Intel Claims CPUs Will Match GPUs in AI, Challenging Nvidia's Dominance

During Intel's earnings call on Thursday, the company's leadership made a significant argument about the future of AI infrastructure: as the industry moves beyond model training into production deployment, CPUs will play an increasingly central role alongside GPUs. CFO David Zinsner stated plainly: "The ratio of CPU to GPU — we now believe we're almost in parity at this point, and could eventually even skew more to CPUs on a unit basis." CEO Lip-Bu Tan echoed this perspective, saying "As AI expands from training to inference and increasingly to agentic and multi-agent systems, general purpose server CPU density continues to increase. Our core server CPU franchise is growing faster than ever." The reasoning behind this shift lies in how AI workloads differ across their lifecycle. While training large language models requires the massive parallel processing power that GPUs excel at, running those models in production—a phase called inference—relies on a broader mix of computing resources, including CPUs. Intel backed its claims with concrete results: the company reported $6.3 billion（約1兆円） in second-quarter data center AI revenue, up 59% year over year. Server growth was described as the strongest in Intel's history, with its Xeon 6 processor ranking among its fastest-ramping product launches, and demand continuing to exceed supply. An analyst on the call referenced an industry projection estimating the CPU total addressable market could reach $220 billion（約35兆円） by 2030. While Intel's CFO did not comment on the specific figure, he did not challenge the broader discussion around growing CPU demand. Importantly, Intel's position does not suggest GPUs are becoming less important to the AI ecosystem. Demand for GPU-powered AI training remains strong as companies continue building larger and more capable models. Instead, Intel is making the case that as AI applications scale in production, the infrastructure supporting those workloads will shift to include a larger CPU share relative to GPUs.

For much of the generative AI boom, GPUs have dominated discussions around AI infrastructure and investment, with Nvidia emerging as the biggest beneficiary as companies raced to build large language models. However, Intel's leadership is now arguing that the next phase of AI will look different. The shift reflects a fundamental change in how AI is being deployed: while training remains GPU-intensive, inference—the step where a trained model produces answers at scale—relies on a broader range of computing resources. Intel's CFO David Zinsner stated the company now believes CPU and GPU deployments are nearing parity on a unit basis and could eventually skew more toward CPUs. CEO Lip-Bu Tan reinforced this argument, noting that as AI expands beyond training into inference and increasingly toward agentic and multi-agent systems, general purpose server CPU density continues to increase. Intel's strong second-quarter results—$6.3 billion（約1兆円） in data center AI revenue, up 59% year over year, with Xeon 6 among its fastest-ramping product launches—lend some support to the company's outlook. An industry projection cited during the call estimated the CPU total addressable market could reach $220 billion（約35兆円） by 2030, signaling substantial growth potential in the CPU segment.

No comments yet. Be the first to share your thoughts!