Intel plans to spin off the Programmable Solutions Group

Intel has announced plans to separate and eventually spin off the Programmable Solutions Group, created after the acquisition of Altera Corporation in 2015. Intel will retain a majority shareholding and sell IFS services to the business. How does this fit in with Intel’s long-term strategy?

Background

Intel acquired Altera Corporation in 2015 for $16.7 billion, creating the Programmable Solutions Group under the Intel Data Centre and AI Group. FPGAs (or Field-Programmable Gate Arrays) are a class of processors that can be programmed by customers to perform a set of repeated tasks, for example, data processing or network packet inspection.

Intel acquired Barefoot Networks in 2019 to add the intellectual property needed for Ethernet switch silicon development, including programming with the P4 programming language. This added to the capability of solutions from the Programmable Solutions Group.

In August 2021, Intel announced the Mount Evans architecture using a custom ASIC platform called the E2000. This incorporates Arm Neoverse N1 cores, a choice of 100Gb or 200Gb Ethernet, 48GB of DRAM and PCIe 4.0 support. This technology was incorporated into Google Cloud C3 virtual instances to support storage and networking functional offloads.

PSG

Intel has roadmaps for what it terms IPUs or Infrastructure Processing Units. These are the collection of devices using either FPGAs or ASICs. Both of these technologies have separate roadmap paths extending up to 2026. You can find out more about SmartNICs, ASICs, FPGAs, and similar technologies in our 2023 Intelligent Data Devices Report (link below, paid content).

Intelligent Data Devices 2023 Edition – A Pathfinder Report

This Architecting IT report looks at the developing market of SmartNICs, DPUs and computational storage devices, as data centres disaggregate data management processes, security and networking. Premium download – $295.00 (BRKWP0303-2023)

Download Now

What’s not clear at this point is whether the spin-out would separate all of these technologies into a new entity or if the focus is just the FPGA business. We believe it must be all IPU products; otherwise, there is no value to the transaction.

The PSG business is important, as we’re seeing a transition in the data centre where the CPU now competes with GPUs, FPGAs, and ASICS. The data processing functions of infrastructure demand new approaches that have resulted in specialisation, improving performance and optimising costs (to a degree), but at the expense of complexity.

IDM 2.0

Why would Intel be spinning off PSG now? As we’ve discussed at length, the data centre landscape is changing. Intel is being challenged on all fronts, and the domination of x86 as the primary data centre architecture is on the wane. Pat Gelsinger’s IDM 2.0 strategy refocuses the company on manufacturing and leading the development of advanced process solutions, with a roadmap looking forward to 2nm and 1.8nm in the future.

However, there is a bigger picture here that we should consider. In recent years, the US has identified the threat of China increasing its stake in chip manufacturing. Most of the world’s advanced chips are made by TSMC, based in Taiwan, right off the Chinese coast. The US government wants to retain microprocessor superiority, and so enacted the CHIPS and Science Act in August 2022, with roughly $280 billion of investment, of which $39 billion is available as subsidies for chip manufacturing in the US. The EU has also introduced its own programme, called the European Chips Act, with €15 billion of investment available.

Lean

With governments willing to spend and a plan that focuses on core products, Intel (and specifically Gelsinger) has set about making the company leaner and more focused on the core mission of IDM. Loss-making Optane was discontinued, NAND products were sold off to SK Hynix, creating Solidigm, while Mobileye was floated off in October 2022. Intel also shut down the NUC business, although this move enables partners to continue manufacturing the technology.

Leading the design and manufacturing of products into the “Angstrom” era is where Intel wants to be.

The Architect’s View®

The recent changes in Intel’s business place the emphasis on data centre computing excellence back at the core of the company. Anything outside of that focus is being pushed aside, although this doesn’t mean an asset sale is in place.

Intel still retains 88% of Mobileye, providing a convenient cash cow for the future. The spinoff of PSG represents the same potential. In addition, Intel can still make money licensing NUC designs. However, in each instance, spinning out these business units removes distractions from Intel management, allowing them to focus on the core strategy of IDM 2.0.

Will we see more business divestment? It’s an interesting question. Although Intel may continue to hold the golden share of Mobileye and PSG, the shares of those entities could be acquired by competitors and force a takeover. Intel would have to justify why the public ownership model is the preferred structure of the company.

Of course, much like the “Federation” of EMC Corporation businesses prior to the Dell takeover, there is a symbiotic relationship in place where Intel can sell IFS (foundry) services to the independent PSG.

We think floating more businesses off would be a risk for Intel, but it may be part of the master plan to get back to core basics. Only time will tell.