Nvidia’s Jensen Huang kicks off Computex 2024 with a highlight on GenAI and an accelerated GPU roadmap
This 12 months’s Computex commerce present is about to be a historic one, with the CEOs of all of the world’s largest semiconductor firms giving keynotes and doing so underneath the aura of the generative AI revolution and the launch of the AI PC class.
The primary keynote at this 12 months’s occasion, appropriately sufficient, was given by Nvidia’s Jensen Huang, a Taiwanese native who has arguably develop into the face of the semiconductor revolution within the AI period.
Huang’s speech, which felt like the following cease in his worldwide stadium tour, basically mixed an entertaining lecture on the historical past and capabilities of GenAI with some particular examples of Nvidia merchandise which are enabling the GenAI revolution. Alongside the way in which, he described the idea of a brand new industrial revolution, with AI factories which are powered – after all – by Nvidia {hardware} and software program.
A lot of the product-specific data was a rehash of the Blackwell platform information that the corporate unveiled at their GTC convention again in March. What was new, nonetheless, was the revealing of a product roadmap that extends into 2027 with the debut of the Rubin platform as the following evolution past Blackwell. American astronomer Vera Rubin will likely be Nvidia’s next-gen GPU structure patron.
Whereas particulars stay scarce, in a market shifting at such a breakneck tempo as GenAI, simply having a way of normal route will likely be comforting for a lot of. In truth, I am anticipating the notion of longer roadmap unveilings to be a giant theme all through the following a number of days of keynotes.
Whereas large architectural modifications are prone to proceed at an each two years, Nvidia is now shifting to make essential GPU product enhancements yearly.
One level that Huang made clear is that the roadmap for his or her core GPUs is taking over an accelerated annual tempo. Whereas large architectural modifications are prone to proceed at an each two years, Nvidia is now shifting to make essential product enhancements yearly. How the corporate is ready to execute on that new roadmap stays to be seen, nevertheless it’s clear they haven’t any intention of slowing down or ceding any of their dominant place in AI accelerators any time quickly.
One other comparatively new space that Nvidia has been discussing extra lately – and that Huang detailed in his keynote – was the corporate’s transfer into software program and providers.
Their CUDA platform for AI acceleration has already develop into one thing of a normal, however the firm is pushing laborious on its NIM (Nvidia Inference Microservices) choices, which are actually usually accessible. NIMs construct on CUDA and wrap collectively all the mandatory software program parts wanted to construct and allow AI purposes throughout particular areas into prebuilt containers.
Their creation and launch is a part of an even bigger firm technique to get into the enterprise software program market and usher in a wholly new set of revenues. Although not mentioned throughout the keynote, it is also value mentioning that Nvidia has lately been partnering with main enterprise {hardware} firms like Dell Applied sciences, Lenovo, HPE and others.
As a result of Nvidia is not usually seen as a software program provider, these companions may assist dramatically enhance the attain of those instruments into the enterprise. That is going to be fascinating to look at as a result of these enterprise distributors may find yourself being an essential channel for Nvidia’s software program choices over time.
What wasn’t introduced within the keynote – however has been the topic of a lot hypothesis right here at Computex – was the corporate’s intention to enter the AI PC market with an Arm-powered SoC, probably constructed along with Mediatek.
The 2 firms are already working collectively on a mixed chip for the automotive market so it isn’t troublesome to think about they might lengthen that partnership into the PC market as properly. In the event that they did – and the timing would probably be someday in 2025 or later – it may have a huge impact on what’s already changing into a crowded marketplace for PC chip suppliers. We will see.
Huang ended his Computex keynote with a forward-looking dialogue on AI-powered robotics and the essential function he believes they’ll play in that area thanks, partly, to their Omniverse platform. There’s little doubt that robotics is shaping as much as be an essential improvement each for industrial purposes and human interactions, so this will likely be yet one more fascinating space to look at within the years to return.
Bob O’Donnell is the president and chief analyst of TECHnalysis Analysis, LLC, a market analysis agency that gives strategic consulting and market analysis providers to the know-how trade {and professional} monetary neighborhood. You possibly can observe Bob on Twitter @bobodtech
