IBM Ponies Up $3 Billion For Advanced Chip Research
July 14, 2014 Alex Woodie
IBM last week committed to spending $3 billion over the next five years on fundamental research into the “post-silicon” generation of chips that will keep us close to a Moore’s Law performance track as processors shrink below the 7 nanometer threshold.
Say what you will about IBM, and many in the IBM i midrange community have. But you can’t argue that Big Blue has been shortsighted when it comes to researching and developing the chips that power its large servers and storage arrays. The company is by far the biggest patent holder in the world, and its annual R&D budget of $6 billion helps keep it at the forefront of the semiconductor industry. It may be shopping around its chip manufacturing business and foundries, but it’s also clearly still in the chip design business, and is looking beyond the 10-year horizon with its next-gen chip designs.
The $3 billion that IBM announced last week will go entirely into research–in particular, coming up with new and creative ways to build the computer chips of tomorrow after the techniques that are state-of-the-art today have been exhausted. IBM is currently building its Power8 chips with a 22 nanometer process, and its scientists have a clear path to get that down to 14 nanometers with the next generation of Power9 chips, and then to 10 nanometers.
But shrinking the circuits below the 7-nanometer level will require some creativity, whether it’s done on silicon or other substrates. IBM’s first research program as part of this investment will be aimed at the so-called “7 nanometer and beyond” silicon technology. Beyond silicon, big chunks of the $3 billion will go towards exploring entirely different materials, including quantum computing, neurosynaptic computing, silicon photonics, carbon nanotubes, III-V compounds (named after the columns in the periodic table), low power transistors, and grapheme.
IBM has already done extensive research into carbon nanotube transistors, which are single atomic sheets of carbon rolled up into a tube, and III-V compound semiconductors (named after the columns in the periodic table), but they haven’t been commercialized yet. Carbon nanotubes may hold the best hope; IBM recently demonstrated carbon nanotubes with 10 nanometer gates with no degradation due to scaling. Research is at earlier stages with the other materials.
One of the most exciting technologies is quantum computing, which utilizes qubit superpositioning that allows super-chilled computers to calculate millions of solutions at once. Neurosynaptic computing utilizes principles of nanoscience, neuroscience, and supercomputing in pursuit of a computer that emulate the human brain. Silicon photonics, meanwhile, uses light to transmit data between different components in a computing system. Graphene uses a layer of carbon one atom thick to move electrons up to 10x faster than in silicon and other conventional materials. Low power resistors, meanwhile, are designed to fight power leakage common in today’s higher voltage chips.
John Kelly, senior vice president for IBM Research, says the company is well-suited to the challenge of building next-gen chips that will power cloud, big data, and cognitive systems. “The question is not if we will introduce 7-nanometer technology into manufacturing, but rather how, when, and at what cost?” he says in an announcement.
Scaling to 7nm and beyond “is a terrific challenge, calling for deep physics competencies in processing nano materials affinities and characteristics,” says Richard Doherty, technology research director and The Envisioneering Group. “IBM is one of a very few companies who has repeatedly demonstrated this level of science and engineering expertise.”