Nvidia (NVDA -1.85%) built an empire in what could be today's most promising technology: artificial intelligence (AI). It all started with a powerful chip -- the graphics processing unit (GPU) -- which is designed to rapidly handle specific types of computing problems: those that can be easily broken down into a large set of smaller problems that can solved simultaneously.
Initially, Nvidia sold this product to the gaming market, because producing high-resolution, fast-moving video game graphics is just the sort of task that parallel processors are made for. But over time, the tech world found other uses for GPUs. And eventually, when AI started really taking off, it became clear that they were the best hardware available for the training and inferencing of AI models.
With the best GPUs on the market and a highly popular platform for developers, Nvidia rapidly became the No. 1 AI chip designer and developed a wide range of products and services to enhance its status as the go-to purveyor of AI hardware.
Still, one threat that Nvidia continues to face is the emergence of rivals, from fellow chip designer Advanced Micro Devices to some of its own big tech customers, which are designing their own AI chips. Competitive offerings from these players could put pressure on Nvidia's prices and slow its red-hot growth.
But Chief Executive Officer Jensen Huang, speaking at Computex (a computer industry expo in Taiwan) earlier this week, announced a key new release from the company -- something that could greatly reduce its risk of losing meaningful market share to those would-be challengers. Did Nvidia just say "checkmate" to its rivals?
Image source: Getty Images.
Nvidia's record revenue
First, a quick summary of the Nvidia story so far. The company has shifted its GPU focus from video games to AI, and added a multitude of related products. All of this has resulted in massive revenue growth. It increased its quarterly revenues by rates in the double- and triple-digit percentages year over year, with these figures setting new records steadily. And, importantly, its top-line growth came at a high level of profitability, with gross margin surpassing 70%.
But other chipmakers released competitive products at lower prices and Nvidia customers like Amazon and Meta Platforms have spoken about their own in-house developed chips. For example, Amazon Web Services (AWS), sells (and rents cloud computing time on servers powered by) its own Trainium chip, a product it says its cost-conscious customers appreciate.
Well, Nvidia in this area has decided if you can't beat them, join them. It's now making it easier for customers to use Nvidia systems along with central processing units (CPUs) and GPUs that are designed and manufactured by other companies. This is through the release of NVLink Fusion. This system will allow customers using NVLink -- a high-speed connection for the sharing of data between processors -- to now include non-Nvidia processors in that network.
"NVLink Fusion opens Nvidia's AI platform and rich ecosystem for partners to build specialized AI infrastructures," Huang said.
Putting Nvidia to work with others
Prior to this, NVLink only could be used with Nvidia processors, meaning customers would have to make a decision: either build with Nvidia or not. Now, with NVLink Fusion, customers don't have to choose. They could put their Nvidia processors to work with its peers' CPUs, GPUs, and application-specific integrated circuits (ASICs), too.
The one potential downside of this strategy for Nvidia is it may lead to slipping sales of Nvidia's CPUs, as customers won't have to invest in everything Nvidia to use NVLink. But I think the positives well outweigh that possible negative.
NVLink Fusion offers chip buyers greater flexibility -- something that could be greatly appreciated as options multiply in this fast-paced market. It's important to note that NVLink Fusion makes it possible for customers to truly tailor their data centers to their needs and budgets. And this offers Nvidia an opportunity to broaden its customer base to ASIC data centers and to ensure it will maintain a key role in the AI infrastructure build-out as the AI boom progresses.
So, this announcement may seem like a detail, but the debut of NVLink Fusion could be a significant move -- one that could keep Nvidia ahead.
Does this mean Nvidia just said "checkmate" to rivals? Yes and no. I say "no" because this new system won't result in its rivals failing in AI or leaving the space. There's plenty of room for more than one player to be successful in this part of the chip space, considering the value of the AI market is expected to soar into the trillions of dollars by the end of the decade.
But Nvidia has said "checkmate" when it comes to ensuring its leadership -- the release of NVLink Fusion may be the move that, along with Nvidia's innovation, keeps it on top of the AI world for the long haul.
