Reading Jim Turley’s article about business models got me thinking about FPGA and EDA companies.
Yeah, I know. What’s new?
We’ve talked a lot about how EDA companies struggle to find a business model that earns them their fair share of the loot that comes from electronics. The full-fledged, modern EDA industry has been around for about three decades and… they still drive a huge portion of the technology while reaping a tiny fraction of the rewards.
The fundamental reason for this, I believe, is that they’re selling software. Since software is free to produce (but certainly not to design), there is always the temptation, when faced with competition, to drop the price. Software development costs are, of course, amortized over the number of copies that are sold. For EDA, those costs can be high because EDA software is some of the most complex stuff on Earth. For EDA, the “number of copies” is very low because, compared with the number of people that need “Windows” the number that need post-layout verification for 22nm ICs is, well, zero, really.
As we learned in math class, when the denominator gets close to zero, the result gets big. What’s more, the cost of developing EDA software is trivial compared with the cost of marketing, selling, and supporting it. If you want to know why, you can get a glimpse by reading my “Par for the Course” blog post.
So – we have these competitive pressures pushing the selling price down, and this unfortunate ratio of development and marketing costs pushing the cost up. For those of you without an MBA, hang on here – your engineering math will suffice. When you subtract the cost from the price, you get – idiots like Carl Icahn knocking on your door saying you don’t know how to run your business.
EDA companies can keep their prices up to a reasonable level when they have a unique, important technology with no competitor (this never lasts long), or when they, as an industry, can sell a particular product based on fear. “Don’t want to lose millions of dollars in a re-spin? Better buy our tool.” Even though the primary purpose of EDA tools is productivity and higher-quality design, those factors have failed time and time again to extract value from the EDA customer base.
EDA: Want to design your next chip five times faster?
EDA: How about saving 75% on power consumption?
EDA: Hey, if you don’t use our tool, your design will need a re-spin…
Customer: Whoa! What? How much does it cost? Wait, I don’t care. Give me six copies, no… seven! Charge me extra, too, I wouldn’t want it to be my fault for trying to skimp on verification budget.
This happens for several reasons. First, the budget for engineering salaries in most companies comes from a completely different place than the budget for design tools. If you can save $1M in engineering time by buying a $100K tool – not interested. The two budgets are unrelated. The engineer doesn’t really always want to make his job easier either. You wouldn’t want to buy a tool that made you redundant.
FPGA companies spend about as much developing EDA tools as EDA companies. Both Xilinx and Altera are rumored to spend more of their engineering budget on software tool development than they do on FPGA design. Really? Yep, really. What’s more, more of their support effort goes to supporting their design tools than their silicon. Hmmm… what do you call a company that spends most of its engineering and support resources developing and supporting EDA tools?
I’d call it an EDA company.
What do you call an EDA company that gives its tools away almost for free, and derives its revenue from a percentage of highly-popular high-margin chips? I’d call that an EDA company with a clever, successful business model.