At a recent investor update, Intel (INTC -9.20%) CEO Pat Gelsinger announced that the company's "Programmable Solutions Group" (PSG) will operate as a stand-alone semi-autonomous chip business.  

This is a bit confusing. Intel used to report PSG-specific segment revenue through 2021, but stopped doing so in 2022. Now the company wants to ready its PSG unit for an IPO to sell to investors within the next few years. Here's what investors need to know about Intel PSG.

What is PSG exactly?

Intel PSG deals with FGPAs, or field programmable gate arrays. These chips are designed so that they can be reprogrammed "in the field," or once put into operation, giving engineers the ability to overhaul the functionality or completely repurpose the FPGA once it's installed in a computing system. 

This inherent flexibility has caused FPGA chips to proliferate across all sorts of applications -- from data centers (primarily as networking chips, but also for AI compute workloads) to industrial (processors controlling specific functions in a car, plane, or piece of machinery) to chip emulation and verification (FPGAs can be programmed to mimic the functionality of a new chip design so it can be tested before it goes to the actual manufacturing floor).  

Intel originally formed its PSG unit in 2015 when it acquired FPGA company Altera for $16.7 billion. As Intel often did over the course of the 2000s and 2010s, it kept buying other small operations to bolt onto its own. In 2018, it acquired the small start-up eASIC, and in 2019 it acquired small FPGA company Omnitek, both purchases getting added to the Altera PSG business.

PSG has done ok the last year-and-a-half since Gelsinger and his company decided to stop reporting it as a stand-alone unit. In 2021, PSG revenue was $1.9 billion. It was mentioned on the aforementioned call with Gelsinger and new PSG CEO Sandra Rivera (who will come over from Intel's Data Center and AI segment) that PSG revenue was up about 50% since the end of 2021 to $2.9 billion over the last trailing 12-month period (at the end of June 2023). However, some upcoming "cyclicality" due to the chip market downturn is likely to reduce this revenue a bit in the quarters ahead temporarily.

By comparison, PSG's top competitors are Xilinx (now part of AMD, known as that company's "Embedded Group") and Lattice Semiconductor (LSCC -0.04%). AMD's Embedded Group reported Q2 2023 revenue of $1.5 billion (or $6 billion on an annualized basis), also up about 50% from where Xilinx was at the end of calendar year 2022. Tiny FPGA designer Lattice -- the final FPGA "pure-play" available to buy stock in -- has grown its sales by about 40% since the end of 2021, but hasn't given any indication its FPGA business will decline this year like its larger peers have.

Intel's need for flexibility, and cash

Intel PSG will once again be reported separately starting at the beginning of 2024. But why now? The idea is to get the unit ready for outside investment, much like Intel has done via the Mobileye (MBLY -5.50%) autonomous vehicle business that went public (again) at the end of 2022, but which Intel still owns 90% of; and more recently the chip manufacturing equipment subsidiary IMS Nanofabrication, which Intel sold minority stakes in to private equity firm Bain Capital and Taiwan Semiconductor Manufacturing (TSM 1.26%).

Intel needs the cash as it doubles down on Gelsinger's IDM 2.0 (Integrated Device Manufacturer) strategy, setting Intel up as a much-needed third-party manufacturing business servicing the many dozens of fabless (no manufacturing) chip designers out there. Taiwan Semi is currently the dominant force in this market, but there's geopolitical risk given China's overtures toward Taiwan that U.S.-based and other chip designers would like to address. 

But building new chip manufacturing capacity is ridiculously expensive. Various projects are going to cost Intel tens of billions of dollars over the next few years, and funding from various governments (the U.S. CHIPS Act and the European Chips Act, for example) will only partially fulfill some of these cash needs. 

Gelsinger pointed out that selling stakes in Mobileye and IMS netted Intel about $4 billion in cash. PSG is another outlet to raise more funding. As for the merits of investing in PSG itself, let's see what kind of financials Intel provides early next year. For now, Rivera pointed out that the FPGA market is currently worth about $8 billion a year in total customer spend (dominated by AMD's Embedded Group, formerly Xilinx), but could grow to $11.5 billion in annual spend by 2027.

Getting PSG ready for sale looks like a good move as Intel continues to sharpen its focus. I'm personally still not buying Intel stock. But I've warmed a bit to the company's strategy as Gelsinger has gotten more serious about offloading non-strategic businesses in 2023 to gear up Intel's manufacturing aspirations.