POET Technologies Inc.

As the followers of Intel (NASDAQ:INTC) know, the company intends to spend a mountain of cash to buy Altera (NASDAQ:ALTR).

The expectation is that Altera's FPGA will provide options in the Data Center that are not available with microprocessor server chips alone. The DC will require some huge and sophisticated FPGAs.

During the recent earnings conference call, Intel CEO Brian Krzanich also mentioned the IoT (Internet of Things) as a beneficiary of the Altera technology. From what I know about IoT chips is that they will have to be quite inexpensive (Cheap) chips with certain basic capabilities. These capabilities will be intelligence, embedded memory, communications ability, sensors or sensor interfaces, and security. These are the minimum capabilities. In addition, each participant in IoT will want "special" capabilities that are unique to their industry or their company.

Those aggregate requirements will cause literally thousands of similar, but different IC SKUs to be produced. Yet the cost to design and produce this variety of chips will be enormous. Many of the smaller volume IoT chip will never see the light of day due to the cost of designing integrated circuits today.

Imagine this alternate possibility: Mr. semiconductor peddler walks into "Little Guy" IoT company and says, "We know that you can't afford to make a dedicated integrated circuits for you application, but we (Intel) can give you a standard, off-the-shelf IC with all the universal IoT functions already done AND a playground of FPGA gates on-chip for your logic designer to make anything he decides to make. And other than his time to design the logic, the cost to Little Guy is the designer's salary. You get much faster time-to-market at very little cost. And you can buy as many or as few of these field programmable chips as you want… even from one of our distributors… who will teach you how to use the parts."

The major selling point of FPGAs has always been rapid time to market and the capability to do and debug the design in-house. The deterrent has always been cost. Much of the cost of FPGAs has been that they were 100% programmable and the interconnect fabric was made of SRAM (Static RAM) cells.

With the approach described above, the known functions are "hard-wired" while leaving a chunk of undedicated FPGA gates from which a custom chip can be made by the customer. Also, as I mentioned in a previous article, there is some possibility that Intel will replace the SRAM fabric in FPGAs with PCMS (Phase Change Memory/Switch) fabric at a much lower cost. Both the PCMS and partial hard-wired circuit will reduce the cost of field programmability to make it affordable for even the smallest customer applications.

So how does this affect Taiwan Semiconductor Manufacturing Co. (NYSE:TSM), you ask?

When not fighting with Samsung (OTC:SSNLF) over winning mobile chip foundry business, TSMC is making probably thousands of versions of less sophisticated chips in moderate volumes for what are, or will become IoT customers. The above "partial-custom" approach will work for the vast majority of these customers as well. The Intel partial-custom approach will lift some of this very profitable business from TSMC and other semiconductor foundries.

So, Altera will be a unique Intel answer to the foundry issue. I love it!

Buy Intel at these depressed prices.