Nvidia (NVDA) signed an expanded multi-year data center agreement with Meta (META) that will see the chipmaker supply millions of Blackwell and Rubin GPUs to the social media giant.
While that was certainly the most high-profile part of Tuesday’s news, the companies said that under the deal, Meta will also roll out Nvidia Grace CPU-only servers in its data centers, marking the first large-scale deployment of the chip.
Grace is Nvidia’s processor paired with two Blackwell or two Blackwell Ultra GPUs to form the GB200 and GB300 AI superchips.
Only Grace’s servers come as Nvidia seeks to capitalize on growing demand for traditional CPUs as hyperscale enterprises increasingly look to these chips to help power some artificial intelligence inference and agent artificial intelligence applications.
That spells trouble for Intel (INTC), which has long dominated the data center CPU space, and AMD (AMD), which is struggling to take market share from Intel.
“Nvidia has been working on delivering more content in the data center for some time,” Gil Luria, managing director and head of technology at DA Davidson, told Yahoo Finance.
“The addition of Mellanox [a networking company Nvidia acquired in 2020] “So when they sell to data centers, they’re actually selling almost the vast majority of the value,” he said. But it makes sense for them to increase that value further by increasing CPU capacity. “
Nvidia’s move comes at a good time for Intel, which is dealing with capacity constraints that prevent it from producing enough CPUs to meet the needs of data center builders.
It’s not just data centers, though. Nvidia is also reportedly pushing into Intel and AMD’s consumer business with its own laptop chips, creating a whole new set of troubles for the PC giants.
Nvidia’s move to sell CPUs doesn’t mean it’s giving up its huge advantage in the GPU market. This is not a sign that the AI GPU market is at its end. Instead, it’s taking advantage of the AI industry’s growing trend to use CPUs to power smaller AI models.
Giant AI models like the latest and greatest cutting-edge models from OpenAI (OPAI.PVT), Google (GOOG, GOOGL) and Anthropic (ANTH.PVT) still require the kind of horsepower that only GPUs can provide. But CPUs are taking back some of the limelight for those more petite models.
CPUs are also a bottleneck in the AI supply chain, one of many bottlenecks in the ongoing AI buildout that could hurt Nvidia’s sales over time. Luria said that by launching its own CPUs, Nvidia is doing everything it can to keep sales flowing.
