AMD’s EPYC server processors and data center graphics processing units (GPUs) are going to play a critical role in the adoption of AI applications. That’s because the deep neural networks that power AI applications, such as self-driving cars or other real-time applications, should be able to carry out hundreds of thousands, millions, or even billions of operations in order to recognize various objects and react to them properly.
AMD’s chips help data centers and supercomputers tackle the massive workload that’s required to enable AI applications. The chipmaker recently revealed the Instinct MI200 series of data center accelerators that are based on the CDNA 2 architecture, claiming that these chips are 4.9 times faster while carrying out high-performance computing (HPC) and AI operations, as compared to competing data center accelerators.
Similarly, AMD’s third-generation EPYC server processors have improved AI inferencing capabilities over their predecessors. Not surprisingly, AMD has been witnessing strong growth in the adoption of its server processors and GPUs. Meta Platforms, for instance, recently selected AMD’s EPYC processors for use in its hyperscale data centers that would be used to power up the former’s metaverse – a virtual environment where people can interact with each other like they do in the real world.
This win could be a big deal for AMD as Meta is expected to invest $10 billion to boost its metaverse capabilities this year. That figure could head higher in the coming years as Meta CEO Mark Zuckerberg had pointed out on the October earnings conference call.
More importantly, AI applications, such as the metaverse and self-driving cars, are expected to increase the deployment of HPC data centers, thanks to their ability to address data-intensive workloads. According to SK Hynix, hyperscale data center deployments are expected to hit 1,060 by 2025, which would be double the current installed base.
Not surprisingly, the demand for data center accelerators can shoot substantially higher in the future. A third-party estimate points out that sales of data center accelerators could hit $53 billion in 2027, as compared to $4.2 billion last year, driven by strong sales of both CPUs (central processing units) and GPUs. This should pave the way for strong growth at AMD and help the chipmaker remain a top growth stock for a long time to come, especially considering that it has other catalysts apart from AI that could boost its top and bottom lines in the long run.