The PC Gets Overheated

     Before long, cracks began to appear. CPUs had advanced to run at speeds of a gigahertz-plus, 25x their rate from my Hong Kong days, but it couldn’t last.

Gelsinger gave an IDF talk about a sharp curve ahead in the roadmap. Chips can’t keep getting smaller, faster, cheaper. Or at least they couldn’t keep getting faster. They were getting too hot. On their current trajectory he said, showing a chart, they would give off as much heat as the surface of the sun in a couple decades.

It wasn’t the end, of course. CPUs switched to becoming silicon buckets of multiple small processor cores working together, like a network of tiny computers on a chip. That and many other advances continue to keep semiconductors advancing every year or two, though now at a statelier pace.

Over time, the friction and heat turned up on the business side, too. It became increasingly clear that –if at all possible -- Intel would quickly absorb any new PC hardware functions into CPUs and their chip sets surrounding them. Likewise, Microsoft expanded to become the leading provider of Windows applications, curbing opportunities for several software companies, like Manzi’s Lotus, a pioneer of digital spreadsheets.

The PC standards themselves sometimes became contentious as system makers clashed with Wintel over who would define the next technology.

For example, I wrote dozens of stories about two conflicting initiatives for how to upgrade PCI, the main bus connecting chips and boards inside PCs. To this day, I’m not sure what inspired the conflict I spent so much ink on. I suspect at least part of it was about who would own the intellectual property, the key patents, on the standard.

In an effort to drive the industry forward faster, Intel evolved from rolling out focused tech initiatives to defining – and manufacturing -- whole platforms, putting itself into direct competition with its PC-making customers like Compaq, Dell, HP and IBM. On one hand, it was an affront. On the other hand, it made sense: If PCs were going to play videos and handle phone calls in a smooth way, you had to carefully craft the entire system.

Stories began emerging about a big but still-secret motherboard factory Intel ran outside Portland, Oregon. The stories caught me flat footed. I talked to every source I knew, followed every lead. Others got the scoop, but I decided I would write a definitive article on it.

I spent several weeks working the phones to piece together a profile of Intel’s system group. I eked out details of the group’s org chart and its product portfolio and gathered plenty of opinions. To put on the finishing touches, I travelled to Hillsboro, Oregon, where Intel was making the PC boards. I walked around Intel buildings, peeking in windows to get a sense of the operations. Intel wouldn’t talk to me on the record, so I stood in its Jones Farm parking lot at shift changes, handing out my business card, asking folks to call if they were willing to be interviewed.

No one did. One employee turned in my card to a manager who sent it to Intel PR execs who quickly called Wallace. “Do you know what your reporter is doing?” they asked him, enraged.

Wallace treated them respectfully, but inside he was laughing with glee.

Ultimately, Intel granted me an interview. I finished the piece and made it a cover story of OEM Magazine. It won the 1996 Jesse Neal award for investigative reporting, a big deal in the trade press.

Weitzner hired a limo to carry us to Manhattan for the awards dinner. I felt both elated and uncomfortable with the attention.

OEM Magazine cover story, September 1995

Eventually, Intel scaled back its systems group, not because of my story or media attention in general, I think, so much as decisions to focus on standards and semiconductors and let Taiwan Inc. make motherboards, a business notorious for its use of hazardous chemicals and pennies-per-unit profitability.

Intel found other ways to promote whole systems designs. Rather than take up the low-profit work of manufacturing, it learned it could define fresh concepts, create market demand for them and set up logo programs so PC makers would essentially be forced to follow its lead. Its most successful effort was Centrino, one of the first Wi-Fi connected laptops. Intel used a Centrino logo to identify systems that used its wireless modules in addition to its CPU and a slimmed down chip set.

Centrino was the brainchild of Paul Otellini (left), a successful Intel salesman who became a group manager. Perhaps as important as the Centrino hardware Otellini’s team defined, Intel also pumped money into marketing the concept of retailers using Wi-Fi to sell internet access as part of their business model. 

