Andy Grove's Big Blunder

Back in 1992 Intel had a choice; Digital Equipment Corporation's (DEC, now part of HP) senior vice-president of operations, John F. Smith, was peddling DEC's hottest property, the Alpha chip, to Andy Grove. Andy Grove rejected DEC's offer on grounds that Alpha required substantial investment in support infrastructure. Intel decided to go with HP's then non-existent EPIC technology.

As events will prove later, the decision to go with HP was one of the most disastrous decisions ever taken by any CEO (John Sculley included). Almost every single one of Intel's current woes can be traced back to that fateful decision.

Grove's reasoning to reject Alpha was akin to suggesting that the best bread on offer is no good because it isn't sliced. Grove's alternative was to go to the market, buy wheat, grind it to make flour, and then bake bread.

The original objection to Alpha is even more incomprehensible in light of Andy Grove's subsequent actions. In 1997, Intel bought DEC's only fab for $700 million, and even committed to manufacturing Alpha for DEC. The $700 million price paid was the book value of the fab and not the market value. In essence, Intel overpaid for something it did not need.

With an Alpha deal in 1992, Intel could have deployed Alpha within 2-3 years. Microsoft was just coming out with Windows NT and the company did make an Alpha port of Windows NT. Even if Microsoft had tried to stymie Intel, it wouldn't have worked. Windows NT contained code derived from VMS, and this could have been used as leverage. Andy Grove either knew this was the case or should have known; it was his job to get familiar with Microsoft technologies.

DEC's FX!32 technology would have allowed Intel to painlessly migrate to the Alpha architecture. FX!32 offered much better x86 emulation performance than the best Itaniums have got. FX!32 did not exist as a product when Grove made the decision to reject Alpha, so it is hard to blame Grove for that. However, DEC had an exceptional record as a technology company, and HP was not quite in the same league as DEC.

The worse consequence of Grove's decision has been the rise of AMD. AMD's Opteron processor has wrested market leadership position from Intel. Moving quickly to Alpha would never have allowed AMD time to bring a competitive offering to the table. Worse yet, the Athlon and the Opteron processors have been designed by members of the Alpha design team.

Intel should have finished its transition to a 64 bit ISA with the release of Windows XP. Unfortunately, Intel has not even started the transition. Persistent delays in the introduction of Itanium have continuously degraded Intel's prospects. The price war with AMD has killed any remaining prospects for consumer versions of Itanium.

The true extent of the damage done will take years to become clear. The damage is not limited to the revenue AMD is stealing from Intel, but also the revenue Intel is losing due to loss of its monopoly position. For every dollar AMD's processor division earns in revenue, Intel loses multiple dollars in profits.

Why did Andy Grove make such a poor judgement and was he alone in making the decision? Grove most likely wasn't alone in making the decision and the rest of Intel management is to blame as well. As to why Andy Grove did what he did, one can only speculate, but paranoia it wasn't. A paranoid person in a war zone will acquire the best flak jacket available for whatever it costs. He won't order a custom jacket from overseas to be delivered by surface post. Alpha existed and was the best available at the time. If Grove was paranoid he would have gone with Alpha. The answer more consistent with Andy Grove's actions is megalomania.

Years of flattering stories in business magazines had convinced Intel management that Intel was the smartest player in the microprocessor market. Intel management believed the company was capable of beating the best, but had unfortunately been locked into the x86 architecture due to IBM's stupidity. They wanted to teach the rest of the world a lesson or two in microprocessor design.

Market conditions only required Intel to come up with an architecture that was competitive with leading RISC architectures of the time; even Alpha was an overkill. Intel though was aiming to out do everyone. By going the way of other RISC processors, Intel could only hope to offer an also ran entry. Intel tried to outsmart everyone else by betting that future improvements in cpu performance will depend on the ability of smart compilers to extract parallelism from code.

Alas, it was the wrong bet; smart compilers never materialized. Moreover, Intel's design capability was not as legendary as the business magazines made it to be. When Intel got tested by AMD it was found to be lacking. AMD's research and development budget is paltry compared to Intel's, but it has been good enough to beat Intel designs.

Andy Grove might be right that only the paranoid survive, but he himself wasn't paranoid enough. His lack of paranoia has already cost Intel billions in lost profits, and the future costs look even more staggering.

by Usman Latif  [Dec 06, 2003]


Why the fastest chip didn't win
Intel, DEC Settle Alpha Chip Dispute
Alpha Learns to Do Windows
How Intel Wrecked Itanium
Itanium to take on Alpha influence
Intel hyperthreading shows Digital roots