Surround And Conquer

The processor wars are back in full swing, this time with some new players in the field. But what defines winning this time around is far less obvious than it was in the past, and it will take years before we know the outcome.

The strategy is the same, though, and it’s one that has been in use for years in the tech world. It began in the 1990s, when IBM came to the realization that it could no longer own the computing market the way it did in the previous decades. That was largely the work of the U.S. Department of Justice, which in 1956 forced IBM to sign a consent decree that barred it from bundling software, services and hardware together in a “market basket.” Three decades later, in the height of the PC era, IBM was struggling to compete under the terms of that consent decree.

At the time, IBM was engaged in an ugly confrontation with Microsoft. IBM had its OS/2 desktop operating system, which was losing out to Microsoft’s Windows 3.0, particularly after Microsoft very publicly pulled the plug on their joint development agreement. Rather than fight a losing head-on battle with Microsoft, IBM began informing customers that Windows was just one operating system among many and that it would support them all. And shortly afterward, it began reducing its reliance on its proprietary version of Unix, called AIX, and embracing Linux. Within several years, IBM was supporting nearly a half-dozen OSes, and Microsoft was helping to sell its hardware.

This “surround and conquer” approach is a common military strategy, and it has been used since the time of Alexander the Great. Applying this strategy to business is a more recent development because it is a way of avoiding government scrutiny over monopolies. The idea is that if only one company makes money, then it’s a monopoly. But if lots of companies make money, even if the same company pulls in the lion’s share of profits, it’s harder to call that monopolistic behavior.

Fast forward to today, and processor companies’ biggest customers are now competing for the same market with an inherent advantage—they are the consumers of their own custom designs. But there is still money to be made in providing chips that live alongside those in-house designs, and the way to win in this market has moved up a notch in abstraction. It’s not just about the chips. It’s about the architecture for connecting data, and the best way to achieve that is to embrace the fastest processors and interconnects, no matter who makes them.

Chipmakers are approaching this from the technology side, and the big cloud providers—Amazon, Google, Facebook, Microsoft and Alibaba—are approaching it from the data side and building architectures to optimize processing of that data. Both of them will meet at the edge in what is almost certain to be a war zone for years to come.

Signs of war are already on the horizon. Companies are competing for soldiers. Cloud vendors can be spotted at almost every major chip design conference and trade show these days, talking about their architectures but providing almost no details about how these systems really work. Instead, they hand out business cards for people looking for further details. But the tactic is transparent. They’re there to raise awareness and to hire experienced architects and chip engineers, who are in short supply everywhere.

What’s needed to make this strategy work is an army of qualified engineers and data scientists. They will be needed to build the customizable heterogeneous architectures and the infrastructure needed to process and store vast amounts of data.

But looked at from a different vantage point, it’s no longer about designing a single chip. It’s about designing a system of chips, no matter who makes them. Big chipmakers and big cloud providers are trying to outmaneuver each other with a surround-and-conquer approach, and at this point it’s too early to tell who’s surrounding whom or how big the final prize will be.