Before founding Intel in the 1960s, Gordon Moore made a bold prediction about the exponential growth in the number of components on integrated circuits, which has been proven remarkably accurate by subsequent history and immortalized under the title of Moore's Law. Intel has been doggedly upholding Moore's Law by roughly doubling the number of transistors in its processors every couple of years, but now that schedule is starting to slip.
For a number of years now, Intel has been adhering to a tick-tock strategy in improving its chips: the tick is a shrinking of the manufacturing process, while the tock is an optimization of the design and architecture at a given size. The last tick was the move to 14nm with the current Broadwell CPUs, and the next tock is the Skylake series of chips expected later in the year — but the usual rhythm will be broken in 2016 with a third 14nm part in the form of the newly announced Kaby Lake processor. That bumps Intel's move to 10nm production into the second half of 2017.
"Our cadence today is closer to 2.5 years than two."
During a conference call to discuss Intel's latest earnings, CEO Brian Krzanich explained that "the last two technology transitions have signaled that our cadence today is closer to 2.5 years than two." Manufacturing advances haven't progressed at the same pace as before, which is why the company has now seen it necessary to inject the Kaby Lake model into next year and is holding off on Cannonlake, the 10nm, chip, until later into the decade. At the 10nm size, Intel is approaching the limits of what can be done with silicon, and any further development beyond it will probably require different materials, such as the silicon-germanium alloy that IBM used to demonstrate a processor built at 7nm recently.
Krzanich is confident that Moore's Law is still alive and well, and cites Intel's recent history as having "disproved the death of Moore's Law predictions many times over." And yet, Intel's decelerated roadmap suggests that the extraordinary pace of improvement in the chip industry is starting to slow down. In any case, if anyone's going to keep Moore's Law ticking along at its established two-year rhythm, it won't be Intel.