the segment without a fight. The company has its own next-generation server platform in development, built on its own process roadmap and its own tile architecture, and it has spent years building software and customer relationships that Intel cannot match overnight. The server market is famously sticky, with customers reluctant to change platforms, and AMD’s share gains have come from persistent execution rather than a single product. Diamond Rapids gives Intel the specifications; winning back customers will take years of deliveries.
Intel’s own recovery is still incomplete. The company has cut costs, reorganized its manufacturing business, and found new customers for its foundry services, but its financial results remain far below its peak, and its position in the AI accelerator market, where Nvidia dominates, is marginal. Diamond Rapids is important to Intel not because it will reverse the accelerator deficit but because it defends the server CPU franchise that still generates much of the company’s profit.
The broader implication is for the x86 ecosystem itself. Intel and AMD have competed for decades, and their rivalry has produced steady gains in performance and efficiency. The arrival of ARM-based server chips, led by Amazon’s Graviton and other designs, adds a third front, and the pressure on both x86 companies to innovate has never been higher. Diamond Rapids is Intel’s statement that x86 remains competitive at the top of the server market.
Customers will judge the chip on delivery, not specifications. Intel has been criticized in the past for announcing aggressive specs and shipping late or below target, and its credibility in the server market depends on hitting the Diamond Rapids schedule. The company has said volume production will begin in the second half of 2027, a timeline that gives it roughly a year to prove the process and the architecture in production.
For now, the specifications have done their job: they put Intel back in the conversation. The company’s data center group has spent years on the defensive, answering questions about delays and market share. With Diamond Rapids, Intel has published numbers that AMD must now match or beat, and it has done so on a process it controls from design to fabrication. Whether the chip delivers on the promise is a question for 2027; whether it changed the terms of the debate is already answered.
