About 13 years ago, Nvidia made a pair of bets on its long-term future as a company. The company launched two major initiatives, branded as Tegra and Tesla at the time, in the hopes of creating new markets for itself. This was partly a response to the decline of its chipset business, which left it active in fewer markets and more dependent on companies like AMD and Intel for a path to consumers.
Tegra boomed in the early dual-core smartphone era, then transformed slowly into Nvidia’s automotive and embedded machine learning initiatives today. These projects contribute positively to Nvidia’s bottom line but their contributions are dwarfed by Nvidia’s consumer GPU business.
Tesla — now known as Ampere, because the “Tesla” brand was retired in May due to potential conflict with Tesla Motors — has been different. Nvidia has built its data center business from scratch and invented an entirely new language called CUDA to do it. The company’s work in this context is helping to drive the growth of AI and associated fields like computer vision.
My work with upscaling video over the past eight months has convinced me tech like this will be an increasingly important part of content processing in the future. Long-term, we may well see the feature baked into AMD and Nvidia GPUs as a real-time, user-controllable setting or capability, and the impact on older content is incredible. One reason I suspect we’ll see this kind of shift is because it gives AMD and Nvidia the chance to market their GPUs as improving older content you might want to watch or stream as opposed to being dependent on studios to reissue materials in 4K, 8K, or beyond.
Nvidia’s data center business doesn’t necessarily have much to do with AI upscaling — it’s more focused on markets in everything from oil and gas exploration to providing cloud instances for GPU virtualization, to powering the work done to analyze diseases like COVID-19 — but the pandemic drove a surge in data base deployments this quarter. Nvidia’s acquisition of Mellanox closed this quarter, which bumped the company’s data center revenue up, but it would have been a record quarter no matter what.
(Note: This is Q2 2021 in Nvidia’s fiscal accounting year. Companies often align their fiscal quarters with the calendar year but are not required to do so.)
Overall revenue was up to $3.866B, up 26 percent quarter-on-quarter and 50 percent year on year. Nvidia booked $622M in net income for the quarter.
Revenue growth trends show how the pandemic has impacted Nvidia’s disparate businesses. Gaming revenue has grown (as you’d expect), but the initial boom in professional visualization driven by the coronavirus pandemic has dried up. The big data center jump is partly Mellanox and partly new deployment cycles for AI and machine learning hardware. It’s possible we may see a smaller-than-expected decline in compute-related business if the pandemic continues to force long-term closures across the United States, depending on how companies react as short-term contingencies begin to morph into long-term commitments.
Nvidia is widely expected to announce consumer Ampere hardware in the near future, with launch expected this quarter. Both of the main companies in the GPU business are anticipating strong profits in the segment through the end of the year — AMD has both RDNA2 and the Xbox Series X / PlayStation 5 launches as major revenue drivers, while Nvidia has the Ampere refresh and overall momentum it’s built in AI and machine learning markets.
- Microsoft Deploys AI ‘Supercomputing’ via Nvidia’s New Ampere A100 GPU
- Nvidia Could Bring Ampere to Gamers for Just $5 a Month
- Report: Nvidia Ends Production of RTX Turing GPUs, Ampere Is on the Way