uploads/2017/02/A10_Semiconductors_NVIDIA_4Q17-Data-Center-Revenue-1.png

Nvidia’s Data Center and the AI Revolution

By

Updated

Nvidia’s Data Center segment

Nvidia (NVDA) is looking to transform its GPUs (graphics processing units) into general purpose processors and use them in new areas like data center and automotive. This effort has made Nvidia a key beneficiary in the data center industry’s shift to AI (artificial intelligence) and deep learning.

More and more, companies are using GPUs and alternative processors for their AI tasks. This transformation has helped Nvidia’s data center revenue grow faster than the industry.

Article continues below advertisement

Data Center revenue trend

As seen in the chart, Nvidia’s data center business reported a sequential revenue growth of 59% in fiscal 3Q17, whereas rival Intel’s (INTC) data center business reported a growth of only 11.7%. This is because GPUs are reducing the need for more server processors by accelerating the processing speed.

The segment is likely to report double-digit percentage growth in fiscal 4Q17 as well, driven by sales of its newly launched TensorRT deep learning framework. The segment is likely to report exponential growth in coming years, as more companies adopt AI and deep learning.

Volta GPUs in supercomputer

In fiscal 2018, Nvidia plans to supply its next-generation Volta GPUs to the US Department of Energy’s supercomputers, Summit and Sierra. While these supercomputer’s host GPU would be IBM’s (IBM) Power9 CPU (central processing unit), Volta is expected to power 90% of their floating point performance.

Tesla P100 in other servers

In fiscal 2018, Nvidia will also see increasing orders for its Tesla P100 as several server makers are expected to launch servers featuring this GPU. System76 recently launched its rack server Ibex Pro, which features Intel’s 22-core Xeon E5 v4 chips and eight Nvidia Tesla P100 GPUs. The Ibex Pro will likely compete with Nvidia’s DGX-1 as the two target the same market.

While both the servers house eight Tesla P100 GPUs, Ibex Pro has newer CPUs than DGX-1. However, Ibex Pro does not have the advantage of Nvidia’s superfast NVLink interconnect, which is reserved only for DGX-1 and Power8 servers.

The proven success of GPUs in AI has encouraged Advanced Micro Devices (AMD) and Intel to enter this market with their Radeon Instinct GPUs and Xeon Phi processors, respectively. But Nvidia has a technology advantage over these companies, which should help it maintain its share in the AI market.

Another segment in which Nvidia is witnessing strong growth is automotive. We’ll take a closer look at this in the next part of the series.

Advertisement

More From Market Realist