![Nvidia B200 Blackwell](https://fittoblog.org/wp-content/uploads/2024/06/Microsoft-and-NVIDIA-Disagree-Over-Artificial-Intelligence-Servers.jpg)
NVIDIA, known for its advanced gaming GPUs and industry-leading artificial intelligence GPUs, has been on a huge rise in recent months. While its share value is constantly increasing, it became the most valuable publicly traded company in the world a few days ago.
The GPU giant is also known for controlling how its graphics chips are used beyond company walls. In other words, it is a company that constantly controls its relations with its partners and can impose restrictions. It looks like there’s another giant that NVIDIA is at odds with: Microsoft.
The Information reports that there is a disagreement between Microsoft and NVIDIA over how the new Blackwell B200 GPUs will be installed in Microsoft server rooms. The green team is aggressively seeking to make huge profits from the world of servers and AI. They already have the biggest slice of the cake, but considering the situation, this is not enough.
Let’s remember the introduction of Blackwell B200 data center chips. Jensen Huang stated many times during the presentation that he no longer considers single GPU production, but sees the entire rack as one GPU. This is an extra effort to generate additional revenue from AI customers and they want to have a say in customers’ B200 GPU installations.
Previously it was the customer’s responsibility to purchase and build the appropriate server racks to house the hardware. Now NVIDIA is pushing customers to buy individual racks or even entire SuperPods directly from them. The GPU manufacturer claims this will improve GPU performance. Considering all the connections between GPUs, servers, racks, and even SuperPods, there is some truth to what is said. But of course, NVIDIA, which takes action when large-scale data centers are being built and wants to do the installation itself, also has dollar banknotes in its sights.
Small customers may be satisfied with the company’s solutions, or they may be reluctantly accepting them. Microsoft is another world giant.
NVIDIA Vice President Andrew Bell has reportedly asked Microsoft to purchase a server rack design specifically for its new B200 GPUs in a form factor that differs by a few inches from the existing server racks actively used in Microsoft’s data centers. Microsoft turned down NVIDIA’s proposal, explaining that it would prevent new server racks from easily switching between NVIDIA AI GPUs and competing offerings such as AMD’s MI300X GPUs.
Reportedly, the greens eventually backed down and allowed Microsoft to design its own dedicated server racks for its B200 AI GPUs. For now, it looks like a deal has been struck, but this probably won’t be the last conflict between the two megacorporations.
Follow our site to read more technology news!