Then AMD acquired Altera’s biggest competitor, Xilinx, in 2022 for a whopping $50B. Now, unsurprisingly, Intel has decided to cash out.
While the anticipated growth for programmable hardware (FPGAs) in AI has yet to materialize, other market segments continued to grow. Intel doesn’t break out PSG sales but said in July that the unit had three record quarters. Given the strength in the business and perhaps a diminishing role for Intel, it looks like a great time to spin PSG out and enhance shareholder value. INTC was up some 2.5% after hours. AMD/Xilinx controls about 60% of the FPGA market, while Intel/PSG controls most of the remaining 40%.
However, the promised strategic synergies with Intel’s AI business never materialized. And never will, according to market forecaster Omdia. See that little yellow bar below? Yeah, that’s FPGA’s role in AI. FPGAs are great in networking, image processing, automobile cockpits, and the vast market of tiny markets where engineers need to build a custom chip, but not hundreds of thousands.
Omdia Projects very little revenue in AI for FPGA architectures. OMDIA
Whatever happened to the supposed synergies between the two companyies? Intel did produced a combo chip as promised, combining a Xeon CPU with an Altera FPGA. However, the Xeon SP-6138P was not a commercial success. It was expensive and complex, and it required specialized software and tools to program. As a result, it was only adopted by a small number of customers. In 2021, Intel announced that it would be discontinuing the Xeon SP-6138P. The company said that it was shifting its focus to other FPGA products, such as its Stratix and Agilex families.
As for AMD, the ex-CEO of Xilinx, Victor Peng, is now running AI for the company, and leveraging the AI expertise in the Xilinx team. While AMD still seems to think a combined CPU+FPGA makes sense, I doubt it will be successful, if it is launched. The Xilinx Versal SoC already has an Arm CPU on the die. Nonetheless, AMD has announced that it will produce a combined CPU+FPGA chip sometime later this year.
Conclusions
While the hoped-for synergies never materialized, and the role of FPGA’s in AI remain a glint in a few enthusiasts’ eyes, buying Altera is proving to be a good deal for Intel shareholders. If Intel can IPO for, say 2/3*$50B=$33B, thats a decent ROI for the acquisition. And of course, PSG will continue to use Intel Foundry Services to fabricate it’s products. Sandra Rivera, having had PSG under her wing for the last several years, is an ideal CEO for the new company, and we expect the company to flourish as a standalone and independent firm.