IBM's focus on servers for businesses prevented it from seeing and responding to what one former employee calls "the biggest server buying spree in a decade."
IBM has had little traction at big datacenters such as Amazon and Google that now represent about 20% of all server sales. Dell and Hewlett-Packard have missed out on a fair amount of it, too, as sales went to the lowest-cost provider of simplified systems -- increasingly Taiwan ODMs such as Foxconn and Quanta.
IBM's focus on services created a product mismatch, says Nathan Brookwood, principal of market watcher Insight64.
"The mega-datacenter guys don't need and won't pay for services; they want lots of dense, low-cost servers. But IBM wants to sell beefier, pricier servers," Brookwood says.
IBM attracted Google into its Open Power group, but it may be too little, too late, some say.
The former IBMer also criticizes the company for not doing enough to create software communities around its Cell and BlueGene processor architectures. He points to a lost supercomputer deal in Illinois as a sign of Intel's increasing dominance in that sector.
In terms of chip technology, "for whatever reasons, IBM decided to go off in directions that did not prove to be mainstream," says former EE Times editor Lammers.
IBM got significant benefits using SOI for embedded DRAM in Power processors, "but they ended up creating a side channel of technology that left them out on their own with higher development costs," he says.
"They had design wins in all three major game consoles at one point, but SOI became too expensive for consumer chips. IBM has super-smart guys with creative ideas, and they are not afraid to be a leader, but it did not work out for them.
"They almost tried too hard to be a leader, an anti-Intel. Now I think the path for IBM is to take their smart people and R&D and work hand-in-glove with the GlobalFoundries people."
— Rick Merritt, Silicon Valley Bureau Chief, EE Times