PCWorld reported that NVIDIA (NVDA +3.97%) no longer plans to develop a processor for servers. This comes after NVIDIA's 2011 announcement of an ARM (ARMH +0.00%) compatible CPU core known as Project Denver. NVIDIA claimed the CPU was intended to extend the range of ARM systems upward to PCs, data center servers, and supercomputers.
However, today, it looks as though Project Denver -- along with any potential successors -- will be aimed at the "traditional" Tegra processor markets of tablets, automobiles, and so on.
NVIDIA initially wanted to sell HPC-oriented CPUs
From NVIDIA's initial Project Denver announcement, it seemed the company wanted to make a play at grabbing more content share in the high-performance computing market. The company already sells very high-performance accelerators (based on the same silicon as NVIDIA's consumer GPUs but with specialized firmware/drivers), which is a good, high-margin business.
However, grabbing the CPU/platform with its own host processors could have driven potentially meaningful incremental revenue for NVIDIA.
Intel (INTC +2.97%) has significant structural advantages in this market via its Xeon family of processors. Perhaps NVIDIA management did not feel comfortable sinking hundreds of millions of dollars to develop host processors for high-performance computing that could ultimately fail to generate the proper return on investment.
Partnering with IBM and the ARM vendors makes sense
In a bid to diversify away from being attached solely to Intel systems in high-performance computing (Intel is trying to cut NVIDIA's GPUs out with its own Xeon Phi products), NVIDIA is partnering with IBM -- which sells POWER-architecture based systems -- and Applied Micro, a company making a play at bringing the ARM architecture to servers.
Source: NVIDIA via AnandTech.
It's unclear whether the ARM vendors will make progress here (though IBM's POWER 8 looks very compelling as an alternative architecture), particularly as Intel's Xeon products are quite strong and enjoy both incumbency and a very large research and development budget. However, making sure its HPC-oriented GPU products are compatible with all potential solutions on the market seems like the correct business decision for NVIDIA.
Foolish takeaway
Though it would have been interesting to see what sort of monster HPC-oriented host processor NVIDIA would have come up with, the economic realities of this market likely led the company to take a more "horizontal" approach. NVIDIA's HPC-oriented GPUs are the best in the world at what they do, and investing to make sure that they stay that way and attaching them to others' platforms seems like the right strategy for NVIDIA at this time.
