A History of Silicon ValleyTable of Contents | Timeline of Silicon Valley | A photographic tour
History pages | Editor | Correspondence
Purchase the book
These are excerpts from Piero Scaruffi's book
The Selfies (2011-16)click here for the other sections of this chapter
SemiconductorsIn the age of the smartphone app people almost forgot why this region was called "Silicon" Valley. However, the semiconductor industry was still very much at the core of the valley, except that a process of consolidation was rapidly changing its shape. In 2015 Intel acquired Altera (specialized in semiconductors for communications), its biggest-ever acquisition, one year after buying LSI's networking business from Avago; and in 2015 Avago (that had been founded in 2005 over the ruins of a 44-years-old division of Hewlett-Packard and Agilent Technologies) acquired Broadcomm, besides CyOptics and Javelin (meanwhile it had acquired LSI and then sold LSI's networking business to Intel).
By then the hardware industry had embraced the multi-core chip, the only way to keep Moore's Law working. Intel's Xeon Haswell-EP of 2015 boasted 5.5 billion transistors but thanks to 18 cores. The cost per transistor had actually been rising since the Taiwan Semiconductor Manufacturing Company (TSMC) had introduced its 28-nanometer chips in 2011. In fact, in 2012 Intel started using a different kind of transistor, the "tri-gate" transistor. The rest of the world called it "FinFet" transistor. Chenming Hu had invented them at UC Berkeley in 1998, and one of his students, Yang-Kyu Choi, had founded the Nanotech lab at the Korea Advanced Institute of Science and Technology (KAIST) that was setting one record after the other in FinFet technology. In 2014 Intel started shipping the 14nm Skylake processor (400,000 times more powerful than the Intel 4004), but in 2015 Intel announced that its 10nm Cannonlake processor would be delayed to 2017. The "nanometer" scale (the separation of the transistors on the chip) was becoming impractical. The first microprocessor, the Intel 4004, had contained 2,300 transistors spaced by 10,000nm gaps. It was just getting too difficult and too expensive to operate at that scale. A Skylake transistor was made of about 100 atoms. With the same technology a 2nm transistor would be just 10 atoms wide. This was technically feasible (in fact, Yang-Kyu Choi's team at the KAIST had already built a 3nm FinFET in 2006), but extremely expensive. The cost of building a factory for microprocessors was already in the billions of dollars. In fact, in 2016 Intel's vicepresident Bill Holt openly admitted that Intel was not planning to use silicon below the 7nm threshold. Then Silicon Valley would stop being "silicon".
Consumer electronics was putting pressure on the memory industry. In particular, the price gap between hard-disk drives and solid-state drives (used mainly for NAND flash memory) kept shrinking. In 2013 Samsung announced a three-dimensional Vertical NAND (or 3D V-NAND) flash memory. In 2014 SanDisk and Toshiba opened a fabrication plant specialized in 3D VNAND in Japan, whose first product, released in 2015, was a flash chip that doubled the capacity of the densest memory chips. In 2015 Samsung unveiled the world's largest storage device, a 16-terabyte 3D NAND solid-state memory; while Intel teamed up with Micron to introduce its 3D-memory chip. 3D memory represented the first new class of memory in a quarter century. Micron and Intel joined forces to manufacture NAND flash memories back in 2005, and in 2010 Micron had acquired Swiss-Italian flash-memory maker Numonyx from Intel.
Quantum computing was developed far from Silicon Valley, in Canada by D-Wave, which popularized the technology, in upstate New York by IBM, which in 2017 delivered a 17-qubit prototype processor, in New Jersey by Bell Labs, where pioneers of quantum computing algorithms such as Peter Shor and Luv Grover now worked for Nokia, in the Netherlands by QuTech, the quantum research institute of Delft University, possibly the largest quantum technology institute worldwide. The center of mass began to shift in 2013, when Google acquired a D-Wave quantum computer and set up the Quantum Artificial Intelligence lab (QuAIL), led by Hartmut Neven and hosted by NASA the its Ames Research Center in Mountain View. Then in 2018 Google announced that it had built its own quantum processor, codenamed Bristlecone, a 72-qubit processor, in collaboration with John Martinis' group at UC Santa Barbara. By that time IBM boasted an experimental 50-qubit processor, codenamed Q System One, available to selected customers via its cloud. In 2015 Intel had entered into a research agreement with QuTech, and in 2018 announced a 49-qubit superconducting chip, codename Tangle Lake. The first major startup of the Bay Area was Rigetti Quantum Computing, founded in 2013 in Berkeley by former IBM physicist Chad Rigetti, followed in 2014 by QC Ware, founded in Palo Alto by Matt Johnson and others. PsiQ was founded in 2019 in Palo Alto by Jeremy O'Brien, a physicist at the University of Bristol, and Terry Rudolph, a professor at Imperial College London, two British scientists who wanted to encode information in photons instead of electrons. The phrase "quantum supremacy" was coined in 2012 by the physicist John Preskill. It refers to the moment when a quantum computer will make a calculation much faster than the fastest supercomputer in the world. Google announced that this happened in 2019 when Martinis' team used a new quantum processor called "Sycamore" to make a calculation a lot faster than any existing supercomputer.
Oddly enough for a place called "Silicon Valley", incubators for semiconductor start-ups were rare. Possibly the first one was Silicon Catalyst, founded in 2014 in Santa Clara by Mike Noonen and Rick Lazansky.
click here for the other sections of the chapter "The Selfies (2011-16)"