Has Intel crushed AMD?

December 12, 2007, 4:34 PM UTC

The scrappy chipmaker has plenty of life left – but mistakes have cost it dearly.

AMD’s manufacturing facility in Dresden isn’t yet producing enough quad-core chips to boost the bottom line. Photo: Sven Doering/AMD

If you’d like to beat up on Advanced Micro Devices CEO Hector Ruiz, now would appear to be a good time. Ruiz has won praise for helping the chipmaker mature into a worthy challenger to industry heavyweight Intel, but as he prepares for a Thursday meeting with Wall Street analysts, AMD has the look of a well-used punching bag.

As a result, AMD (AMD) is bleeding cash — the company reported a loss of $396 million last quarter alone. That puts AMD in a tricky financial position. The company has more than $5 billion in debt, much of it from the acquisition of graphics chipmaker ATI, but it it needs cash both to finance future projects and to pay for earlier acquisitions. (The ATI purchase is still causing headaches; AMD today said it will take a significant writedown from the $5.4 billion transaction, a possible sign it overpaid.)

AMD got a boost from the original Opteron server chip and other advances, but Intel has come back with a vengeance

That’s not all the bad news. AMD’s answer to the Intel onslaught, its much-hyped Barcelona server chip, is plagued by glitches and the company hasn’t managed to ship it in time to save the day. Word of the latest setback came last week — AMD now says the chip that was originally due this summer won’t reach most customers until just before spring 2008. There are even rumors that Ruiz could face pressure to step aside.

AMD’s success or failure will have a broad impact on the computer industry. Though Intel dwarfs the company, AMD has produced innovative designs that have given PC makers more options. Perhaps most important, AMD gets credit for keeping Intel from getting too complacent, spurring both companies to produce better technology.

But all is not lost for AMD or its investors. True, Ruiz had recently built the company’s reputation on technology that business customers use for back-end computing tasks, and AMD executives bragged too soon that Barcelona would leave Intel in the dust, leaving egg on their faces. But there are still plenty of other areas where AMD can hope to make money. For example, industry observers are buzzing that the company seems to be winning back customers to its graphics processors. And if AMD continues to gain market share in the fast-growing laptop segment, it will go a long way.

“If you plot our trajectory from Q1 all the way through Q3 in our public numbers, every quarter we have done a little bit better than the quarter before,” says Mario Rivas, AMD executive vice president of the computing products group, which handles processor development. “We have better average selling prices, better margins — still in the red, but now we are very close to the black.”

Analysts agree. “AMD has actually gained a lot of share, a lot of acceptance,” says ThinkEquity Partners analyst Robert Burleson. “I don’t think it’s a disaster for AMD. Certainly they need to recover.”

At this week’s analyst meeting, however, Ruiz is sure to field plenty of questions about servers and Barcelona. For AMD to have a fighting chance against a resurgent Intel, it will have to prove it can fix these sorts of problems. So what went wrong?

Pride Before a Fall

For starters, it’s clear that AMD’s top bosses drank too much of their own Kool-Aid. At the same time that they over-estimated the impact Barcelona would have, they under-estimated Intel’s ability to respond with its own server chip improvements. Gordon Haff, principal IT advisor for consulting firm Illuminata, says that cocksure stance was on full display. “Execution aside, one of the problems with Barcelona is that they set up this expectation that Barcelona was just going to whip Intel,” Haff says. “As recently as mid-to-late summer, that was the kind of message they would bring out: When this chip’s out there, Intel’s just going to have to duck and cover.”

It’s easy to understand how Ruiz & Co. could get a bit overconfident. Just three years ago, AMD was pumping out profits and its stock was rocketing upward. A server chip code-named Hammer had busted open the corporate market for AMD. Rather than focus too much on raw speed, AMD’s chip designers made sure Hammer would be energy efficient and would do a good job running both older software and newer 64-bit applications.

A financial edge from edgy PC design (Photos 1-7)

AMD caught Intel flat-footed. Not only did the company win sales with the Hammer chip, which it marketed as Opteron, but it also got something more valuable: respect. No longer did computer makers view AMD products as just a last-ditch option when they couldn’t afford Intel chips. AMD was an innovator. The customer roster reflected the change. Though only IBM (IBM) had signed up with AMD to use Opteron when it launched, AMD was soon selling chips to every major server shop except Dell (DELL).

