Nvidia Looks to Tap AI Opportunities in Fiscal 2018
Nvidia expands its Data Center segment
Nvidia (NVDA) has expanded its offerings in its Gaming platform, providing a boost to its revenues. However, what drove Nvidia’s stock to new highs was its breakthrough in the Data Center segment.
Nvidia converted its Pascal GPUs into general purpose cards that can boost a server’s processing speed and support the deep learning efforts of cloud companies. This helped the company expand its data center offerings to cloud, enterprise, HPC (high-performance computing), and GRID graphics virtualization. The company also launched DGX-1, the world’s first AI supercomputer, which was well-received in the HPC market.
Interested in NVDA? Don't miss the next report.
Receive e-mail alerts for new research on NVDA
Data Center revenues
Nvidia’s (NVDA) Data Center segment’s revenues rose 145% YoY (year-over-year) to $830 million in fiscal 2017. The segment’s revenue growth has accelerated with every passing quarter, growing 63%, 110%, 193%, and 205% YoY in fiscal 1Q17, 2Q17, 3Q17, and 4Q17, respectively. Strong demand in the education, automotive, and energy markets doubled Nvidia’s GRID graphics virtualization revenues.
Overall growth in NVDA’s Data Center segment was driven by the increasing adoption of GPUs by cloud service providers like Microsoft (MSFT), Facebook (FB), and Google (GOOG) for their deep learning initiatives. The next growth spurt would come with the adoption of GPUs by enterprises in the fields of healthcare, transportation, finance, and retail.
Goldman Sachs analyst Toshiya Hari stated that Nvidia is already a leader in deep learning, owning an ~90% share in the deep learning chip market. Intel (INTC) is trying to enter the AI market, but it is behind Nvidia in terms of technology.
What would drive growth in fiscal 2018 and beyond?
Nvidia’s DGX-1 has been well-received by tech companies and governments worldwide, including the following applications:
- Fujitsu would use 24 DGX-1s for its RIKEN Center for Advanced Intelligence Project’s system, which is scheduled to come online in April 2017. The system would have a theoretical performance of over 4 petaflops.
- The Tokyo Institute of Technology would use Nvidia’s Tesla P100 GPUs to power new AI computer TSUBAME3.0, which is expected to deliver a performance of 47 petaflops.
- Nvidia has also unveiled its new supercomputer DGX SATURN V, which would be used in the CANDLE (Cancer Distributed Learning Environment) project. DGX SATURN V would feature 124 DGX-1 systems and would be the most energy-efficient and the 28th-fastest supercomputer in the world.
Nvidia is looking to accelerate its Data Center segment’s growth in fiscal 2018 by rolling out AI software-as-a-service, which is currently in its nascent stage. The company’s DGX-1 AI supercomputer is powering Microsoft’s GPU-accelerated Microsoft Cognitive Toolkit available on the cloud and on site.
Nvidia has also undertaken several initiatives to boost AI adoption. We’ll look into this in the next part.