AI titans Microsoft and Nvidia reportedly had a dispute over Microsoft's use of B200 AI GPUs in its server rooms.

WhatsApp Group Join Now
Telegram Group Join Now
Instagram Group Join Now

Nvidia is known for its high-performance gaming GPUs and industry-leading AI GPUs. However, the trillion-dollar GPU maker is also known to control how its GPUs are used outside the company's walls. For example, it can be quite limited with the graphics card designs of its AIB partners. Perhaps unsurprisingly, this level of control appears to extend beyond card partners to AI users, including Microsoft. Microsoft and Nvidia reportedly had a standoff over how Nvidia's new Blackwell B200 GPUs were to be installed in Microsoft's server rooms.

Nvidia is aggressively pursuing bigger pieces of the data center pie, which is immediately clear if you look at how it announced the Blackwell B200 parts. Several times during the presentation, Jensen Huang indicated that he doesn't think of individual GPUs anymore – he thinks of the entire NVL72 rack as a GPU. This is a transparent effort to generate additional revenue from its AI offerings, and it influences how customers install their new B200 GPUs.

Previously, the customer was responsible for purchasing and building the appropriate server rack to house the hardware. Now, Nvidia is pushing customers to buy individual racks and even entire SuperPods — all directly from Nvidia. Nvidia claims this will increase GPU performance, and makes sense considering all the interconnections between different GPUs, servers, racks and even SuperPods. But when you're building data centers at scale, a lot of dollar bills also change hands.

Nvidia's smaller customers might be fine with the company's offerings, but Microsoft wasn't. Nvidia VP Andrew Bell reportedly asked Microsoft to purchase a server rack design specifically for its new B200 GPUs with a form factor a few inches different from Microsoft's existing server racks that are actively deployed in its data centers. are used.

Microsoft pushed back on Nvidia's recommendation, revealing that the new server racks would prevent Microsoft from easily switching between Nvidia's AI GPUs and competing offerings such as AMD's MI300X GPUs. Nvidia eventually backed down and allowed Microsoft to design its own custom server racks for its B200 AI GPUs, but this probably isn't the last such disagreement we'll see between the two megacorps.

Such controversy is a sign of how big and valuable Nvidia has become in just a year's time. Nvidia became the most valuable company (briefly) earlier this week, and that title is likely to change hands several times in the coming months. Server racks aren't the only area Nvidia wants to control, as the tech giant also controls how much GPU inventory is allocated to each customer to keep up with demand, and it also controls its software and networking systems. AI is using its dominant position in the space to push forward. To maintain its position as the market leader.

Nvidia is benefiting in a big way from the AI ​​boom, which started when ChatGPT exploded in popularity a year and a half ago. Over the same period, Nvidia's AI-focused GPUs have become the GPU manufacturer's most in-demand and top-grossing products, leading to incredible financial success. Stock prices for Nvidia have risen and are currently eight times higher than early 2023, and 2.5 times higher than early 2024.

Nvidia continues to use all that extra revenue to great effect. Its latest AI GPU, the Blackwell B200, will be the world's fastest graphics processing unit for AI workloads. A “single” GB200 Superchip delivers up to 20 petaflops of compute performance (for a wasteful FP8) and is theoretically five times faster than its H200 predecessor. Of course the GB200 is actually two B200 GPUs plus a Grace CPU on a big daughterboard, and the B200 uses two big dies connected together for good measure. The 'regular' B200 GPU, on the other hand, only offers up to 9 petaflops of FP4 (with it), but that's a huge number crunching capability nonetheless. It also provides up to 2.25 petaflops of dense compute for FP16/BF16, used in AI training workloads.

As of June 20, 2024, Nvidia stock closed at $126.57, down slightly from its high of $140.76. We'll have to see what happens once mass deliveries of the Blackwell B200 begin. Meanwhile, Nvidia continues to be at the forefront of AI.

WhatsApp Group Join Now
Telegram Group Join Now
Instagram Group Join Now

Leave a Comment