Even better, a halo effect from Opteron spread to AMD’s other products. The same PC makers who flocked to Opteron for their servers began using more AMD chips in their desktops and laptops too, boosting the company’s market share and revenues.

So in 2004, as AMD’s executive strategy committee including CEO Ruiz and President Dirk Meyer planned the next act for Opteron, they were thinking big. The future, they knew, was in multi-core processors that have more than one computing brain on a single chip. The next server battle would be won or lost, they decided, with four-core processors – so they would try to design a quad-core chip more elegant than the one Intel was bringing to market.

The Barcelona Gamble

While Intel planned to reach quad-core quickly by taking two dual-core chips and tying them together, AMD would instead try to leapfrog Intel with a strategy it calls “true quad-core,” where all four cores were designed to work together optimally and efficiently. Sure, the complexity of the Barcelona design meant Intel could claim quad-core bragging rights first, with its dual-core compromise. So AMD would have to hope customers would wait a little longer to check out Barcelona, and count on AMD to deliver another Hammer-like breakthrough.

By mid-2007 though, things clearly weren’t working out as well as AMD executives had hoped. An AMD team in Texas was working with customers to run tests on the chip when the design flaw surfaced. After an internal panel reviewed the glitch and determined that it could possibly afflict real-world customers, they decided to take the chip back to the drawing board, Rivas says. “The design itself, intrinsically, is a good design,” he says. “We will get silicon in early January. We will take it through its paces, we will test it, and start delivering samples to our partners.”

New design in HP’s business displays (Photos 1-5)

Delays weren’t the only problem. AMD also announced that the first versions of Barcelona would run more slowly than expected. The rationale was that most server customers don’t buy the fastest chips, anyway; they prefer slower, more affordable chips that run more efficiently. But with AMD’s belated chip expected at speeds slower than Intel’s available offerings, Barcelona was looking much less like an Intel killer. Last week’s news of further delays were the third blow; AMD executives now say Barcelona won’t ship in healthy volumes until the end of the first quarter, when it will be nearly a year late by some measures.

“Barcelona is the most complex x86 processor ever designed,” says AMD’s Rivas. “We encountered some unexpected challenges when we were wrapping up the part in the middle of November. … Unfortunately it requires another revision of the silicon.”

Rivas says the revised Barcelona design should be available in mid-January, and the chips will ship in volume at by spring 2008.

Meanwhile, Intel has taken full advantage of AMD’s gaffes. In November, the chip giant announced an updated version of its quad-core Xeon, dubbed Penryn, that’s made with an advanced 45-nanometer manufacturing process. (AMD’s 45-nanometer Barcelona chip isn’t expected until late next year.) Intel’s competitive move weakens the chances that AMD will pull in big profits when the first big volumes of Barcelona chips finally arrive.

Two Competitors

Despite AMD’s Barcelona headaches, much of the server market seems to be rooting for AMD.

“I think AMD’s proven themselves as being an enterprise-class product — if they’d had this kind of glitch before they had proven themselves, it would be far more detrimental,” said David Driggers, co-founder and chief technology officer at Verari Systems, a specialty server maker in San Diego. “People are cutting them a bit of slack. I think most of the marketplace knows that to have really healthy technology, we need two competitors, not just one. I think even Intel knows that now. They’re sharper and stronger now because of AMD.”

In the laptop and desktop market, where trends will ultimately decide AMD’s fate, customers are taking a similarly patient tone. AMD is gaining market share in consumer laptops, but still has a long way to go in the higher-end business laptops that command a premium and deliver profits. A new AMD laptop processor due in 2008 should help the company make gains with smaller businesses, and another in 2009, code-named Puma, could lure larger businesses.

Philip McKinney, Hewlett-Packard’s (HPQ) chief technology officer for its PC division, said the major hurdle for AMD to clear in the business laptop market is power efficiency.

“If you look at the product line today and you stack up Intel against AMD on a chipset basis, the AMD chipsets are still at a negative power impact. In the enterprise market, that’s the number-one criterion for selection – it’s battery life,” McKinney says. “But the current roadmap at AMD we think is very, very strong.”

If AMD can deliver the planned laptop chips on time, we might see a day when these Barcelona woes seem like a bad dream. For AMD executives, and scarred investors, that day can’t come too soon.