Thursday, November 19, 2009

HP and Silicon Valley

The book has generated some 'new insight' re HP's involvement in 'silicon' -- some of those unwritten stories that James Burke with "Connections" always treasured.

The book notes that HP had more semiconductor processes "in production" by a factor of two than any other chip manufacturer, ever. This was due to unique needs for instrumentation leadership mostly. But it translated into more than half of all Gallium Arsenide chips for satellite communications, and more than half of all Light Emitting Diodes (the flashing red lites on your VCR or digital clock) for years -- HP was the undisputed III-V compound semiconductor manufacturer, whether for truly high frequency signals or for light output.

The book also documents HP leadership in microprocessor architecture, design, and manufacture, leading Intel usually by about three years, up to and including the Itanium design, still the world's most "powerful" microcomputer chipset which was designed by HP and licensed for manufacture by Intel.

But memory chips are what built Silicon Valley, actually, and HP's role here was more interesting and more lightly mentioned. Yes, HP helped craft the specs for the Intel 1103 DRAM, the first 1K memory chip, and yes, they bought 40% of the output from Intel for the first two years (East Coast computer folk were still wed to magnetic core memory). But it had never been documented, and I only pieced it together last week, that the Anderson Bombshell (in the book) which nearly derailed the U.S. push for MCC and Sematech, was heavily stimulated by the YHP (Yokagawa HP joint venture) semiconductor testset, derived from HP's acquired R-L-C meters from Boonton Radio. I reported on this in last week's blog, so this will just add a bit.

This is one of those stories where a resourceful group bucked HP top mgmt, survived long and difficult development cycles, found only ONE customer (Hitachi) which figured out how to build radically more reliable, higher performance chips cheaper as a result. YHP never was able to sell Silicon Valley manufacturers; the net result is that today the $20 Billion (yup) memory chip business is entirely elsewhere -- a shift that happened in a mere five years (between the 8K and 64K memory chip 'space'). I have now interviewed four HP top managers involved with YHP, and I don't believe that HP top mgmt or Intel top mgmt ever had a clue that this is all related; if Intel had known it, Andy Grove's epiphany in 1985 might have been hastened by several years (or he might instead have been able to adopt the 'winning tools').

Great speculation at any rate. Fun to hear the stories come forward...

No comments: