Look Who’s Making Chips

Newcomers to the semiconductor world have an opportunity to make lasting changes, but will they?

popularity

The entry into the chip business by companies such as Apple, and possibly Google, Amazon and a handful of others, may seem like a land grab in the semiconductor world, but the reality is that system companies have always done their own semiconductor design. Only the names have changed.

IBM made its own PC processors, and it still makes them for its high-end servers. HP made chips for its PCs and printers—so many, in fact, that HP executives used to brag that the company was the largest semiconductor manufacturer in the world. Cisco made chips for its networking equipment.

This is hardly a new trend. But what is new is that big systems companies are now building systems on chip, and some of those big systems companies have never built their own semiconductors before. The advantage is they can co-develop the hardware and the software, and eke out both performance improvements and power savings while controlling costs.

It’s unlikely this will continue forever. The economics of designing, building, integrating and debugging chips versus designing them and handing off the details and integration to third-party chipmakers don’t work. Efficiency is in the volume and the expertise that comes with making lots of chips for multiple markets and understanding what goes together best. If you only make one or two chips a year for a single platform, you don’t ever gain the level of expertise necessary to make the necessary tradeoffs that improve performance, lower power and reduce costs enough to make it all worthwhile.

This is all rather opaque to the outside world, particularly when earnings at these systems companies are astronomical and end product prices are not under pressure. But add in some competition, maturing platforms, market shifts and saturation—exactly what’s happening now in the smart phone market—and financial analysts begin beating up on these very public companies. The flatter the earnings numbers, the harder the punches.

For the newcomers who are now building chips, having almost unlimited resources is like giving new toys to kids and telling them to go build something interesting. The big question is whether they have the processes and methodologies in place to really do that—and interesting in terms of what it means for many companies, not just one product.

One of the great legacies of companies such as IBM, Xerox PARC, Bell Labs, and even Intel, Texas Instruments and Toyota on manufacturing (despite the recent bad press) is that they infused engineering discipline into the toy box and figured out ways to solve problems for an entire industry. Decades later, we are all beneficiaries of those efforts.

It remains to be seen just how much of lasting value the newcomers bring to the table of value that can grow a multitude of industries for many years to come. There’s a sense of responsibility that goes along with having megabucks for R&D. So will this next period of development provide real breakthroughs, or will it simply be a squandered opportunity? It will be years before we know the answer to that question. It takes an industry and lots of hard work over time to make significant changes, and opportunities have to be nurtured and recognized every step of the way.



  • Barbara Kalkis

    Sounds like you read my 26 January blog posted on the Fab Owners Association website (www.waferfabs.org). We will continue to see companies like Google and others outside the classic sector known as “semiconductor company” designing chips. And why not? Design is the ticket to superior performance, silicon efficiency, IP protection, cost, The impact will hit fabless companies and the IDMs (integrated device manufacturers). I maintain that IDMs can come out winners if they hone their manufacturing expertise to offer foundry services. Altera and Intel announced such a deal the other day. It promises to become an alluring model – one that can fuel regrowth of IDMs and fill the fabs while waiting for EUV to come about.