HPE awarded over $1 billion contract to build xAI servers, the artificial intelligence created by Elon Musk and which can already be experienced through Grok on the social network
This represents a major success for HPE, traditionally not a leader in the AI server sector. The size of the deal highlights the growing importance of artificial intelligence for social platforms. Although the technical details have not been disclosed it is practically certain that the servers will use GPU NVIDIA (specifically the new H200), the key components of these servers which can constitute half of the total cost.
A $1 billion deal is significant even by AI industry standards.
In fact, assuming the use of H200 GPUs based on Blackwell technology, their unit cost would be $50,000. X should be equipped with a cluster with at least 10,000 GPUscapable of offering performance up to 90 FP4 ExaFLOPS. We leave it up to you to do the math to find out the cost of these servers just for the GPUs.
Obviously it remains to be seen how X intends to exploit this computing power, given that in addition to Grok there are several projects in the pipeline for xAIespecially considering the constant influx of investors and investments that it attracted following the presentation of the Colossus supercomputer.
Before Elon Musk’s acquisition in 2022, Twitter sourced from companies like MiTAC, Supermicro, and Wiwynn. Subsequently, it only maintained relationships with Supermicro. Interestingly, xAI, Musk’s venture into artificial intelligence, commissioned Dell to build the Colossus supercomputer, equipped with 100,000 Nvidia H10 GPUs0, but even following that result, the company was unable to sign the agreement to build the xAI servers.
Source: www.tomshw.it
