"I've been told by quite a few people not to discuss this topic publicly," says Joe Costello, CEO of one of the world's largest EDA companies: Cadence. "Other EDA CEOs said not to discuss it because it'll only anger customers. The PR people carefully tried to tell me to avoid this topic altogether because of Ônegative customer reaction.' My big customers warned me not to discuss this with other customers because they probably won't like what's being proposed. Even my lawyers said not to talk about this because it might violate some antitrust law—and that the customers won't like it," said Costello in a recent phone interview.
Costello isn't trying to publish some closely held business secret like the formula for Coca-Cola. No illegal diversion of federal funds or North Korean nuclear-weapon plans involved here. Joe just wants to challenge how the EDA industry views its software pricing, support costs, funding new software development, and paying for upgrades. If the industry won't meet his challenge, he at least wants the engineering community as a whole to reexamine this topic seriously.
Turnkey pricing
Historically, EDA companies sold hardware and software together as a turnkey package. Early on, it was much easier to convince a purchasing manager to fork out the $30,000 in capital expense for a workstation that he, the accounting department, and the IRS could physically put their hands on. Even nontechies knew that workstations had fairly well-defined product life spans that fit rather nicely into standard IRS-acceptable depreciation schedules. EDA software, on the other hand, wasn't something you could touch, was easily copied, and had a problematic quasi-indefinite life span that wreaked havoc on accounting depreciation tables. Let's just ask those EDA salesmen to throw in the software for free, OK?
Of course, everyone also agreed with the concept of paying EDA companies "protection money," in the form of annual maintenance fees of 12 to 15% of the purchase price, to keep hardware (and software) running smoothly. This bought the right to phone the EDA company when things didn't work—and even have technicians sent out if things got really bad. You need experienced hands to keep the electronics functioning. And software "repairs" weren't real anyway, just an occasional tape sent for free from some address in California.
After a few years, better workstations emerged that made old workstations look, well, old. "No problem!" says the company accountant, "Just buy new workstations. I'll depreciate all the old ones!" Management then, to the glee of the EDA vendor, approved big bucks for the purchase of new EDA workstations—and new EDA software.
As time went by, the EDA industry changed. Eventually, the EDA vendors got out of the hardware business altogether because customers wanted one workstation with one operating system that a variety of different EDA and gaming software packages could run on. Also, accounting departments became more sophisticated on how to handle software purchases. The trick is to buy EDA software the same way we bought workstations: Purchase price plus the annual protection money to keep those tapes arriving from that address in California!
Here's where Costello stepped in with some observations: "This business model works great for the first few years if you have a hot product that's flying like a rocket ship to the stars. You make your money selling licenses for $25,000 a copy. The 12% annual maintenance fee rarely covers the actual expense of staffing a hot line plus development of the next generation of tools; no one's complaining when you get 33% pretax profits on sales.
"The gotcha comes when you've sold your hot product to customers that expect the next generation of the product to be covered by the 12% annual maintenance fee. If you look at it, you see it's a bankrupt strategy because you never get the influx of cash that used to come when the customer bought new EDA hardware and software in the turnkey pricing model.
"Instead," said Costello, "customers develop elaborate schemes to get maximum use out of EDA software with an eye on paying minimal dollars to the EDA companies. Exemplifying this are customers who demand floating licenses in lieu of buying extra fixed licenses. This absorbs tremendous demand for EDA software while successfully keeping money out of the EDA vendors' budgets. I even have customers who are talking about floating licenses across T1 lines so they can run one license over three rotating world shifts: first in Europe, next in the US, then in Japan... and then back to Europe! Where does the EDA vendor win—or possibly break even—with this business model?"
Costello went on to describe an industry scenario where customers who pay maintenance fees establish a double standard, which I witnessed a few months back: In a small-scale Cadence customer rebellion on the Internet news groups comp.lang.verilog and comp.cad.cadence, engineers complained about being asked to pay $10,000 to upgrade their Verilog-XL license to Turbo-Verilog. Established customers had no qualms with Cadence's charging new customers the extra $10,000, but, because they paid maintenance fees all these years, the long-term customers felt the upgrade should be free.
The double standard that Costello pointed out is that a completely new company like Chronologic can offer a Verilog upgrade for a higher price and customers don't balk because they're buying a "new" EDA tool. "This dooms all EDA companies to being purely one-trick ponies," Costello said, "where new companies cycle in to supplant the old companies purely out of a poorly thought-out EDA-industry business model."
A secret marriage
Before you get teary-eyed and pull out your corporate checkbook to support the EDA vendor of your choice, there are some other factors an EDA consumer should be aware of. During the interview, I did some back-of-the-envelope calculations to figure out how much money Cadence made from selling Verilog. I came up with a total of $250 million for a 7-year period. Verilog was a real cash cow for Cadence. Yet, five guys in a start-up called Chronologic wrote a screaming Verilog simulator in 15 months.
What's unstated in Costello's overall observations is that he's seeking an unwritten contract that, once you make a significant purchase from an EDA company, you are making a commitment to that vendor for better or for worse. This means that you're giving this EDA company control of your design process, even though its goals may not always match yours.
With a quarter of a billion dollars in Verilog income, Cadence could have easily spun off a compiled Verilog project. Instead, it used that cash to finance all sorts of projects, many of which were against customers' own vested interests! That is, a customer that has tens of thousands of man-hours invested in using Verilog design methodology isn't particularly well served when Cadence takes Verilog income to finance a VHDL simulator to sell for new-license income. Or how about the Synopsys/Cadence customers who see Cadence Verilog dollars going into writing a competing synthesis tool? (Synopsys is no saint either—it used Verilog-synthesis dollars to finance a VHDL simulator and an ATPG tool, among other products.) The point is that sometimes big EDA company's vested interest is to sell any type of license to keep cash flow positive.
The positive side of helping small EDA companies is that virtually all of your EDA dollar goes to where you think you're putting it. That is, if you buy an ATPG tool from Sunrise Test, your money most likely isn't going to be financing a new schematic-capture tool, because the company is concentrating on keeping its ATPG tool competitive.
To remedy the situation, Costello is experimenting with spinning off companies for new products, as opposed to keeping them an incorporated part of Cadence. Instead of absorbing Comdisco and newly purchased Redwood, Cadence is merging the two companies into one business unit named Alta, which will retain its own sales staff, R&D team, and support group.
A second way Costello is thinking about remedying the problem business model is to adopt the PC software pricing-and-services way of doing business. Customers buy a specific version of an EDA tool. If they want an upgrade, they pay for it. Also, customer service is broken into various support levels—all based on what the customer pays for in support.
Costello closed the interview with an interesting statement: "Although absolute dollars in the EDA market have grown steadily over the past seven years, you'll find that the percentage investment by EDA consumers for EDA tools has decreased by 50% in those same seven years. Seven years ago, electronics companies used to spend 2% of budgets on EDA tools; now they spend only 1% of their budgets on EDA tools."