The server processor market has been pretty boring to watch as Intel, the 800-pound gorilla in this space, has held onto more than 99% market share of unit shipments for the last several years. The industry has been begging for a viable competitor, and several have thrown their hats in the ring, including IBM with OpenPOWER and Advanced Micro Devices (AMD). In addition, Cavium, Applied Micro Circuits, Broadcom, and others have developed server chips based on a low-power architecture from ARM Holdings. These competitors remain in various stages of development, but to date no one has been able to gain enough traction to make a meaningful impact against Intel’s dominant position.
Just over a year ago, Qualcomm publicly announced a plan to enter the server processor market and demonstrated its first pre-production server chip based on the ARM architecture. Today, Qualcomm announced that its first server processor, the Qualcomm Centriq 2400, has reached the commercial sampling phase. Unfortunately, the announcement did not include many details about product specifics or performance expectations. This is not surprising, since there is no benefit for Qualcomm to tip off the competition on its plans ahead of production shipments. Despite the lack of details, I believe it was important for Qualcomm to make the announcement to demonstrate the company’s commitment and progress in attacking the server market.
Qualcomm Centriq 2400 Processor (Source: Qualcomm)
Here are my top 3 takeaways on the state of the union of Qualcomm’s server initiative:
- Qualcomm is executing to its committed schedule. Last year when the company unveiled its plans to go after the server market, Qualcomm committed to begin commercial sampling in 2016. Today’s announcement confirmed the company’s ability to achieve this milestone. Qualcomm also committed to begin shipping Centriq 2400 in production in 2017. It will be critical for Qualcomm to continue to meet its commitments in the datacenter space and to avoid being distracted by large corporate level changes, such as the pending acquisition of NXP Semiconductors. The fact that the organization leading the server effort—Qualcomm Datacenter Technologies—is set up as a separate subsidiary may help the group remain focused and on track.
- Qualcomm has the ammunition needed to make a competitive product. Competing effectively on performance for datacenter-class workloads against Intel Xeon is not something anyone in the ARM server space has been able to do to date. One of the key messages in Qualcomm’s announcement was that the Centriq 2400 is the “first and only” server chip built on a leading edge 10nm manufacturing process. It is important to note that process geometry alone doesn’t indicate size and power, and there are many other factors to consider that will determine if the 10nm advantage can translate to real value for customers. Also, Qualcomm has significant expertise in designing custom ARM cores from its mobile business. While these groups use different design teams, there may be some leverage in terms of process, resources and IP from the company’s mobile experience for the server-optimized Falkor core used in the Centriq 2400. Using a leading-edge process node with a new custom core and chip design presents its own set of risks but positions Qualcomm to better compete on performance and performance per-watt-against Intel than the rest of the ARM crowd. It is way too soon to say if Qualcomm will be able to execute to its competitive performance goals once they start shipping in production, but their arsenal of R&D capabilities is solid and positions them well.
- Qualcomm remains focused but recognizes the long-term opportunity. Qualcomm’s primary targets for Centriq 2400 are “megadatacenter” customers like Google, Facebook, Amazon and others who build their own datacenters and deploy servers at massive scale. These customers have the buying power to purchase directly from chip manufacturers and have significant expertise on how to optimize their applications. New technologies that provide efficiency gains are worthwhile to megadatacenter customers, as the sheer scale at which they purchase can mean huge cost savings over the long term. While Qualcomm sees opportunity for Centriq 2400 in other segments such as high performance computing, telecommunications and financial services, the company’s executives made it clear that they are being careful to remain focused on the megadatacenters at this early stage. Another long-term opportunity I believe Qualcomm has with Centriq 2400 is Internet of Things (IoT) infrastructure, where the proliferation of connected devices drives new demand for edge computing and cloud datacenter resources. There may even be an opportunity for products like Centriq 2400 to reside in the IoT devices themselves. For example, Qualcomm competitors like NVIDIA and Intel are targeting self-driving cars with datacenter-class products. Qualcomm has an increased focus on IoT at the corporate level, with the pending acquisition of NXP. Tying the datacenter and IoT initiatives together is not a near term priority for Qualcomm, it but could be a key part of the company’s long term strategy.
ARM-based servers have been being hyped in the market for 6+ years, with little to show for it in terms of customer adoption. Qualcomm is not likely to begin to ramp in production with Centriq 2400 until late 2017. But the industry’s desire for an alternative to Intel remains. Other competitors like AMD Zen and IBM’s OpenPOWER are also positioning themselves to fill this gap and may be able to beat Qualcomm to market. However, if Qualcomm can deliver a solution that provides a competitive advantage over Intel, large datacenter customers may see benefit in moving some of their workloads to Qualcomm Centriq-based servers. It will be key for Qualcomm to commit the resources needed to execute on its commitments over the next couple of years. And if Qualcomm plays this right, I believe the company can create a unique and compelling corporate strategy that combines its datacenter focused initiative, strong position in mobile and investments in IoT.