Nvidia (NVDA 3.12%) made a giant splash on the firm’s annual GPU Know-how Convention (GTC), known as the “Woodstock of synthetic intelligence” (AI). Throughout CEO Jensen Huang’s keynote deal with on Monday, he launched the long-awaited Blackwell structure. He known as the brand new flagship processor, the GB200 Grace Blackwell Superchip, the “world’s strongest chip” for AI.
Past the industry-leading technical specs, a revelation on Tuesday is being seen as a blow to rival Superior Micro Units (AMD 0.54%) and will considerably hamper the corporate’s AI aspirations.
The GB200 Grace Blackwell Superchip. Picture supply: Nvidia.
The “world’s strongest chip” for AI
There have been nice expectations relating to Nvidia’s Blackwell structure, and the corporate didn’t disappoint. The Blackwell graphics processing unit (GPU) boasts 208 billion transistors — up from 80 billion for its predecessor. The GB200 combines a Grace Hopper CPU (central processing unit) with two B200 Tensor Core GPUs and a lightning-fast chip-to-chip hyperlink to type a “single, unified GPU.”
Nvidia additionally launched the GB200 NVL72, a “multi-node, liquid-cooled, rack-scale system … which mixes 36 Grace Blackwell Superchips, which embody 72 Blackwell GPUs and 36 Grace CPUs interconnected by fifth-generation NVLink.” The corporate notes that its rack-scale system offers as much as 30 occasions the efficiency whereas lowering vitality consumption by 25 occasions.
The technical capabilities of the Blackwell processors alone will make life harder for AMD, however Nvidia goes even additional to outpace its competitor.
Lengthy-time rivals
There is a long-standing rivalry between Nvidia and AMD, as the businesses have traditionally jockeyed for place to dominate the GPU house. For instance, at an occasion to introduce its Intuition MI300X AI chip in early December, AMD went to nice lengths to indicate that its flagship processor outperformed its nearest competitor, Nvidia’s year-old H100, utilizing all kinds of benchmarks.
Quickly after, Nvidia dropped a weblog put up that took concern with the claims, noting that AMD purposefully did not use Nvidia’s optimized software program, which skewed the leads to AMD’s favor. Nvidia famous that if it was benchmarked correctly, the H100 was “2x quicker.”
One space by which the rivals have not traditionally competed is value. Regardless of AMD’s protestations on the contrary, Nvidia has lengthy been acknowledged because the {industry} chief, although AMD’s comparable chips are a lot cheaper. A report launched final month means that whereas AMD’s top-of-the-line MI300X chips value between $10,000 and $15,000 every, Nvidia’s H100 offered for as a lot as $40,000, demonstrating the corporate’s pricing energy.
Is Nvidia going for the throat?
In an interview on CNBC, Huang revealed that the Blackwell GPUs, anticipated to ship later this yr, could be priced between $30,000 and $40,000, suggesting the next-generation processors shall be priced equally or at a modest premium to its present processor — the H100.
Nvidia’s aggressive pricing technique was a shock to analysts who cowl the inventory. Morgan Stanley analyst Joseph Moore famous that Nvidia is pricing its new Blackwell chips “at extra aggressive value factors than our earlier assumptions … [which will] probably dent the passion for options to Nvidia, each service provider and customized silicon.”
Mizuho Securities analyst Jordan Klein notes, “The extra modest pricing (for Nvidia’s Blackwell GPUs) was perceived as constructive for (Nvidia) inventory … It opens the door for extra prospects to purchase vs. solely the extremely concentrated large capex spenders in cloud hyperscalers and rich sovereign nations.”
The pricing will give new and present prospects another excuse to decide on Nvidia over AMD, as the value differential is shrinking.
One other nail in AMD’s coffin?
To be clear, AMD has established itself as a worthy, although considerably distant, competitor. Nvidia’s latest processor will definitely up the ante, as will its aggressive pricing coverage. There’s the likelihood that Nvidia might enhance its already dominant market share within the AI house or preserve its rivals at bay for some time longer.
Whereas these strikes will definitely put AMD at a drawback, it is unlikely that AMD shall be going anyplace anytime quickly.