By some analysts’ accounts, Intel controls close to 99% of the market for microchips used in servers. The current domination is partly based on Intel’s innovations and savvy marketing skills, as well as on missteps by competitors.
But 2017 may be the year that competition reemerges for Intel in the fast-growing server market from challengers like Advanced Micro Devices, Qualcomm, and Nvidia (NVDA). Each is rolling out new products that could offer better performance than the market leader, particularly in some of the hottest sectors like cloud data centers and big data analytics.
“The competitive environment in datacenter appears stronger than it has in many, many years,” longtime chip industry analyst Stacy Rasgon at Bernstein Research noted last week in a report urging investors to sell Intel’s stock. “Further risk exists not only from possible share losses, but also from pricing pressure, should new competition, in whatever form, prove credible enough to crack Intel’s strong pricing umbrella.”
The $50 billion server market is critical because it’s still growing even as the market for desktop and laptop computers has been shrinking. Cloud data center giants Amazon, Microsoft, and Google are feeding the frenzy, buying huge quantities of chips for their ever-expanding facilities.
Intel investors have been growing increasingly concerned about possible problems in the data center business. Intel shares are down 6% since it reported fourth quarter results on January 26, including a 2% drop since Rasgon’s report. Having missed the smartphone revolution, Intel can ill afford to get pushed out of the data center.
The latest competitive hit to Intel came on Wednesday, as Microsoft announced a partnership with Qualcomm to integrate server chips made by Qualcomm with servers in Microsoft’s Azure cloud service. Though in the early stages, the partnership threatens Intel’s leading position in cloud service until now.
The Qualcomm chips are based on designs from ARM Holdings, which have dominated in smartphones but made little headway in older, more traditional computing segments. Qualcomm says its chip is aimed at exactly the kinds of tasks performed in cloud data centers, such as analyzing big data sets.
Get Data Sheet, Fortune’s technology newsletter.
It’s been five years or more since Intel’s competitors started hyping the potential of ARM-based server chips to outperform Intel’s chips. But those rivals failed to make any serious gains.
Qualcomm entered later with ARM-based chips for what it now calls the Centriq 2400 server chip platform, and after two years it won’t ready to sell the chips in large quantities until later this year.
Microsoft has taken a major step in moving away from Intel by creating a version of its Windows Server software that is compatible with Qualcomm’s chip and another ARM design from smaller producer Cavium (CAVM). The ARM-friendly Windows software is set to run only in Microsoft’s own Azure data centers, at least for now, but it could eventually be available to Microsoft’s (MSFT) many thousands of customers as well.
“Our collaboration with Microsoft is multi-faceted and multi-year,” Ram Peddibhotla, head of product management for Qualcomm’s data center group, told reporters in a briefing last week.
Still, the Qualcomm (QCOM) chip won’t be widely available until the second half of the year, and the company wouldn’t disclose its performance relative to Intel’s top chips.
Also on Wednesday, Nvidia and Microsoft announced a new computing set up intended to speed artificial intelligence tasks performed in cloud data centers. The new HGX-1 computer chassis combines eight of Nvidia’s Tesla P100 graphics accelerator boards in one box for faster data exchanges. Although AI is only a small part of the workload in cloud data centers today, it is one of the fastest growing areas and Intel is battling to keep up.
A day earlier, on Tuesday, AMD unveiled its own attack on Intel by introducing a variation of its Ryzen chip line called Naples. As with its Ryzen chips for desktop computers, AMD says its new Naples server chips are fully compatible with Intel’s architecture but will be cheaper than Intel’s and offer better performance. AMD says it combined the computing cores in Naples with other required systems on the same chip, providing greater speed and efficiency over Intel’s multi-chip design.
“We’re energized to once again join the global community of technology leaders who also see the value in rethinking data center hardware,” AMD senior vice president Forrest Norrod wrote in a blog post unveiling the new chip.
Analysts aren’t certain AMD will follow through and be able to manufacture enough chips. The company was near-death, trading at under $2 a share, just over a year ago. But the promise of Ryzen and its spin-offs have tantalized investors, who have pushed the stock price as high as $15.55 in recent days. On Wednesday, the shares were at $13.37 in midday trading.
Intel isn’t standing still, either. It’s finally upgraded its Xeon server chip line with its Skylake design that debuted in desktop computer chips in 2015. Google (GOOGL) announced last month that it was adding the new chips to its data centers pronto while Amazon (AMZN) has also committed to Skylake. Facebook (FB) said on Wednesday that it will use a mix of Intel Skylake chips and Nvidia graphics chips in its newest servers.
And although in the past, Intel brought each new major revision, such as Skylake, first to desktop and laptop chips and only later to server chips, it’s flipped those priorities for the future.
“We’re confident that our products will maintain their performance leadership,” Intel (INTC) said in a statement about AMD’s (AMD) performance claims. “We take all competition seriously, but we have to wait and see if the claims match the realities, especially with real-world applications.”
It may take the whole year to play out, but the server market will be testing Intel’s bold claims like never before.