According to analysts of Market Research Media Ltd the governmental High Performance Computing markets shift from Vector to Cluster, from Custom Supercomputer to COTS (see COTS Is the Name of the Game for Defense HPC Market), with parallel designs based on “off the shelf” server-class microprocessors, such as the PowerPC, Opteron, or Xeon, and most modern supercomputers built on computer clusters using commodity processors combined with custom interconnects. It seems that Cray fall victim to this technology shift. In a recent release of preliminary 2009 results Cray announced about DARPA cuttuing back on Cray super project:

“During the fourth quarter, a contract modification was not finalized and consequently a milestone was delayed in the agreement covering Phase III of the Defense Advanced Research Projects Agency’s (”DARPA”) High Productivity Computing Systems program. This contract delay will have an approximately $7 million negative impact to net research and development expenses in the fourth quarter of 2009. Substantive talks regarding this contract are ongoing and the company expects to complete a contract modification and milestone in the first quarter of 2010. As with the previous contract, the balance of the modified contract is expected to be received through achievement of a series of pre-defined milestones culminating in the delivery of a prototype system in 2012. Consistent with this change, certain deliverables will be eliminated from the contract, reducing the overall scope and cost of the project. After the anticipated contract modification, the remaining amount of the contract is expected to be reduced by $60 million, to $92.5 million. As a result of the reduction in overall scope, we expect future research and development expenses related to this program, net of reimbursement, to be lower than previously anticipated. “

While Cray is rather vague on the nature of the DARPA requirements and specifics of the DARPA contract, it seems that DARPA preference of competitors’ HPC systems relying mostly on COTS played its role. According to a recent market report, Worldwide Defense High Performance Computing (HPC) Market Forecast 2010-2015, HPC systems built on COTS is the only way to meet exponentially growing military computing requirements and lagging behind budgets.