In 2015, Intel (NASDAQ:INTC) announced that it intended to purchase programmable logic vendor Altera (UNKNOWN:ALTR.DL) for $16.7 billion in cash. The deal, Intel claims, should create value in two significant ways.
The first, according to the chip giant, is that by transitioning Altera's existing stand-alone FPGA products to Intel's chip manufacturing technology, Intel can improve Altera's competitive positioning in the market for such products. This, the company says, comprises about 40% of the value that Intel expects to create from this deal.
Next, the remaining value that Intel expects to create from this gargantuan buy is by taking Altera's FPGA technology and integrating it into its server processors for significant performance/power gains.
Although I really like this deal conceptually, it has the potential to be a significant flop. Here's why.
The stand-alone FPGA thesis appears tenuous
The core of Intel's thesis vis-a-vis stand-alone FPGAs is that it should be able to gain a manufacturing advantage over rival Xilinx (NASDAQ:XLNX), which builds its FPGAs at contract chip manufacturer TSMC (NYSE:TSM).
However, it's worth noting that Xilinx appears to have beaten Altera to the punch with a 14/16-nanometer class stand-alone FPGA. Indeed, Xilinx apparently began shipping samples of such a product in late 2015, while it doesn't appear that Altera/Intel are sampling their own 14-nanometer FPGAs yet.
Xilinx is also now talking about skipping over TSMC's 10-nanometer technology and landing its next high-end FPGA on TSMC's 7-nanometer technology. Although the specifics of the technology are not yet known, TSMC's 7-nanometer should offer equal to better transistor density than the Intel 10-nanometer technology that Altera's next high-end FPGA will be built on, potentially allowing Xilinx to pack in more features/functionality into a given chip area.
At this point, I don't think it's prudent or even reasonable to assume that Altera will be ahead of Xilinx in terms of migrating to next-generation chip technologies. If anything, it's looking to me like Xilinx has the edge.
Integration could go about as well as cellular integration did in mobile
In 2010, Intel announced that it planned to acquire the struggling wireless division of Infineon (NASDAQOTH:IFNNY) in a bid to become a major player in the market for cellular modems and system-on-chip designs.
Intel had originally planned to build future Infineon Wireless modems on its in-house chip manufacturing technology, integrated onto the same piece of silicon as the company's low-power Atom processors.
Intel eventually succeeded, but it did so by moving its Atom processors to TSMC's older 28-nanometer chip manufacturing technology. The company had originally planned to put out an integrated Atom and modem on its 14-nanometer technology in late 2015/early 2016, but this product was pushed out to late 2016.
I fear that the same thing will happen with Intel and Altera. Intel/Altera have yet to ship the first 14-nanometer FPGAs for revenue to my knowledge. Unsurprisingly, when one analyst asked management to give investors an update on when to expect a fully integrated Intel Xeon with Altera FPGA on the same die, Intel CEO Brian Krzanich had the following to say:
Our roadmap will then-we are still working on our roadmap beyond that of when do we integrate the full IP into our silicon
It should make investors just a little uneasy to know that after spending more than $16 billion of its shareholders' money on this acquisition, Intel still can't tell investors when to expect this integration to happen.
A deal looking at risk of failure
As an Intel shareholder, I would like to see the Altera acquisition play out as Intel hopes. However, with Xilinx and TSMC aggressively migrating to new chip manufacturing technologies, and with Intel's previous track record of integrating major third-party technologies into its own chips being, shall we say, less than stellar, it's tough to have confidence that the deal will create the value that Intel expects it to.