Now that Intel Corp. (INTC, Financial) has realized CPUs alone will not be enough to keep it relevant, the need to expand its offerings is at an all-time high. As demand rises for a complete internet of things (IoT) solution, Intel appears to be well positioned to make the most of this opportunity.
Intel initially reported it expects its fastest-growing unit, the DCG (data center group), to register approximately 15% annual growth every year for the foreseeable future. As it stands, this feat is going to be hard to pull off considering GPUs (graphic processing units) have started taking up most of the cloud computing load, which has in turn reduced the demand for Intel’s server processors.
The cloud is a concentrated market dominated by seven large companies. The DCG’s growth tends to vary depending on orders received from these companies. With data centers gradually moving towards AI (artificial intelligence), growth in this segment is expected to slow to about 7% this year. This slowing growth is due to cloud companies adopting accelerators such as FPGAs (field programmable gate arrays) and GPUs to boost performance without necessarily having to add more servers.
While this provides a substantial win for the cloud companies in terms of cost savings brought about by reducing the number of servers and increasing their power efficiency, this has been terrible for Intel’s growth prospects. On the other hand, this has been great news for GPU suppliers such as NVIDIA (NVDA, Financial) and FPGA suppliers such as Xilinx (XLNX, Financial). The latter recently received an order from Amazon Web Services to supply FPGAs for its data center.
Over the course of last year, NVIDIA and Advanced Micro Devices' (AMD, Financial) stock prices skyrocketed. The two companies are now trading at multiple times their 2015 valuations. These surges were fueled by the increasing hype around AI, which is underpinned by an array of algorithms and mathematical calculations that harness methods of machine learning and deep learning so as to evolve its awareness.
All this technology requires a great deal of processing power, and it just so happened that both NVIDIA and AMD’s GPUs were perfect for the task. Intel, which had long been the go-to option for most processing needs, found its chips were at a major disadvantage and could not effectively capitalize on the AI opportunity.
This is because, unlike CPUs, graphics chips are less versatile in the tasks they perform and are best suited for running a large number of parallelized workloads. With thousands of cores efficiently handling small recurring tasks, GPUs have become the standard for machine learning. As a matter of fact, Google’s cloud platform will be powered by AMD’s FirePro server GPUs.
So where does Intel fit in the broader context of AI?
Intel’s plan for AI
Although Intel has not secured the customers or made as much progress as its rivals, the chipmaker is not just sitting on its hands. It has acquired several AI startups and has entered into a number of partnerships with automakers and hyperscale cloud providers to get to the core of AI.
One such acquisition is the buyout of FPGA supplier Altera Corp. Thanks to this deal, Intel was able to launch the Xeon Phi processor with AI support. Intel revealed that the goal of this acquisition was to integrate Altera’s FPGA inside its server processors to deliver comprehensive data center solutions. The company is also developing data center products based on its 3-D Xpoint technology to boost data analytics performance.
In addition to this, the company hosted its inaugural AI day in San Francisco in November of last year to further highlight the company’s strategy, products and ecosystem for the fast-growing AI market. At the event, the company revealed it would add Nervana to its portfolio as a scalable accelerator and would be broadly investing to build an ecosystem for its AI portfolio.
Intel also committed to productizing the Nervana Engine and Neon DNN (deep neural network) software for Intel’s chips later this year. The company went further and boldly claimed Nervana would achieve100-fold improvements for DNNs over today’s best GPU solutions by 2020.
At the same time, Intel revealed some details of the Nervana Engine codenamed “Lake Crest” and the advantages it hoped to exploit. One such advantage was that each chip would have an on-die fabric that enables strong scaling with multiple nodes per CPU, which results in up to 20 gigabytes per second. This is important as it provides support for model parallelism beyond what can be attained with CPUs alone.
According to analysts, the overall AI market is expected to grow at a compound annual growth rate of 62.9% to $16 billion by 2022. Although Intel is struggling to catch up, the strategy outlined by management should help narrow the gap. At the recent Consumer Electronics Show, Intel reaffirmed its plans of leveraging all of its capabilities to deliver end-to-end IoT solutions through its AI strategy.
Takeaway
With the company’s fastest-growing segment under threat from AI advancements, Intel could benefit by focusing more on developing its own solutions for the market. Now that management has outlined a clear roadmap for success, I believe a flawless execution should drive major growth for the company in the long term.
Disclosure:I have no positions in any stocks mentioned and no plans to initiate any positions within the next 72 hours.
Start a free 7-day trial of Premium Membership to GuruFocus.
