IBM Bets $3B That the Silicon Microchip Is Becoming Obsolete


IBM's Silicon Nanophotonics chip technology.

IBM’s Silicon Nanophotonics chip technology. IBM



IBM is worried that the age of silicon may be drawing to a close. So it’s going to spend $3 billion over the next half-decade to try and find new ways to power the future generations of microprocessors.


“We really do see the clock ticking on silicon,” says Tom Rosamilia, Senior Vice President of IBM Systems & Technology Group.


Today’s state-of-the-art IBM chips use silicon components that are already tiny — just 22 nanometers in width. But looking about five years into the future, parts become so small that it becomes extremely difficult to maintain a reliable on or off state. “As we get into the 7 nanometer timeframe, things really begin to taper off,” Rosamilia says.


So the first step of IBM’s $3 billion quest will fund research into ways of making these smaller chip components work — even if they don’t use silicon. IBM has high hopes for a silicon alternative, called carbon nanotubes, but the concept still needs work if it’s going to become as easy to crank out carbon nanotube chips as their silicon alternatives. Another promising area is silicon nanophotonics: a way of using light instead of electrical signals to send data around the chip.


The microprocessor industry has had an amazing ride, but already you can some fraying at the edges of Moore’s Law


Beyond that, IBM is also investing in brand new ways of number-crunching—quantum computing, or Neurosynaptic chips — that use computing models that go beyond the digital computing paradigm that has dominated the tech industry for fifty years.


The microprocessor industry has had an amazing ride, but already you can some fraying at the edges of Moore’s Law—the observation that chip performance doubles ever two years or so.


Microsoft is starting to use custom-designed chips for its Bing search engine — they say that it may be better to get performance boosts this way, rather than relying on general-purpose server chips to speed up. And scientists who maintain a list of the world’s top supercomputers worry that today’s supercomputers aren’t becoming powerful at quite the same clip.


If you ask the world’s largest chipmaker, Intel, the company will tell you that Moore’s Law is alive and well, and that it expects to crank out faster and faster chips for the foreseeable future.


So why the disconnect? Most likely, it comes down to the different business models of Intel and IBM. Although both companies make their own microprocessors, Intel rules the high-volume desktop market. It sells server chips too, but at higher volumes than IBM. Big Blue, on the other hand, is looking to carve out profitable niches for its systems, taking advantages of its expertise in software and system design to build unique systems. It’s looking for competitive differentiation—not necessarily high-volume sales.


And if quantum computers, carbon nanotubes, and computers modeled on the human brain are what it takes to get there, then IBM may be buying itself a $3 billion head start.



No comments:

Post a Comment