The Year (and Next Year) in Semiconductors

This has not been one of the great years for the economy, although it ended better than it started. Semiconductor companies started out the year in full retreat as spending just stopped dead in the fourth quarter of 2008. The true impact became clear across the industry when grim earnings reports hit in January 2009.

Semiconductors are a good leading indicator of the hardware industry's relative health, since all finished devices -- PCs, servers, printers, smartphones, storage systems -- use semiconductors as their base components.

Because of this, many market research firms and financial analysts monitor the chipmakers. If Dell is seeing increased demand for servers or a Chinese supplier is building more laptops, for instance, those firms will be buying more CPUs, memory, and other base components.

And as 2009 progressed, it was the chipmakers who first saw the beginning stages of the recovery. Intel CEO Paul Otellini was one of the first to declare a recovery had started as early as midyear because his company was receiving an increasing number of orders, especially in the month of June.

Many IT giants managed to stay profitable through the recession by cutting their way to profitability. Even Microsoft had its first layoffs in its 30-year history. One thing the major semiconductor players would not touch, though, was R&D. Otellini said the way to exit the recession was with "tomorrow's technology," not yesterday's.

As it turns out, though, the biggest shot in the arm for the hardware industry came from Microsoft. With Windows 7, Redmond created genuine excitement around an operating system release. Many critics said Win 7 was simply "Vista done right," but it was the "done right" part that evidently counted.

In the months leading up to Windows 7's debut, system builders increased their orders for parts in anticipation, helping Intel, AMD, and Nvidia exceed third-quarter projections. When the new OS launched on October 22, PC sales enjoyed a considerable spike, and now one report shows businesses are accelerating plans to roll out new PCs to their employees because of Windows 7.

Sundown

Of all the M&A (merger and acquisition) activity this year, the sad saga of Sun Microsystems provided the most headlines, not all of them good.

IBM reportedly planned to purchase Sun in March but the deal fell apart at the last minute for reasons still unknown. What is known is that Sun walked away, and ended up walking right into the arms of Oracle.

That proposed merger was announced in April and still isn't consummated due to protracted government investigations in the U.S. and then Europe. Finally, in December, the EU seemed to be convinced to let the deal proceed, although it's still not a done deal as of press time.

Through this eight-month ordeal, Sun lost staff, customers, and market share at an increasing pace. Its share of the SPARC server business has plunged precipitously as HP and IBM have vigorously pursued nervous Sun customers. Oracle chief Larry Ellison has pledged to stick with the SPARC line and stick to selling high-end servers like the Exadata 2 servers introduced in September.

While SPARC may survive overall, one future product did not: the processor codenamed "Rock." Never given a formal product name, Rock was killed off late in the year. The chip was behind schedule and under development for a long time but still never made it to market.

Sun still continues to look for ways to grow, thanks in part to plans to continue SPARC development with an iterative chip to its current Niagara processor line, codenamed "Rainbow Falls." All that's known at this point is that it is a 16-core processor.

GPU Rumble

In the graphics business, AMD's graphics arm ATI and archrival Nvidia found themselves with a common problem: their manufacturing partner, TSMC of Taiwan, was having a terrible time migrating to a new manufacturing process. ATI and Nvidia started the year with GPUs made from a 55-nanometer manufacturing process and wanted to go to 40nm, but TSMC was having a devil of a time with yields.

Yields, the number of usable chips from a wafer created in the manufacturing process, are one of the most closely guarded secrets in the industry, so no one knows for sure what was happening at TSMC. But rumors placed initial 40nm yields in single-digit percentages, clearly not enough for Nvidia and ATI, which both need millions of chips.

Nvidia took a little longer than usual to come out with a new architecture and ATI passed it, if the benchmarks are to be believed, with its Radeon HD 5800 series. It wasn't until October that Nvidia announced plans to strike back with Fermi, a processor with three billion transistors -- and a ship date that's slipped to the first quarter of 2010.

For Intel, its attempt at the discrete GPU market, "Larrabee," became its first high-profile failure in years. After five years of development, the company was forced to shelve the technology and said it would repurpose it elsewhere. All the learning from Larrabee would not go to waste, but any plans for a GPU to compete with ATI and Nvidia ended in December of this year.

Coming Up

So what will 2010 look like? Much of the semiconductor industry is betting on new technology to help drive its fortunes in the coming year. But for some of the biggest players, navigating rocky business issues will be just as critical.

Intel: Intel continues to execute with almost boring efficiency. The silicon giant has kept to its road map and has not deviated from it in years, which is admirable considering how much legal trouble it has had to deal with -- most recently stemming from the Federal Trade Commission's suit over Intel's competitive activities.

That trial is set for September, and Intel, which usually settles its cases, is probably hoping to settle this one. But Nathan Brookwood, research fellow with Insight 64 and a long-time semiconductor analyst, doesn't think the FTC will be as willing to strike a deal.

