Competition picks up in the data center space
In the previous part of the series, we saw that Intel’s (INTC) monopoly in the data center market is likely to be threatened by ARM-based processors and IBM’s (IBM) Power processors. While Intel is preparing for the HPC (high-performance computing) market by developing its Knights Landing technology in its Xeon Phi processors and 3D XPoint technology in its Optane SSDs (solid state drives), IBM is competing with Intel head to head with its Power9 processors.
Other than IBM, ARM Holdings is also looking to challenge Intel’s dominance in the data center space.
ARM-based server-class microprocessors
At the Computex trade show in Tapei, Taiwan (EWT), ARM Holdings CEO, Simon Segars, stated that the company aims to increase its share in the data center market from less than 1% at present to around 25% by 2020. It aims to do so by tapping public cloud providers such as Amazon (AMZN) Web Services.
This market share is expected to come from ARM-based server-class microprocessors from Applied Micro Circuits, Cavium, Broadcom (AVGO), Qualcomm (QCOM), and Advanced Micro Devices (AMD), according to IDC (International Data Corporation).
- Broadcom will launch its Vulcan 64-bit ARM server processor.
- Applied Micro Circuits would launch its X-Gene ARM server processor.
- Cavium would launch its ThunderX ARM processor.
- AMD has already started shipments of its “Seattle” Opteron A1100 ARM processor and plans to launch an x86 architecture-based server processor “Zen.”
Google and Rackspace collaborate on data center server
Google (GOOG) and Rackspace (RAX) are jointly developing a data center server code name “Zaius.” They are currently deciding whether to use IBM’s new Power9 processors or ARM-based processors. Once the designs are in place, Google plans to make them freely available on Facebook’s (FB) Open Compute Project. This would enable Chinese companies to develop data center servers at a lower cost. Despite this, Google won’t give up using Intel’s server platform.
Google’s ASIC threatens GPUs and FPGAs
Google has developed a custom ASIC (application-specific integrated circuit) named TPU (Tensor Processing Unit), which it uses for deep learning applications such as object identification, voice recognition, and translation of text to different languages.
If Google makes its TPU available in the market, it will pose competition to NVIDIA’s GPUs (graphic processing unit) and Intel’s CPU-FPGA (central processing unit-field programmable gate arrays) hybrid chips.
Next, we’ll look at the impact of the changing face of the data center space on Intel’s earnings.