Microchip and Atmel

10/26/2008 12:17 AM EDT

With all due respect to Bernie, who really does know this industry, I disagree.

Microchip is an interesting and somewhat unique company. While most technology startups are founded by techies with a vision, Microchip came about when a group of venture capitalists bought General Instruments' line of processors. GI had all but failed trying to peddle these parts, which wasn't all that surprising considering their frankly lousy architecture. Some of us engineer-types chortled: only dumb old VCs would buy that brain-dead and clearly obsolete design.

We were wrong. Thanks to brilliant sales and marketing, a relentless focus on driving prices down, and a commitment to proliferate the parts into hundreds of flavors, the PIC went on to be one of the most successful microcontrollers ever. Microchip was perhaps the first to exploit, on a mass scale, the idea of putting a little bit of processing power everywhere.

In the early 90s, when in one product we found our Xilinx FPGAs failed to boot reliably, we stuck a PIC on board simply to watch the boot sequence and issue a reset in case it didn't succeed. PICs appeared in sneakers with blinking LEDs. To date over 5 billion have been shipped.

So these are smart people and I've learned never to discount any move they make. If they want Atmel, I have no doubt they know exactly what they are doing.

I was surprised when Microchip's based their 32 bit offerings on the MIPS. There's probably some clever business reasoning there, but acquiring Atmel will give them the ARM products that dominate that segment of the industry. What gets interesting is when we stop thinking about microprocessors and focus on peripherals.

A key strength of the PIC line is the huge range of I/O that's available, and the stunning numbers of PIC variants that exploit those devices. Suppose Microchip offers Atmel-derived parts with PIC peripherals? Few of us really care much about CPU architecture in this C world.

Well written code is pretty easy to port from one micro to another. But that's not true of the I/O. We invest a lot of time into getting our drivers right, and do design the hardware component of our products around a part's peripheral mix. This acquisition could give Microchip's customers a wide range of P&P (processor and peripherals) options.

Both companies sell low-end processors targeted at, in many cases, extremely low-cost applications. I believe that as costs continue to fall the number of products that benefit from a little bit of compute power will rise sharply.

Though not a glamorous market segment, expect plenty of growth. The acquisition will give Microchip both the 8051 and AVR architectures. While the 8051 is nearly as brain-dead as low-end PICs, it has a huge following. The AVR brings the company a more modern CPU that addresses both low-end and more demanding requirements.

The deal's numbers are certainly interesting. Microchip will put in $1.3B and ON another $1B. For that ON gets Atmel's nonvolatile memory business which does about $400M in sales. Microchip also hopes to sell off Atmel's $500M ASIC products. Suppose they just get a half-billion for that. Then the net cost to Microchip is $800M, for which they get Atmel's $700M micro business, a great product line, and some very smart engineers.

What do you think? Is this acquisition a shrewd move or a dumb idea?

Jack G. Ganssle is a lecturer and consultant on embedded development issues. He conducts seminars on embedded systems and helps companies with their embedded challenges. Contact him at jack@ganssle.com. His website is www.ganssle.com.