By 2025, NVIDIA’s long-awaited Blackwell DGX AI servers will be making their way to the market, marking the next major growth phase in the industry that is set to take place. The latest and next generation of the DGX AI server series, based on NVIDIA’s Blackwell chip, is expected to result in the impetus of the market, making NVIDIA an unexpected beneficiary of this demand
More About NVIDIA Blackwell DGX GB200 AI Servers
The DGX AI servers bearing the GB200 label should have considerable architectural improvement as well as performance, with significant claims that the efficiency will help the series be adopted even more than the Hopper variety. Reports confirm the mass production campaign for the GB200 servers will kick off in the latter phase of 2024, with approximately 40,000 of the equipment in circulation by 2025.
NVIDIA will reportedly be distributing the Blackwell DGX GB200 AI server supply, which is comprised of three distinct segments: DGX NVL72, NVL32, and HGX B200. For the NVL72, which will be the flagship, NVIDIA will deploy 72 units of Blackwell B200 AI GPUs and employ the Grace Hopper chip. DGX NVL72 will see around 10,000 units sold, generating $30 billion in revenue with the price for one unit “cabinet” expected to be around $3 million.
The supply chain will be heavily dependent on Foxconn and Quanta, both of which expect a boost in their server revenue, with Foxconn estimating a turnover of 1 trillion yuan, or $30 billion by 2025. This underscores the immense revenue potential in the AI markets, with NVIDIA and its suppliers poised for substantial financial growth.
As demand for AI computing power continues to escalate, the data center markets are expected to expand concurrently, with NVIDIA positioned as a dominant player in this evolving landscape. This marks just the beginning of what promises to be a period of sustained growth for NVIDIA and its partners in the AI ecosystem.
