Monday, May 02, 2005

Will Moore’s law guide Intel for the next 20 years?

Recently there was an unusual request on EBay (like many other items sold in this e-flea market) for a copy of April 19, 1965 edition of now defunct magazine called Electronics. Intel posted a reward of $10,000 for a mint condition of the original copy of this edition. In this edition, Gordon Moore, one of the Intel’s co-founders, famously observed that the number of components in a microchip (ICs) will double every two years. No other technology revolutionized the field of electronics like integrated circuits. Many modern day electronic gadgets owe their presence to the rapid miniaturization of integrated circuits. When the actual number of components in a microchip followed the trend in the first decade after Moore’s prediction, his observation came to be known as Moore’s law (thanks to one of Moore’s friend at Cal Tech).
Three years after this publication, Gordon Moore along with a colleague at Fairchild Semiconductor Robert Noyce started a company called Integrated Electronics, shortly called Intel. Away from the stifling bureaucracy of Fairchild, the young company started innovation in a big way. By 1971, Intel launched its first microprocessor. Till 1981, Intel was supplying microprocessors for non-PC uses such as traffic lights. Only in 1981 IBM chose Intel to supply microprocessor to power the IBM PC. In 1985, Intel made a strategic decision to exit the memory business, which then had become a commodity business, and to focus on its PC microprocessor development.
The above recap is pertinent to highlight that Intel faced two big decision points separated by a decade. First one in 1985 when it faced the onslaught of cheap memory imports. Since Intel had been working on microprocessors even before 1985, it had an alternative to fall back when it jettisoned the memory business. Intel came out of this restructuring smelling roses.
Second inflection point?
In the ensuing decade after 1985, Intel made relentless march in its process technology by doubling the number of components in an integrated circuit every two years and microprocessor performance increased commensurately. Moore’s law became a self-fulfilling prophecy. Intel’s sales grew at an annual average of 28% between 1985 and 1995. When growth rates started slowing down after this scorching pace, Intel for the first time in its history devised a strategy that did not leverage its process technology prowess. In 1997, in order to capture the low end of the PC market, it introduced Celeron processors. Instead of putting more components in the microprocessor, Intel removed features from its premium product Pentium and called it Celeron. It tried to capture whatever consumer surplus that was left on the table. It is during this time (mid 90s) that Intel missed a second inflection point to diversify its business.
Intel, instead of vigorously expand into other market segments such telecommunications and networking, kept its focus on the PC market throughout the 90s. Intel let companies like Cisco and Lucent become leaders in their respective markets. It can be argued that the engineering skill sets and process technology needed for design and manufacture of microprocessor is very different than that for making network routers and cell phones. But Intel could have acquired these complementary assets by buying companies such as Analog Devices and Conexant (a Rockwell spin off) and added analog products portfolio to serve the networking and communication markets. Intel had enough cash and rich stock price to make these acquisitions in mid 90s.
During this time frame (from 1995 to 1999), Intel was not really pushed to wall as it was in 1985. Intel was extremely profitable with net margins above 25% and annual sales growth of 10%. Hence it exhibited inertia similar to any successful company. It waited until 1999 to make big acquisitions in the networking market. Also in 1999, Intel changed its mission from being a PC-centric company to being a building block supplier to the internet economy. However, none of the acquisitions were successful. Five years after change in mission, Intel still derives 85% of its revenue and all of its operating profits from PC-centric (and enterprise computing) products.
New strategy – old wine in new bottle?
It is in this context, the arrival of new CEO Paul Otellini on May 18th should be seen. Self-proclaimed products guy, Paul was responsible for the creation of Celeron and Centrino brand. His “platformization” strategy derives its strength from the successful Centrino campaign. Intel branded a collection of chips in the laptop, such as microprocessor, chipset and wireless component as Centrino. In a recent interview to Fortune magazine, the incoming CEO talks about creating similar lucrative clusters of components in high tech gadgets other than PCs such as in home multi-media, hospital doctor’s room. This assumes that Intel will have compelling product offerings in the new growth areas similar to Pentium M microprocessor that was central to Centrino brand. It also assumes that the competition in the new growth areas will be as meek as in the PC arena.
We, the authors of this article, would like to take a different view of Intel’s near term prospects. The success of the Pentium and Centrino brand centered on Intel’s strong processor line-up. Intel was able to maintain its lead in microprocessors (barring recent slip-ups in dual-core and 64-bit processors) against AMD through its faithful adherence to Moore’s law and manufacturing prowess. But to extend its success in the PC arena to other areas such as multi-media and medical electronics, Intel needs strong analog and signal processing products. Intel currently has neither of them. In fact, in the same Fortune article, Texas Instruments chief dismisses Intel competition in TI’s product space.
Has Moore’s law reached its limits?
Also Intel (in fact the semiconductor industry as a whole) is facing technical issues in process technology. As microprocessors become faster and faster, they throw up lot of heat. As the size of the devices in a microprocessor become smaller, the devices leak a lot of power. Hence the cooling solutions have become costlier. Also there are physical limits to Moore’s law. The devices with dimensions 16nanometer or less (a nanometer is one-billionth of a meter – currently minimum length of devices in microprocessors is 90nm and approaching 65nm this year) are predicted to be unreliable. Hence Intel cannot maintain its dominant position in the PC-arena solely through faster microprocessors in another five years.

We, the authors of this article, recommend that the incoming CEO recognize the fact that Intel is no more a growth company it once was. We also recommend that Intel should use the $13billion cash to make acquisitions in areas outside of its core competency and also return a portion of cash to its shareholders.

Raj Subramanian, Sona Mooliyil, Kunal Shah


Post a Comment

<< Home