TL;DR
- Partnership: Google committed to using multiple generations of Intel’s Xeon 6 processors for AI training and inference in its data centers.
- CPU Demand: Agentic AI workloads are driving CPU demand beyond what GPUs alone can handle, with server processors effectively sold out across the industry.
- Intel Turnaround: Intel’s shares have nearly tripled in a year after investments from the U.S. government and Nvidia, plus new partnerships with Elon Musk’s ventures.
- Competition: AMD hit a record 28.8 percent x86 server CPU market share, while Nvidia and Arm launched competing processors targeting agentic AI workloads.
Google is doubling down on Intel CPUs for its AI data centers, even as it develops its own Arm-based Axion processor and AMD gains record server market share at Intel’s expense. In an expansion of a partnership that dates back nearly three decades, Google committed to using multiple generations of Intel’s Xeon 6 processors for AI training and inference workloads.
Agentic AI workloads are pushing CPU demand beyond what GPUs alone can handle. According to Futurum Group, CPU market growth could exceed GPU growth by 2028, and high-core-count server processors are effectively sold out across the industry. Agentic AI and reinforcement learning workloads are pushing CPU-to-GPU ratios in AI clusters back toward 1:1, creating what the research firm calls a quiet supply crisis.
Intel shares gained 2% on the news while Alphabet shares fell more than 1%, a divergence suggesting the market views the commitment as costlier for Google than beneficial for its bottom line. No financial terms or timeline were disclosed.
Google and Intel Deepen AI Infrastructure Collaboration
A multiyear collaboration covers performance, energy efficiency, and total cost of ownership across Google Cloud’s infrastructure. Google Cloud already deploys Intel Xeon 6 processors across its C4 and N4 compute instances, and the expanded commitment extends to future chip generations. Google has relied on Intel processors dating back to its earliest server rack ambitions nearly three decades ago, making Intel one of its longest-standing silicon partners.
Beyond CPUs, both companies are expanding co-development of custom ASIC-based IPUs, programmable accelerators that offload networking, storage, and security functions from host processors. According to Intel, Google and Intel have collaborated on the infrastructure processing unit since 2022, when Google described it as a first-of-its-kind chip. IPU work goes beyond standard procurement, with both companies jointly developing silicon tailored for cloud-scale networking demands.
Intel CEO Lip-Bu Tan framed the expanded partnership as evidence that AI infrastructure demands balanced systems beyond just GPU accelerators, with CPUs and IPUs central to delivering the performance modern workloads require.
“CPUs and infrastructure acceleration remain a cornerstone of AI systems — from training orchestration to inference and deployment. Intel has been a trusted partner for nearly two decades, and their Xeon roadmap gives us confidence that we can continue to meet the growing performance and efficiency demands of our workloads.”
Amin Vahdat, SVP & Chief Technologist, AI Infrastructure, Google
Multi-generational CPU procurement combined with joint IPU development gives Intel both revenue stability and a co-design relationship with one of the world’s largest cloud operators. For Google, locking in Intel access provides a hedge against the very CPU supply constraints driving up prices across the industry.
Intel’s Comeback Draws Major Partners
Intel has struggled for years to keep pace with industry trends, but its trajectory has reversed under Tan’s leadership. Shares have nearly tripled in the past year as major partners and governments have rallied behind the chipmaker. In August 2025, the U.S. government bought a 10% stake in Intel, touting the chipmaker’s ability to manufacture advanced chips on American soil.
Nvidia followed with a $5 billion stake a month later, further cementing industry confidence in Intel’s direction. Elon Musk has also tapped Intel to design, fabricate, and package custom chips for SpaceX, xAI, and Tesla at his Terafab project in Texas. Each new high-profile backer validates Intel’s viability, making the next partnership easier to secure.
Intel recently announced repurchasing 49% of its Fab 34 chip facility in Ireland, paying $14.2 billion according to CNBC. Intel had sold the stake to Apollo Global Management in 2024 for $11.2 billion, and the buyback at a $3 billion premium reflects the chipmaker’s improved financial position. Intel CFO David Zinsner pointed to the company’s stronger balance sheet and evolved business strategy as evidence of the turnaround’s staying power.
Intel also builds chips on its most advanced 18A process at its Arizona fab, though Intel’s own processors remain the largest customer at the plant and no major external foundry customer has been announced. Intel’s manufacturing ambitions still outpace its execution: its Arizona fab remains primarily a captive operation, and the foundry business meant to attract external customers has yet to land a marquee deal.
Google’s commitment is particularly notable given its 2024 launch of the Arm-based Axion CPU, a direct alternative to Intel’s x86 architecture. Google has already migrated 30% of its internal applications to Axion, yet is simultaneously deepening its Intel dependency, a dual-sourcing strategy that reflects the severity of current CPU supply constraints. Tan has been candid about Intel’s competitive shortcomings, acknowledging that “moving away from SMT put us at a competitive disadvantage” and pledging to reintroduce simultaneous multithreading in future Coral Rapids processors.
CPU Bottleneck in the Agentic AI Era
Broader industry recognition that GPUs alone cannot handle the next wave of AI workloads underpins the partnership. Dion Harris, Nvidia’s head of AI infrastructure, told CNBC in March that CPUs are becoming the bottleneck as agentic systems shift compute demands. Agentic AI systems that orchestrate multiple model calls, manage tool use, and maintain persistent state impose heavier loads on CPUs than traditional single-model inference pipelines, reversing the GPU-centric trend that has defined AI infrastructure investment for the past four years.
According to Mercury Research, AMD’s x86 server CPU shipments hit a record 28.8 percent in Q4 2025, and AMD claims its EPYC Turin processors hold a significant performance lead over Xeon 6. Intel’s upcoming Diamond Rapids processors are expected to lack simultaneous multithreading, a decision that AMD has seized on as a competitive advantage. Nvidia unveiled its Vera CPU and Arm introduced its AGI CPU, both targeting the same agentic AI workloads driving demand upward. Once a two-horse race between Intel and AMD, the server CPU market now faces new entrants from both ends of the compute stack.
Intel occupies a paradox: its CPU business is more relevant to AI infrastructure than at any point in the GPU era, yet the number of credible competitors has seldom been higher. Google’s deal provides a high-profile reference customer and revenue anchor as Intel defends its dominant unit share against AMD’s rising revenue share and looming Arm-based alternatives from both its customers and competitors.
“Scaling AI requires more than accelerators — it requires balanced systems,” Tan stated. Google’s multi-generational commitment substantiates that argument, but it also underscores how much Intel’s turnaround depends on execution.
Beyond its immediate commercial terms, the deal carries strategic weight. For Intel, it converts a narrative of decline into one of strategic relevance: its CPU business is now central to AI infrastructure at the very moment hyperscalers are scrambling for supply. For Google, the arrangement provides procurement stability as lead times for entry-level and midrange Xeon processors extend to as much as six months for new orders. With CPU demand surging, supply constrained, and new entrants flooding into server processors, the partnership gives Intel the backing of one of the industry’s largest infrastructure buyers, and raises the stakes if the chipmaker cannot deliver on its manufacturing promises.