Intel holds its annual investor meeting today, giving analysts a chance to hear about the company’s strategy directly from top executives.
This year’s meeting comes at an interesting time. The PC market seems to be recovering a bit, but still isn’t growing much. Intel’s vaunted tick-tock product development cadence broke down as it worked out the kinks with the industry’s first 14nm processors, but, the Core M line, is now ramping and the delay doesn’t seem to have had much impact on Intel’s market share.
Intel’s mobile business, on the other hand, still seems to be struggling to find its footing. The data center business continues to be a key growth engine, propelled by the launch of the Haswell-based Xeon E5s, and is diversifying into all sorts of areas including low-power chips, custom processors, and networking and storage gear.
Finally there’s the nascent wearables category where there are lots of questions about Intel’s Quark processors and the sorts of products in which they will be used.
Here are some of the things I’ll be listening for:
How will Broadwell and Skylake roll out next year?
The 14nm Broadwell processors are a year late, but the first systems using Core M processors are now starting to reach the market — including the Acer Aspire Switch 12, Asus Zenbook UX305 and Transformer Book T300FA, Dell Latitude 13 7000 Series, HP Envy x2 series and Lenovo ThinkPad Helix. The ultra low-voltage Core M chips (around 4.5 watts) are specifically designed for these sorts of hybrid 2-in-1 devices, which are still a relatively small portion of the PC and tablet markets.
The question is how quickly Intel will be able to introduce more powerful Broadwell processors for mainstream laptops. In addition, Intel says it will get back on track with the next microarchitecture, Skylake, which is scheduled to ship next year. How will Intel and its customers, which design the motherboards and systems, handle the overlap between Broadwell and Skylake? Will they target different market segments or will Broadwell simply have a very short life?
What happened to the roadmaps?
There was a time when chipmakers gave detailed roadmaps, which gave a good idea of what they planned to release quarter-by-quarter over the next couple of years including the microarchitecture, number of cores and other features. Nowadays we get codenames and the year in which they might be released. (AMD’s official roadmap is particularly bad; it still ends at 2014.)
There are a couple of possible reasons for this.
First, the pace of innovation has picked up. Mobile chipmakers in Asia can design high-performance chips based on turn-key solutions from ARM and Imagination Technologies, and have them running in TSMC’s fabs in as little as nine months. In that world it is harder for companies like Intel and AMD to publicize their plans three or four years out.
Second the stakes have become bigger. Chips are more complex, integrating many features, and it is harder to stay on Moore’s Law as transistors approach physical limits. As a result the costs to design and manufacture chips have increased significantly, and few chipmakers want to tip their hands.
Finally, for Intel, there is little competitive pressure in PCs and servers. AMD has been distracted by “growth areas” — semi-custom chips for game consoles, embedded processors and low-power servers using both ARM and x86 cores — and has removed all major x86 PC or server redesigns from its roadmap (though here’s hoping this will change under new CEO Lisa Su). As a result there simply isn’t much incentive for Intel to talk about how it plans to stay ahead on performance and power. I get all this.
Is Intel making headway in mobile?
This is the question on everyone’s mind, in particular after The Wall Street Journal reported earlier this week that Intel would merge the mobile business into its much larger PC Client Group under Kirk Skaugen. Intel seems to be on track to meet its goal of providing the processors of 40 million tablets this year, which makes it a real player in the merchant market (Apple designs its own chips for iPads and Samsung supplies the processors for its tablets).
The problem is that Intel is essentially giving away these chips to gain market share, losing billions of dollars in the process. That’s becauseare really designed for high-end tablets, while much of the volume in the market has shifted to low-cost Android devices.
Intel’s mobile portfolio should become more competitive with the release of the low-cost SoFIA platform—with 3G wireless by the end of this year followed by a 4G LTE version in the first half of 2015—as well as the 14nm successor to Bay Trail, code-named Cherry Trail, also in 2015. Intel says SoFIA was built to have no “contra revenue,” which means they plan to sell it at a profit.
Last month Intel announced its first 4G LTE smartphone in the U.S., the Asus PadFone X mini, which has an Atom processor and Intel XMM 7160 LTE modem, but this is really a niche product. And The Wall Street Journal’s report stated that the wireless chip business would not shift to the new client computing group, but would instead be part of a new “wireless platform research and development organization.” This makes me wonder whether Intel is fully committed to competing head-to-head with Qualcomm in wireless or whether—like Broadcom, Nvidia and many others before them—it will scale back its ambitions here.
What exactly are Intel’s plans in China?
To help break into China, Intel has made investments in two major chip designers there,and Spreadtrum, and forged partnerships to co-develop mobile processors with them. CEO Brian Krzanich has also hinted that more deals like this could be on the way. But Intel hasn’t really explained the extent of this co-development.
The first chip, Rockchip’s XMM6321, seems to be more of a repackaging of a low-power ARM Cortex-A5 dual-core processor and an Infineon AG620 3G baseband (Intel purchased Infineon’s wireless business in 2011) to target low-cost Android devices with screens from 3.5 to 7 inches. Surely Intel has bigger things in mind here. The first SoFIA-based chips from these partnerships are due to arrive in 2015 and ultimately Intel hopes its partners will shift altogether from ARM CPU cores to its x86 Atom designs.
How serious is Intel about the foundry business?
Intel’s decision to manufacture chips for other companies was a big departure. It started very small, but over time the foundry business seems to be gathering momentum. Intel now has more than the six customers—the most recent announcement was the addition of Panasonic in July—and two of them are in volume production with a third ramping. Intel currently offers customers one 22nm manufacturing process, but at 14nm it will offer both general-purpose and low-power processes. By the end of next year, it may offer foundry customers a chance to test out its 10nm process on shared wafers know as shuttles.
In the past, Intel executives have said they won’t manufacture chips for other companies that compete directly with its own products. But as Intel expands into more areas, and as more companies want access to its leading-edge manufacturing, it will become increasingly difficult to draw that line.