Intel May Challenge Qualcomm in the Data Center Space



Data center: A new battleground for Qualcomm

Qualcomm (QCOM) has been eyeing the high-margin server chip market. This is visible from Intel’s (INTC) earnings—it earned 29% of its fiscal 3Q16 revenues and 47% of its operating profits from its Data Center segment. It is critical for Qualcomm to meet its data center commitments, as any delay would make it tougher for the company to compete with Intel, which is constantly developing advanced nodes.

Article continues below advertisement

Competitive challenges

Although Intel owns more than 90% of the server chip market share, there is high competition for the remaining share. IBM (IBM) has developed the Power 9 processor, Advanced Micro Devices (AMD) is launching the Zen processor, and several small companies are struggling with their ARM-based processors. None of these companies have been successful in gaining a decent market share against Intel.

In this scenario, Qualcomm is launching its ARM server processor. The company is optimistic that it would be able to gain a decent share of the server chip market. QCOM’s Centriq 2400 is better than all available ARM processors and is competitive with Intel’s Xeon. Moreover, it would be the first 10nm (nanometer) server chip to enter the market.

However, Intel is not bothered by the 10nm feature, as no foundry has been able to match Intel’s advanced node technology.

Technological challenge

Many analysts stated that no foundry has been able to compete with Intel with respect to its chip manufacturing technology. Although TSMC (TSM) and Samsung (SSNLF) were able to develop smaller nodes, they could not leverage the full benefits of shrinking chip densities at each node.

This is visible from the chart above. Samsung’s 16nm chip and TSMC’s 14nm chip performed better than Intel’s 22nm chip, but they were not competitive with the latter’s 14nm. Hence, Samsung’s 10nm node is likely to have performance similar to that of Intel’s 14nm.

Data center market shifts toward deep learning

Another major challenge is the shifting of the data center trend toward deep learning and AI (artificial intelligence). This is one place where Intel lags behind Nvidia (NVDA), whose GPUs (graphics processing units) are used by most data centers for their AI needs.

However, Intel, IBM, and AMD are slowly incorporating this AI technology into their server chips. Against this backdrop, Qualcomm’s Centriq 2400 can only handle basic data center loads.

Because of the above challenges, Qualcomm has adopted a wait-and-watch approach. If it succeeds with mega–data centers of web service companies, it may target high-performance computing, telecommunications, and financial services.

In the next few parts, we’ll see how these product announcements could impact the earnings of Intel, Qualcomm, and Microsoft.


More From Market Realist