June 2, 2024
By Bob O'Donnell
As I described in a recent LinkedIn post, this year’s Computex trade show in Taiwan is a historic one, with the CEOs of all the largest semiconductor companies in the world giving keynotes and doing so under the aura of the Generative AI revolution and the launch of the AI PC category.
The first speech at this year’s event, appropriately enough, was given by Nvidia’s Jensen Huang, a Taiwanese native who has arguably become the face of the semiconductor revolution in the AI era. Huang’s speech, which felt like the next stop in his worldwide stadium tour, essentially combined an entertaining lecture on the history and capabilities of GenAI with some specific examples of Nvidia products that are enabling the GenAI revolution. Along the way, he described the concept of a new industrial revolution, with AI factories that are powered—of course—by Nvidia hardware and software.
Much of the product-specific information was a rehash of the Blackwell platform news that the company first unveiled at their GTC conference back in March. What was new, however, was the unveiling of a product roadmap that extends into 2027 with the debut of the Rubin platform as the next evolution beyond Blackwell. While details remain scarce, in a market moving at such a breakneck pace as GenAI, just having a sense of general direction will be comforting for many. In fact, I’m expecting the notion of longer roadmap unveilings to be a big theme throughout the next several days of keynotes. Companies have been finding it hard to plan their AI strategies given how fast things have been moving, so they’re bound to be looking for this kind of information.
One point that Huang made clear is that the roadmap for their core GPUs is taking on an accelerated annual pace. While big architectural changes are likely to continue at an every two years pace, Nvidia is now moving to make important product enhancements every year. How the company is able to execute on that new roadmap remains to be seen, but it’s clear they have no intention of slowing down or ceding any of their dominant position in AI accelerators any time soon.
Yet another area relatively new area that Nvidia has been discussing much more recently—and that Huang detailed in his keynote—was the company’s move into software and services. Their CUDA platform for AI acceleration has already become something of a standard, but the company is pushing hard on its NIM (Nvidia Inference Microservices) offerings, which are now generally available. NIMs build on CUDA and wrap together all the necessary software components needed to build and enable AI applications across specific areas into prebuilt containers.
Their creation and release is part of a bigger company strategy to get into the enterprise software market and bring in an entirely new set of revenues. Though not discussed during the keynote, it’s also worth pointing out that Nvidia has recently been partnering with major enterprise hardware companies like Dell Technologies, Lenovo, HPE and others. Because Nvidia isn’t generally seen as a software supplier, these partners could help dramatically increase the reach of these tools into the enterprise. This is going to be an interesting area to watch because those enterprise-focused vendors could end up being a very important channel for Nvidia’s software offerings over time.
What wasn’t announced in the keynote—but has been the subject of much speculation here at Computex—was the company’s intention to enter the AI PC market with an Arm-powered SOC, potentially built in conjunction with Mediatek. The two companies are already working together on a combined chip for the automotive market so it’s not difficult to imagine they could extend that partnership into the PC market as well. If they did—and the timing would likely be sometime in 2025 or later—it could have a big impact on what is already becoming a crowded market for PC chip suppliers. We shall see.
As with his GTC speech, Huang ended with a forward-looking discussion on AI-powered robotics and the important role he believes they can play in that field thanks, in part, to their Omniverse digital twin software platform. There’s little doubt that robotics is shaping up to be an important development both for industrial applications and human interactions, so this will be interesting to watch.
Here’s a link to the original article: https://www.linkedin.com/pulse/computex-part-1-nvidia-expands-genai-vision-bob-o-donnell-xdcfc/ Bob O’Donnell is the president and chief analyst of TECHnalysis Research, LLC a market research firm that provides strategic consulting and market research services to the technology industry and professional financial community. You can follow him on LinkedIn at Bob O’Donnell or on Twitter @bobodtech. |