Nutanix plans to release a line of server appliances with IBM’s Power processor for artificial-intelligence applications, the first time it has shipped one of its data center bundles with anything but Intel inside.
Nutanix sells a package of hardware and software to companies looking to manage their own data centers, and Fortune reports that it has decided for the first time to include appliances based on Power as part of those packages. IBM has been pushing Power as uniquely suited for AI or machine learning applications, given that it runs its Watson service on the chips.
Nvidia is probably the first chipmaker AI researchers think of when building a new system. The graphics company has adapted nicely to the growing demand for powerful chips to train AI: GPUs (graphical processing units) excel at processing a relatively simple task over and over again, which is ideal for training AI systems, and IBM is actually using some of Nvidia’s chips inside its own public cloud service.
Still, this is very new territory, and there’s lots of time for other companies to find a niche. Intel acquired Nervana last summer to address this area, and IBM is hoping its Watson experience will move a few Power chips.
There could also be some more movement later this year on the idea of Power in cloud data centers, where Intel dominates.
The Register noted Monday that Nutanix has invited Google Cloud Platform boss Diane Greene to speak at its yearly conference this summer. Google is the most prominent supporter of the OpenPower movement (which also includes Nvidia), and confirmed last year that it is running some of its public cloud infrastructure on IBM’s chips. This could mean (as The Register speculated) that Google is striking a hybrid cloud partnership with Nutanix, with Power processors playing some role in the package.
You had to expect that the public cloud providers would do something to try and keep Intel on its toes in this market. It was long believed that Amazon Web Services and others would focus on designing their own chips around the ARM instruction set, which Microsoft has discussed in some detail, but Power processors are available now and backed by a rather large company.