The number of participants in the US Semiconductor complex continues to shrink. Yesterday, AMD announced it intends to acquire Xilinx. AMD with a $90 billion market cap and Xilinx with its $30 billion valuation are both ”mid-sized” companies in a field increasingly dominated by a handful of mega-cap companies, to which the combined entity can now claim a seat at the table.
Some background. AMD is best known for making general processors – CPUs for compute and GPUs for graphics and AI. Long the second place behind Intel in the CPU market, AMD over the past decade appears to have finally righted itself after a tumultuous history. It has been steadily gaining share from Intel most notably in the profitable data center market. Xilinx makes FPGAs, which are ‘programmable’ chips used in a wide array of markets.
The combined company will have significant share in PCs, data center and a range of ‘infrastructure’ and networking products. In theory, there are some meaningful synergies where AMD products sit next to Xilinx products. (We say in theory because integration is always the hardest part of an M&A deal.)
We have long held that the number of chip companies in the US is going to continue to shrink, and this deal fits neatly into the underlying logic of this consolidation.
There are some clear implications for the broader industry here. Most notably for Intel. Xilinx’s biggest competitor is Altera which Intel acquired five years ago. So AMD is now ratcheting up the competition with its long-time nemesis. AMD now competes with Intel across Intel’s core product line. Those with a long history in semis will remember that AMD was once very much the laggard in this race. They almost fell out of the market and were essentially saved by Intel 15 years ago with a timely licensing agreement. Back then Intel needed AMD in the market as a credible competitor to avoid greater anti-trust scrutiny, but they seem to have created a bit of a monster. As world governments enter a new wave of tech anti-trust furor, this is a useful example to keep in mind. Competitive advantage in technology can be fleeting.
It is also worth noting that AMD’s fortunes really turned around ten years ago when they spun off their fabs (chip manufacturing line) assets into the company now known as Global Foundries. Unable to compete in both the fab market and the CPU market, AMD split itself in two, and the CPU side at least has thrived ever since. We have to think this fact is not lost on Intel.
Beyond Intel, there are other important considerations for the industry. Xilinx and Altera are, by far, the largest FPGA companies out there. There are a handful of other companies but they are all much smaller than either of these leaders. FPGAs are important in two ways.
First, FPGAs are a great solution for products that need intense compute but are not large enough in volume to merit building a special-purpose chip (aka an ASIC), hundreds of thousands of units, not tens of millions. We find them in network switches, autos, some data center uses, and a wide range of industrial and aerospace applications (more on this in a moment). Both major FPGA companies are now part of large organizations with the ability to cross sell their products into a range of end products and the ability to afford the latest manufacturing processes.
Secondly, FGPAs are a critical tool for designing advanced chips. Designing large chips takes a considerable amount of time, and building physical chips is a big upfront expense at the foundry. Chip design teams need to simulate their designs long before they can get physical silicon components to test. Instead, they use FPGAs to do that. As such FPGAs are a critical tool for chip companies. And the FPGA makers have a lot of insight into the broader chip market. That means there are now a lot of other companies, the ones who lost the auction for Xilinx, who are going going to have to work with a competitor to build their own products.
We do not think this is going to be sufficient grounds for US/EU anti-trust regulators to block the deal, but it is going to be a concern for many companies going forward.
And then there is China. China’s regulators are going to have a say in approving this deal, and as with all such topics today, this approval is a big question mark. We saw a report recently which listed China’s highest priority areas of investment for semis. These includes the obvious ones like advanced fabs and AI chips, but FPGAs are high on this list as well. It turns out that a big use for FPGAs is in radar, missile and avionics systems. China’s regulators probably do not care which US company owns this, they are equally worried about all access to US chips. That being said, we imagine this deal can join the list of US companies waiting in line at the China anti-trust office.