- Intel’s Xeon platform will soon soak-up opportunities in telecommunications due to NFC or C-RAN.
- Furthermore, Intel will soon launch a new CPU that tackles the most in-demand workload (video streaming).
- DCG results will likely fluctuate as it has become more dependent on the launch and adoption of new technologies.
Intel's (NSDQ:INTC) efforts in the cloud are at least noteworthy, as the data center market is growing at respectable rates. Unfortunately, this doesn’t represent much of Intel’s total revenue but corresponds to roughly half of Intel’s gross profits. Therefore, the higher mix from the enterprise segment should translate into healthier operating profits and EPS growth. So, bottom line improvements remain highly likely in 2016.
Diane Bryant, the head of Intel’s datacenter division made some presentation points at Computex 2016 worth analyzing and observing in greater detail.
Key takeaways from Computex 2016
Intel is planning to launch a new CPU for datacenters called the Xeon E3 1500 V5 with Iris Pro 580 integrated into the enterprise class CPU. Diane mentions that the new CPU should drive enterprise adoption, as much of the consumer internet traffic is driven by video usage (72% of internet traffic is composed of video streaming from sites like YouTube, Netflix, Facebook and online gaming).
Basically, the new Xeon CPU addresses high bandwidth use cases. The Xeon E3 1500 can process 2 HEVC 4k live streams or 15 HD live steams with real-time compression and transcoding, so live broadcasts of events can be handled via datacenter architecture rather than a live-feed, which tends to have lower quality outputs (current cable operators tend to offer 1080i) in the United States. 1080i utilizes interlacing to boost the frames per second, reducing motion blur, but also outputs at lower quality (roughly 720p) versus standard pass HD and UHD (1080p and 4k).
Basically, Intel’s keynote on Xeon architecture inclusive of 1080p and 4k is interesting, because if a datacenter can handle both the livestreaming compression and transcoding while also running network function virtualization via packet processing along with decryption of encrypted data packets, the standards may start to converge in broadcast and standard online streaming.
Intel continued the discussion of network function virtualization, which separates compute processing in separate environments within a datacenter allowing provisioning for workloads across standardized X86 architecture for different applications. This may sound confusing, but there are a lot of overlaying synergies when you start to standardize network architecture around a single hardware ecosystem that services both the internet and live broadcasting.
Also read: Should You Buy Intel After Computex 2016?
How this affects cable television and broadcast operators
Currently, the United States operates off of the ATSC standard for live broadcasting, which is MPEG-2 a technology for video compression and transcoding that was standardized for digital signals and introduced in 2002 (the technology is quite old, but is used as the standard TV signal to this date). In an CNBC interview the CEO of CBS, Leslie Moonves mentioned that there will be a convergence of data pipelines over the next 10-years.
The convergence implies that the video standards utilized in today’s streaming, i.e. MPEG 4 and HEVC (high efficiency video coding) will become standardized for digital broadcasting of live events. Essentially, the only reason why Netflix is able to offer newer standards like 4K and bandwidth saving encoding features is because it operates out of a datacenter environment with end-devices that can quickly adopt new streaming standards.
Cable operators on the other hand have limitations of bit rates at roughly 19.39 megabits/second, whereas high bandwidth cable is closer to 300 megabits/second with fiber to the premises driving 1 gigabit/second. Therefore, today’s television channels will eventually adopt higher bandwidth standards via a conventional network interlink that operates via packet processing and standardized internet-specific data channels. This will eventually operate outside of bitrate restrictions for MPEG 2. Broadcast signals will converge to a singular network interface in other words, which means that it will operate in the same environment of fixed and wireless broadband.
Intel doesn’t dive into the finer points of standard broadcast, but I believe standard broadcast/cable will utilize internet infrastructure, which implies that the back-end infrastructure for the 4K standard equates into adoption of data center solutions. In the end, Intel is best positioned to capitalize on the transition of data transmission standards to both 5G and television/internet convergence.
