When there’s a massive potential market disruption, one looks for initial signs that all the investment and market vectors are about to push things in a completely new direction.
I’m not sure we’re quite there yet with regard to ARM chips in the data center, but Bloomberg Technology’s report yesterday that Google is thinking about building its own ARM based servers for the data center is in line with all of the other server trends that have been accumulating over the past 5 years–AMD’s acquisition of SeaMicro, the massive investment Calxeda has received, the focus on the power envelope of chips, the concentration of buying power in the hands of fewer and fewer IT giants.
The statement from Google on this question shouldn’t be comforting for Intel:
“We are actively engaged in designing the world’s best infrastructure,” said Liz Markman, a spokeswoman for Google. “This includes both hardware design (at all levels) and software design.”
At the end of the day it comes down to cost. Is it more cost effective for Google to build its own hardware, cut Intel out of the value chain, and produce chips that are better geared for Google’s specific needs? Google still needs a foundry which could, ironically, mean that it would have to have Intel build the chips or some other foundry like TSMC.
But the reality is that when you’re as big as Google and your buying power is that large, economies of scale begin to raise the question: Do I have the IP and the talent to build servers at a custom level that are optimal for my business?
4 percent of Intel’s revenue comes from Google, according to Bloomberg.