Microsoft’s internal memo: So Long, Intel (1992) [pdf]

Article note: What a neat artifact! Nathan Myhrvold did some spectacular prognosticating here. The Apple/IBM prediction is _dead on_ up to the point where he didn't foresee that _after_ the collapse of Pink they would just buy a successor platform. Technically he's right that 64bit architectures wouldn't move into the consumer space for a decade, but he dismissed 64-bit extensions to x86 and that's what eventually won as "the" ISA. He failed to predict that Intel would build a superscalar P5 and captures-all-the-RISC-advantages-behind-a-dyanmic-reflow P6 that let them get away with all kinds of bullshit for a decade. He's right that one of the _problems_ that messed things up for the RISC world (for Unix and NT and ...everyone) was that supporting multiple ABIs in a platform is a huge burden. He correctly points out that DEC was in position to do Modern-OS-on-modern-RISC in the mid 80s and blew it so hard that Microsoft got "the guy they spurned" (Dave Culter). He correctly anticipates that and how many-commodity-microprocessor systems would eat the bespoke high end market. He caught that low-end RISC would eat Intel from below. He missed that the great VLIW snowplow would fuck up the whole RISC market then collapse into itself...and somehow Intel would come out on both ends of that, but that story is unbelievable.