Financial takeaway for Intel
Intel mentioned at the 2016 Computex Keynote that it was going to partner with Foxconn for Network Function Virtualization. The two companies (Intel and Foxconn) signed a memorandum of understanding to develop the hardware, whereas Foxconn will focus on design and manufacturing capabilities ahead of full deployment. Some of the bigger mobile carriers are adopting C-RAN (cloud radio access networks) which Intel labels network function virtualization. NFV or C-RAN also utilizes software defined networking (SDN) where resources are pooled and operated out of any number of virtual machines that get provisioned access to pooled data center resources.
C-RAN or cloud virtualization is both modular and standardized under the most prevalent enterprise x86 hardware ecosystem. There are significant cost savings to adopting C-RAN, as China Mobile has rolled out this technology to various markets with very promising results in high-density urban environments. This basically implies that companies like AT&T and Verizon will also save on Capex (capital expenditures) allowing a quicker rollout of 5G technologies.
The new 5G standards are convoluted, but there’s an expectation on the order of 1,000x improvement in network capacity, 100+ billion devices can be supported, 10Gbit/second peak speeds and below 1 millisecond latency. Standards have yet to be defined, but the implications of next-generation speeds are transformative for Intel’s datacenter and mobile computing business.
Intel is still doubling down on the development of mobile modems capable of reaching the necessary signal packet processing within a mobile form factor for 5G. There are also promising tests over at AT&T where they’ve achieved speeds in excess of 10 Gbit/second as it has a roadmap for deploying its 5G base station footprint by the 2019 timeframe. The development of AT&T’s 5G network was done in coordination with Intel and Ericsson. Furthermore, Verizon is also anticipating initial deployment of 5G by 2017, which implies that Intel’s Xeon-series CPUs will start soaking up the network processing market in the 2017 to 2019 timeframe.
The addressable market is a bit convoluted, but when looking at a past data center day presentation there’s roughly $8 billion worth in enterprise CPU revenue that can be derived given the adoption of AT&T, Verizon, China Mobile, Vodafone, SK Telecom and Telefonica. However, the adoption of these technologies seems more contingent on 5G rollout for North America base stations, which means that the impact financially is hard to model and doesn’t seem like an immediate contributor to datacenter group results.
However, the build out of 5G base stations and the transition to C-RAN will become a bigger tailwind for FY’17 results with some uptick to CPU volumes in 2016. It’s hard to tell though, because Intel didn’t produce exciting numbers in its Q1’16 datacenter group (DCG) results, as it went through a period of shrinking IT budgets.
Essentially, the Q1’16 9% y/y growth for DCG implies that expectations for growth by management were unrealistic as near term drivers to server SKUs abated due to slowing public and private cloud investment into IT. For these trends to reverse it will require the onset of new market opportunities in the 2017 timeframe to drive the figure closer to the company’s long-term outlook range of 15% y/y growth. I believe, Intel’s long-term outlook of continuous mid-teen revenue growth for DCG will need to be inclusive of CPU-volume for network transformation to 5G and C-RAN, which is why I’m revising my expectations on the current fiscal-year, but am optimistic on FY’17 as roll out of new network technologies has healthy implications for top line growth in the most profitable segment.
I still remain optimistic on Intel’s efforts in the datacenter group. However, I would focus on long-term growth drivers outside of conventional cloud based consumer applications. The growth opportunities have shifted to network-specific workloads, and opportunities in consumer workloads will become more meaningful once bandwidth needs are addressed with the migration to 5G. Datacenter workloads will correspondingly increase with network bandwidth improvements.
We’re witnessing some cyclicality in Intel’s datacenter business, as evidenced in the prior quarter. It’s dubious that growth will continue on a linear path, so investors should see fluctuations in y/y comps due to the timing of new technologies.
Furthermore, I believe cable/broadcast television could become a long-term tailwind to datacenter results. I think investors can anticipate cable/broadcast television to transition to network-based processing in the 2018 to 2025 time frame. So, Intel will go through multiple phases of growth, but each one building off of core technologies like software defined networking, and standardization of data transmission through inter-operable standards running on PC architecture.
These technological building blocks are crucial to long-term financial models, but become quickly convoluted by the technical language. It is my hope that this article helps fill the missing holes.
I continue to reiterate my hold recommendation and $30.54 price target.