"Intel, I'm sure, would like to settle," he said. "I think the FTC probably wants to get something out of Intel and [am] not sure if Intel wants to go along with it."

It's The Dirk Meyer Show: AMD CEO Dirk Meyer has no one to blame or credit but himself now. He is firmly in charge of the company with a new executive team.

This year, AMD settled its long-running legal fight with Intel and has a huge chunk of change to show for it. Its fabs are now completely spun off, and ATI is going gangbusters. As a result, AMD will stand or fall on its own products now. Thus far, Wall Street seems pleased, with AMD's stock going from $2 to $10 this year. Who else can claim a fivefold ROI? If he executes on Fusion, 2010 will be a good year for the engineer-turned-CEO.

Brookwood thinks 2011 will be an even bigger year, however.

"They need to demonstrate they are making progress. I think 2011 will be more of an interesting year here for AMD. That's when we'll see the next generation," he said.

"When you look at the total platform, AMD is in pretty good shape. They have a better integrated graphics story than Intel, plus they have a GPU offering," Brookwood added.

ATI gains major ground: Nvidia has put its eggs in the basket of big silicon. The new Fermi chip is said to be huge and require considerable cooling. And therein lies the challenge: desktop PCs are at best stagnant, at worst a shrinking market. Mobile GPUs are going to be the new wave, and ATI has a better story in that area.

Going into 2010, Nvidia and ATI will be armed with powerful new processors capable of supporting DirectX 11 and, more important, supporting libraries like OpenCL and DirectCompute, which will let those GPUs speed up graphics-intensive applications ranging from the Windows 7 interface to Adobe Photoshop and Illustrator.

Right now, ATI is claiming numerous design wins and that it has passed Nvidia in mobile discrete GPUs, but it all depends on what comes out from Nvidia. If the next generation of Fermi-based GeForce cards (GeForce is the brand name for Nvidia's consumer GPUs) is an overheating dog, Nvidia will be facing some real problems.

The GPU guys have also found a new market in which to rumble: supercomputing. Nvidia has touted its GPUs as supercomputers, able to provide teraflops of computing power. However, it was ATI to score the first supercomputer on the Top 500 list with a Chinese system using thousands of ATI cards. Expect the next lists, due in June and November 2010, to feature more GPU-powered computers.

What won't happen to Nvidia: What could be truly fatal is if Nvidia chief Jen-Hsun Huang lets his enmity with Intel get out of hand. I've never, ever seen a company succeed when its CEO gets into a public war of words with a bigger rival. In fact, the opposite happens: The company hits the skids. Nvidia's track record of success can't be denied, but then again, Sun, Borland and SCO were successful companies for quite a while, too.

You can be assured of two things that won't happen in 2010: Nvidia won't go into the x86 market, and it will never merge with Intel. The acrimony is so deep that analyst Jon Peddie of Jon Peddie Research said there would likely be "a few homicides" if the two companies, just two or three miles apart on the same stretch of road in Santa Clara, Calif., were to merge.

But people will keep talking about it, Brookwood figures.

"No matter what the reality on the ground is, there will continue to be people who want to talk it up," he said.

In any case, it will be a transformative year for Nvidia's income statement as it shifts markets. Its chipset business is looking doomed. Brookwood feels there's still enough Core 2- and AMD-based platform traffic to keep it alive, but just go to a store such as Fry's Electronics or Micro Center and see how many pre-Nehalem desktop motherboard choices are available.

Fortunately for Nvidia, 2010 will see a growth in its high-performance computing business with Fermi and Tesla, and its Tegra business will finally take off. Expect to see netbooks, smartphones, mobile Internet devices and other gadgets using the latter chip, which combines Nvidia GPU technology with an ARM processor in a very low-power package.

A big rollout could begin as soon as CES next month. Plus, Nvidia has started sampling the Tegra 2, which is said to offer big improvements over the prior generation.

"So they do have some momentum and technical advantage with the Tegra line over alternate ARM designs," Brookwood said.

Storage, speed and memory: SSDs will continue their slow adoption by the enterprise, hampered more by the slow, cautious rollout of new hardware than anything else. This technology needs to prove itself over time and will not be rushed. Seagate will enter the market in a big way, giving validation to the concept of SSDs as replacements to 15,000 RPM drives.

Computers will get a lot faster. Between Nehalem, DDR3, new GPUs, SSDs, Windows 7, USB 3.0 and possibly Intel's fiber optics, PCs are going to become ridiculously fast and responsive.

A memory shortage will cause memory prices to rise throughout the year, and products will get more expensive as a result. The transition to DDR3 is expensive and many memory makers in Taiwan are so laden with debt that they can't afford the capital equipment to move to DDR3. That means the companies that can make the memory, like Samsung, Micron, and Toshiba, will profit handsomely as 2010 rolls along.