clock menu more-arrow no yes

Intel Beats Q2 Expectations and Cautiously Boosts Outlook, Buys Extra Time for Moore's Law

Server revenue grew 10 percent, offsetting sales to PC makers, which didn't fall as much as expected.

Asa Mathat

Shares of chipmaker Intel rose by more than 6 percent after hours as the company posted second-quarter results that were stronger than analysts expected.

Intel posted a per-share profit of 55 cents on revenue of $13.2 billion. The results were solidly better than the 50 cents a share and revenue of $13 billion that analysts had expected.

The company also said in a statement that it expects revenue of $14.3 billion in the third quarter, which is better than the consensus outlook of analysts, who said they expected sales of $14.1 billion and a profit of 56 cents per share. It said it expects gross margins — a closely watched metric relating to its profitability — of about 63 percent, or about half a percentage point higher than previously forecast. For the year, Intel said it expects sales of $54.8 billion, which would amount to a year-on-year decline of about 1 percent, and said it will cut its expected spending on capital expenditures by $1 billion to about $7.7 billion.

The stronger results are a surprise in part because of slackening demand for personal computers, Intel’s largest market for chips. Consumers and companies are thought to be holding off on buying new computers ahead of the new version of Microsoft’s Windows operating systems and a new generation of chips coming later this year.

“The PC has less consumer competition that it has had in years, tablets are on the decline and PCs that are thin, light and sexy are finally affordable,” analyst Pat Moorhead of the research firm Moor Insights and Strategy said about Intel’s cautious outlook. “Consumers need to buy new machines sometime, right?”

But not yet. Intel said sales in its Client Computing Group (Intel calls PCs clients) fell by 14 percent year on year, slightly better than the expectations set by market research firms IDC and Gartner, which clocked a 17 percent global decline in sales.

Those weak results in PCs were offset by sales in the Data Center Group, where Intel sells chips for servers, that rose 10 percent.

“Second-quarter results demonstrate the transformation of our business as growth in data center, memory and [Internet of Things] accounted for more than 70 percent of our operating profit and helped offset a challenging PC market,” CEO Brian Krzanich said in a statement.

But the bigger news concerning Intel’s long-term future came on a conference call with analysts. Krzanich essentially pushed out the company’s plan to upgrade its manufacturing technology by about six months. In recent years it has followed a two-year manufacturing cycle it calls “tick-tock,” where it shrinks the size of the chips it builds in one year and introduces a new design in the second.

Krzanich said a more realistic cycle has increased to two-and-a-half years. He said this created an opportunity to introduce a third product during the current 14-nanometer cycle. That third product, coming in late 2016, he said will be code-named “Kaby Lake” and will be built on the foundations of the “Sky Lake” generation of processors due this fall.

For perspective: Intel’s current generation of chips are built on a 14-nanometer process technology. A nanometer is a billionth of a meter and an object of 14 nanometers in size is smaller than a typical virus particle, and about the equivalent to the thickness of the outer cell wall of a typical germ.

In fairness, Intel never expressly said when it would start building 10-nanometer chips, but analysts have considered late 2016 or early 2017 about the right time to expect it. Now it won’t arrive until mid-2017 at the earliest, and will last as Intel’s leading-edge process until early next decade.

Getting from one generation to the next is getting increasingly difficult and is starting to take longer. Intel said in a regulatory filing last year that it is taking progressively longer to complete each “tick-tock” transition. What once took about two years and change may now take as long as three.

That will add to the time it takes for Intel to research and ramp up the manufacturing generations that follow. Seven nanometers would arrive sometime in, say, 2022, and five nanometers roughly a decade from now. The date of reaching five nanometers — about twice the size of a strand of DNA — is important because after then, the best experts say that Moore’s Law will have to yield to the reality of physics. It will be impossible, the thinking goes, to shrink transistors on a chip any further, and it will reach its logical end.

Here’s why: When transistors on chips reach the size of individual atoms, physics says you can’t go any smaller. To get the expected doubling of computing power every few years, each successive generation of chips would have to be bigger, and thus more expensive to produce. This would reverse Moore’s Law which has been a fundamental underpinning of the improvement of computing during the last five decades.

Extending the “tick-tock” pattern by six months amounts to an admission that Moore’s Law is getting increasingly difficult to keep on schedule, and has the appearance of an attempt to buy researchers an extra few years in hopes of finding a way forward.

CFO Stacy Smith appeared on CNBC after the results were announced. The video is below.

This article originally appeared on Recode.net.

Sign up for the newsletter Sign up for The Weeds

Get our essential policy newsletter delivered Fridays.