It’s been almost exactly a year since AMD launched its first Epyc server CPUs. Over the last 12 months, AMD has fed the market with a slow, steady drip of new products and design wins as it begins the process of re-entering the server market in earnest for the first time in over half a decade. Of course, there’s a small problem standing in between AMD and its attempts to re-enter the server business — Intel, and it’s ~99 percent market share.
Intel CEO Brian Krzanich recently sat down to talk to Instinet’s Romit Shah, who warns that the CPU giant faces “stiff competition” from AMD and its Epyc line of processors. According to Shah, Krzanich was remarkably matter-of-fact in his discussion of the situation. He admitted that Intel is likely to lose some market share to AMD in the back half of the year, which isn’t surprising — when you have 99 percent of a market, the only place to go is down.
What’s a little more surprising is how Shah characterizes the rest of the conversation with Intel. Shah found it significant that “Mr. Krzanich did not draw a firm line in the sand as it relates to AMD’s potential gains in servers; he only indicated that it was Intel’s job to not let AMD capture 15-20 [percent] market share.”
There’s an awful lot packed into that sentence and billions of dollars at stake. Let’s unpack it. Intel’s CEO, Brian Krzanich, has been very clear that cloud data centers are absolutely critical to Intel’s long-term future. While the PC space is obviously still important to Intel — the company’s economies of scale depend on selling the huge number of processors it ships in that segment — it doesn’t view PCs as a growth opportunity given that the market has shrunk by 25 percent since 2011. Cloud and data center revenue, however, have skyrocketed over the same period. In 2013, Intel reported $12.2B in Data Center Group (DCG) revenue and $5.5B in operating income. In 2017, Intel reported $19.1B in DCG revenue and $8.4B in income. Its consumer client group (CCG) operating income for 2017 was $12.9B, which gives you some idea how important DCG is to Intel’s overall bottom line.
The following sheet from Intel’s Form 10-K at the end of Q4 2017 gives some useful additional details on how the DCG business has grown of late. For Q1 2018, Intel reported DCG revenue of $5.2B and operating income of $2.6B, with the latter up a whopping 1.75x year-on-year.
Intel’s data center business is growing quickly, it’s richly profitable, and it’s the company’s only other business segment that really brings home the bacon. IoT, non-volatile memory, and the programmable solutions group are less than half the size of the DCG combined. Yet Krzanich’s comments don’t sound like a CEO concerned that a competitor could eat a hefty chunk of his overall data center business.
There are some possible reasons for this. One is that Intel, while willing to acknowledge that AMD could take some market share in the near-term future, doesn’t believe its smaller rival will prove to be a significant long-term threat. There’s historical reason to take this view. AMD launched its first server products in 2001 with the 760MP chipset, but didn’t take significant market share until the debut of dual-core Opteron processors.
AMD’s market share surged from 2005 – 2006, but the company lost a significant percentage of those gains from 2006 – 2007 and the rest degraded by slow degrees over the next five years. Intel may be betting that history repeats itself — and the company has a history of being very defensive of its server market. When companies like Calxeda began talking about building high-density ARM servers, Intel quickly responded by repositioning Atom as a low-power Xeon part. It then invested in evolving and building on that platform, from 2012 through to the present day, even though Calxeda has long since died and the threat to Intel’s server business from high-density ARM CPUs isn’t nearly as pressing as it appeared to be six years ago.
Holding AMD to 15-20 percent market share would give Intel protection against accusations of monopoly while still allowing it to command the lion’s share of the profits in the data center business, especially since it knows big iron customers are more conservative than smaller companies. Server customers are conservative by nature and they only become more so the more the hardware costs.
Some of this chatter about a 15-20 percent market share for AMD is premature no matter what; CEO Lisa Su has previously said that she’s targeting 4-6 percent of the server market this year, not 15-20 percent. But as a matter of principle, I’d expect Intel to fight back harder in servers than anywhere else, including the mainstream PC market. Optane, long term, is a major server memory play. Cloud computing, machine learning, and AI are all expected to be hot topics the next few years, and those are all part of the DCG as well. Krzanich may not have come out swinging by claiming Chipzilla would hold AMD to a 3-5 percent market share, but based on Intel’s response to ARM’s potential threat and the importance of DCG to its long-term earnings, I’d expect the company to be nothing less than ruthless when it comes to protecting its server revenue.
Windows 10X, Microsoft’s Chromebook Competitor, Supposedly Coming Later This Year
Windows 10X will reportedly ship on systems in H2 2021. Originally intended to compete as a high-end OS variant for premium devices, Windows 10X is now intended for educational markets, where Chromebooks have recently made huge strides.
As AMD and Intel Struggle, a Competitive CPU Market Re-emerges
Everything old is new again — and in this case, what's new are process node problems and transition challenges.
Intel Positions Xeon as Machine Learning Competitor in Inference Workloads
Intel is tossing its own hat into the machine learning ring, positioning Xeon as a potent inference workhorse.
Nvidia AI Compensates for Your Poor Photography Skills by Erasing Noise From Images
Even if the Noise2Noise system has never seen an image before, it can de-noise it to get something very close to the original.