In the days of dial up connections, it was a big bet. No one knew how or whether it would take off. IBM struck a deal with McDonalds to put Wi-Fi in thousands of hamburger joints, one of many efforts that grabbed headlines but misfired in the market.

But thanks to Intel’s vision and big bucks – and a bazillion cafes trying to attract customers --Centrino was a hit. Otellini got rewarded with the CEO job, leapfrogging Gelsinger and becoming the first non-engineer to lead Intel.

At EE Times, Otellini’s promotion was another brick in a wall of aversion to the company. Years earlier, Intel had shunned the newspaper’s engineering readers, taking its “Intel Inside” advertising campaign directly to consumer tech publications. Intel’s ad budget was probably the biggest of any semiconductor company by far in those days, and EE Times got very little of it. That didn’t stop us from covering the company’s every move, but it no doubt irked the ad department.

Otellini, as much as anyone in tech, helped popularize the concept of platforms, super products that spawned small industries that built other products around them. Every new generation of Intel processors became not just CPUs but platforms, a collection of mainly Intel chips and Intel-backed technologies and standards.

Before long, the speed and amount of memory you could connect to a CPU became a bottleneck for system performance. Intel struck a partnership with a startup called Rambus that was designing a new type of memory. It was a hot story. The PC was a huge user of memory chips. If Intel insisted on the new technology, giant memory chip companies might become like PC makers, metal benders cranking out products that made royalties for Rambus.

About this time, I got the chance to interview Otellini after he gave a keynote. Wallace joined us. I hammered the Intel exec hard from every angle I could think of about the Rambus deal. He didn’t give an inch. Finally, perturbed, he smacked the table and said he wouldn’t respond to another question on the topic. I gulped and moved on to other, less sexy topics.

Ultimately, Intel did not insist upon the Rambus memories, other solutions emerged. But the climate had soured.

Many folks rallied behind Intel archrival AMD or startups and their x86-like CPUs. But the startups didn’t get much traction. They and their customers risked being sued out of existence; they lacked a license to Intel’s intellectual property. AMD had a license and good chip designs, but stumbled many times trying to manufacture them in high volumes.

As Wintel’s dominance grew, rumors started flying about antitrust cases. Without solid sources in government and legal circles, I found myself playing catch up again. Somewhere along the way I got a tip from a PC maker about one small but significant piece of the story -- market development funds aka MDFs.

Both Intel and Microsoft were said to use MDFs. They had become a standard business practice in the industry. It was a simple arrangement: Folks who used Wintel’s logos for things like Centrino and other initiatives on their products and in their marketing materials would get rebates. I was told the rebates could amount to millions of dollars. Often the MDFs arrived in the last few days before the end of a financial quarter, and they might be large enough to affect whether the company’s results were in the black, given PC makers often ran on low single-digit profit margins.

I had a story, but of course no named sources. In the flood of antitrust reporting, it amounted to just another interesting anecdote.

Folks generally played along with the Wintel agenda to stay in business. Even then, many went bankrupt or were bought up as the industry consolidated to a handful of giant PC builders. The joke was that they lose money on every unit but make it up in volume.

Years later, I snagged an interview with David House (below), a former Intel executive, in his home in the Saratoga hills overlooking Silicon Valley. The avid helicopter-skier bragged about Intel’s aggressive management style and its hard-ball tactics.

He recounted his role in creating the company’s Intel Inside branding campaign under which it would share the cost of advertising for OEMs that used the new logo. "The Taiwanese [computer makers] loved that they could put Intel Inside on their systems, but Compaq and IBM hated it because it was taking away from their brand,” he said.

To get them on board, "I took a check of what they would have accrued [in the program] two weeks before the end of their quarter and I signed it right in front of them," he recalled. "I said this could be on your P&L or mine.

"They said they couldn't do it -- the IBM check was for, like, $125 million, and the margins on PCs by then were already razor thin," he said, recalling how he then tore the check up in front of them. "About three weeks into the next quarter, they signed up," House said.

Next: A lesson in ethics