
putilov_denis - stock.adobe.com
Supermicro launches Nvidia Blackwell systems
The Supermicro Blackwell systems are good news for Nvidia, but enterprises should calculate the ROI carefully, an analyst says.
Super Micro Computer Inc., a key Nvidia partner, has started shipping AI data center systems with the chipmaker's advanced Blackwell GPUs, indicating that the latest generation of AI accelerators has entered the market as planned.
On Wednesday, Supermicro introduced its Building Block portfolio with the core infrastructure elements for Blackwell. The portfolio includes air-cooled or liquid-cooled systems with multiple CPU options for running applications alongside Blackwell-powered AI workloads.
The systems also include a data center management software suite, network switching and cabling. Supermicro has tested the hardware, software and interconnects to validate their ability to run Blackwell clusters under real-world conditions.
During an earnings call in November, Nvidia CEO Jensen Huang faced investor questions about a media report of overheating problems with Blackwell systems. Such problems could delay the systems' arrival at customer data centers.
Huang dodged the question while assuring Wall Street analysts that Blackwell's production was "full steam."
"Blackwell is in great shape, and as we mentioned earlier, the supply and what we're planning to ship this [fiscal] quarter is greater than our previous estimates," he said.
Although Blackwell systems are entering the market, buyers should expect a wait for receiving the chips, Gartner analyst Chirag Dekate said.
"Blackwell demand continues to outstrip supply, and we see several months of wait time for Blackwell today," Dekate said. "However, a lot of the initial spend in large GPU acquisitions is being executed by cloud and hyperscalers, and not necessarily by the broad set of enterprises."
Supermicro offers a broad portfolio of rack-scale systems covering traditional enterprise data centers and hyperscalers. For the former, Supermicro offers a 10U air-cooled system with eight Blackwell B200 GPUs.
However, enterprises should carefully calculate the return on investment before writing the check.
"AI is more than buying a bunch of GPUs," Dekate said. "In fact, buying the GPUs, however scarce they may be, is the easy part. Using them and delivering returns is the hard part."
Pharmaceutical companies, financial institutions and retail giants have found sufficient value in deploying AI on-premises. However, because of the technology's immaturity, deploying AI in-house remains challenging for enterprises with less sophisticated IT departments, analysts said. Until there is more innovation, leaning on the cloud might be the better option for those organizations.
"Cloud providers like Google, Microsoft and Amazon are at the forefront of how customers can access Blackwell," Dekate said.
Supermicro's product release comes two months after the company said it would appoint new legal and financial leadership following a board-appointed special committee's investigation into allegations that the company overstated revenues. The committee found no evidence of misconduct, and a restatement of the company's finances was unnecessary.
The allegations, made in federal court by a former employee, led to the resignation of Supermicro's accounting firm, EY. In its resignation letter, EY said it could no longer rely on information provided by Supermicro management and the company's audit committee.
Antone Gonsalves is an editor at large for Informa TechTarget, reporting on industry trends critical to enterprise tech buyers. He has worked in tech journalism for 25 years and is based in San Francisco. Have a news tip? Please drop him an email.