INTC will make chips for Achronix in its newest 22nm fabs

INTC and Achronix

Achronix, a small privately-held fabless semiconductor design firm in Silicon Valley, announced on Monday an agreement that it and INTC have been working on for almost a year.  INTC will manufacture Achronix’s 22-nanometer Speedster 22i, which it calls “the world’s most advanced field programmable gate array (FPGA),” using one of the company’s most advanced fabricating plants, slated to open late next year.

INTC didn’t make an announcement, but did post the news on a company blog.

On the surface this looks like a lopsided deal in Achronix’s favor.  The key feature of FPGAs is that their programming can be improved or upgraded even after the chip is in a device.  So they’re great for putting new technology into telecom or networking equipment.  Achronix claims the Speedster 22i will offer 4x performance vs. competing chips and will use 50% less power, but will cost 40% less.  A lot of this will be due to the INTC process technology.

What does INTC get in return?  The company isn’t really saying and there’s no agreement either among analysts or technology bloggers.

my thoughts

I think there’s a good news/bad news aspect to the answer, which is why INTC is mum.

A semiconductor fab in today’s world costs at least $3 billion to build.  It’s capable of producing about $7 billion in yearly output.  Who can afford to build one?  Who needs that much output?   No one–other than INTC, which has a near monopoly in processors for PCs, Samsung, the dominant force in memory chips, and maybe TXN.

To the degree I’ve thought about these facts–and I haven’t done much–I’d assumed that being big enough to continue to have your own fabs, and to relentlessly push for more advanced fab technology, would prove a critical market advantage.  But it hasn’t.  That’s the bad news.

For most semiconductor design firms, fab ownership is out of the question financially.  They turn to third-parties, the so-called “foundries” like TSMC or UMC, to do their manufacturing for them.  For the fabless firms, then, manufacturing is not a competitive advantage.  What separates winners from losers among them is the flexility of products, the ability to do more creative things within the confines of a standard fabrication technology.  As a result, the fabless industry has made a virtue of necessity and learned how to do a lot more with a lot less.  Arguably, the foundries have done the same.

In other words, the rules of the game for semiconductor device makers–and especially for chips in mobile devices–have changed in a way that puts INTC at a disadvantage.  More bad news.

Where’s the good news?  –it’s that INTC understands what’s going on and has decided to try to change the game once again.   Companies like Achronix get manufacturing prowess that will give their chips features the competition may not be able to match.  INTC learns the mindset of the designer of products for the mobile universe.  Together, the two firms may be able to create system-on-a-chip products that make future generations of the Atom processor more attractive to makers of cellphones and tablets.

My guess is that this move isn’t at all about INTC itself becoming a foundry.  It’s more about changing the chip design culture within INTC to the point where it can meld the best of the fabless world in with its own traditional skills.

If I’m right, this is a real sign of management strength.  It might work.  I don’t see anything in the company’s stock price that suggests this possibility is being discounted at all. Let’s see what happens.





2 responses

  1. That’s an interesting take on Intel’s motivations. If I were to write up a list of the boldest/most interesting of fabless start ups, Achronix would definitely be near the top. Intel has been in a bit of an x86 rutt for the past few years. If you look at most of their research endeavors like Larrabee, they always start with a stripped down x86 core as their basic building block. x86 is a bloated instruction set and x86 cores are designed around latency hiding rather than throughput maximizing (GPUs’ forte). It’s that difference in compute strengths that makes heterogenous computing so enticing.

    Anyway, let’s hope this is Intel’s attempt to get some fresh ideas.

    • Thanks for your comment. I started looking at INTC again after a long absence (I covered the company as an analyst beginning in 1981?) a year or so ago. I was as much interested in getting a read on the PC business as I was in the firm. It seems to me that the current management understands the company has a deep problem and is trying to fix it. The stock price makes it clear, however, that Wall Street doesn’t believe it will be successful.

Leave a Reply

%d bloggers like this: