At the moment, Intel (INTC 0.43%) is the undisputed king of the data center. The company enjoys a near-monopoly in the server chip market, allowing its data center segment to generate incredible profits. Serious competition is on the horizon: AMD will launch new server chips next year, and various companies are attempting to make ARM-based chips viable for data center applications. But the biggest threat to Intel may be International Business Machines (IBM 0.40%).
IBM, which sells expensive mainframe systems and high-end servers built around its POWER processors, completely missed the boat on the rapid rise of hyperscale data centers. Major cloud and internet companies often design their own systems to meet their specific needs, and Intel chips have been the only real option for years.
The OpenPOWER foundation, started by IBM in 2013 and now boasting more than 150 member companies, aims to open up IBM's POWER architecture and create a viable alternative to Intel in the data center. There's been slow and steady progress, with third-party POWER servers now available. But the next iteration of IBM's chips, POWER9, will be the real test of IBM's strategy to win back share in the data center market.
Taking on Intel
IBM's goal is to achieve a 10% to 20% share of both the high-performance computing and hyperscale data center markets. The company is starting from close to zero in both markets, so it will certainly be an uphill climb. The good news is that the POWER9 processor, set to be launched during the second half of next year, has the potential to be game changer for IBM.
IBM's POWER processors are highly parallel, supporting up to eight threads per core. Each POWER9 processor is expected to come with up to 24 cores, making the chip well suited for applications that benefit from being run in parallel. POWER 9 is expected to be built on a 14-nanometer process, which will help narrow Intel's manufacturing advantage.
Details beyond the basics are scarce, but IBM has already managed to win two important POWER9 deals. Back in November 2014, IBM announced that the U.S. Department of Energy had chosen IBM's POWER9 processors to power two of its upcoming supercomputers. Summit, the larger of the two, will be capable of 200 petaflops of peak processing power, which would easily make it the fastest supercomputer in the world when it's delivered in 2018.
Along with IBM's POWER9 chips, Summit will be accelerated with NVIDIA's (NVDA -2.81%) next-gen Volta GPUs. NVIDIA was a founding member of the OpenPOWER foundation, and its NVLink technology, which allows for CPUs and GPUs to exchange data at a faster rate, was contributed to the foundation and is supported by IBM's processors. NVIDIA currently dominates the market for accelerators, but Intel is now shipping a competing product that has a good chance at winning some market share.
While the supercomputer deal is a one-off event, the recent announcement that Alphabet's (GOOGL -2.75%) Google and Rackspace were developing a server architecture based on IBM's POWER9 processors has broader implications. Google's goal is to make its infrastructure architecture-agnostic, giving it the freedom to choose the processor that best fits its needs. At this point, Google's entire toolchain supports the POWER architecture, and with the company also being a founding member of the OpenPOWER foundation, it seems likely that POWER9 processors will eventually find their way into Google's data centers.
Given how much cloud and internet companies like Google spend on server processors each year, there's a major incentive to have options beyond Intel. IBM's POWER9 processor is one of a few threats that Intel's data center business will need to deal with in the coming years. Intel is unlikely to lose its crown as the king of the data center anytime soon, but there's a good chance that IBM can carve out a portion for itself.