Intel has owned Altera for almost a year and a half. Since that purchase, there has been continuous speculation that Xilinx, the other major FPGA player, would be next to get acquired. However, despite Altera recently having a tough run in getting the latest generations of solutions to market in time, which provide greater density, performance, and power advantages, the fact that Xilinx is still independent, so far, is probably not a fluke.
While FPGAs generate high margins, there is nothing inherent to the business that will allow it to outgrow the overall semiconductor market significantly in terms of revenues. As a general rule, once a particular volume is achieved in the space, there is movement to ASICs. In addition, Xilinx’s market cap is disproportionately high compared to its pretty much static annual sales over a long period because there has been this expectation of an acquisition at a large premium.
Moreover, it is widely known that Intel hardly has a flawless record in aggressively delving into new technological realms. While there continues to be movement of FPGAs for applications into the data center sector, such as for accelerating deep learning, convolutional neural networks, and for deployment of video search enhancement (as well as for video compression and transcoding), the size potential of such opportunities appears to still be unproven.
At the same time, Intel’s potential to embed its Xeon processor with the FPGA provides a more powerful offering. Unquestionably, it has tremendously more sales and marketing muscle, compared with Xilinx, to sell such a combined package, particularly to certain large users of these programmable components.
Nonetheless, it seems that Xilinx’s sole focus on FPGAs currently gives it a widely perceived advantage in the marketplace, particularly outside of the data center space. Especially with the realization by Taiwan Semiconductor Manufacturing Company (TSMC) that it was no longer getting Altera’s business, Xtera appears to remain quite cost competitive with Intel, despite the fact that the giant chip company owns its fabs. Xilinx is also getting the necessary support from TSMC for its aggressive movement to even lower nanometer technology.
On a side note, we like Microsemi’s FPGA game plan in at least potentially securing a very viable niche position, such as in the military space. In the past, we have endorsed the idea in general of smaller competitors taking a couple of steps back to avoid the fray among the bigger players, preoccupied with the latest solutions.
Microsemi is targeting the more mature, highly stable, as well as widely deployed 28-nanometer node, and looking to provide minor refinements, including greater security, lower power, and less weight. Substantial sales at that size are likely to last for several more years, as the cost continues to decline on these devices.
[written by Mark Lutkowitz]