EDA Has a Value Capture Problem — An Outsider’s View

By Liyue Yan (lyan1@bu.edu)

Fact 1: In the Computer History Museum, how many artifacts are about Electronic Design Automation (EDA)? Zero.

Fact 2: The average starting base salary for a software engineer at Netflix is $219K, and that number is $125K for Cadence; the starting base salary for a hardware engineer at Cadence is $119K (source: levels.fyi)

Fact 3: EDA industry revenue has been 2% of semiconductor industry revenue for over 25 years, and only recently climbed to 3%.

Part 1: Overview 

Starting Point

I started my inquiry on EDA’s value capture issue as a research project, puzzled by the fact that this critical technology has historically captured only 2% of the revenue generated by the semiconductor industry. As an outsider, it was not obvious whether this proportion is reasonable or not. While the question of how exactly EDA is undervalued or overvalued (and under-charging or over-charging) warrants a whole other discussion, it is not surprising that EDA folks are wanting a larger share. When I was strolling around conferences pitching this 2% number to engage people in my research, the responses converged:

When asked, employees from the bigger vendors, particularly senior engineers, expressed strong opinions that the technologies are under-appreciated. They believe that the whole industry should get paid more and they should get paid more. “10%!” some said. Well, that might be too greedy. Some of the engineers questioned what the salespeople were doing exactly that the industry did not get higher revenue. In general, employees from bigger vendors can’t discuss details but requested me sharing my findings when they are complete. Because they want to know. And everybody wants to know, but nobody talks.

Smaller vendors (salespeople, engineers, founders, sometimes that one person playing all these roles), uniformly pointed the finger at the big players: “It is them! Those companies squeeze us by keeping their prices down. It is so unfair. Not only do we suffer, but the whole industry is also underpriced! You should ask them what they are doing!” The expression, the tone, and the structure of their blaming account are shockingly similar, as if they were rehearsed from a secret union whose sole purpose was to recount their suffering to each other. While small EDA companies have no problems sharing with me, they believe they are not so relevant as the industry-level revenue and profit are driven by the big three.

There is an advantage as an outsider. My ignorance provides me courage to throw any random question to any random person who is willing to catch it. When I threw my question at an executive panel, Joe Costello, former CEO of Cadence, responded along the lines of “Profit margin? I don’t think about it. I think about the value we provide through our products. I always believe that if we bring value to the customer, then we will get value.” Oh no.

Wally Rhines, former CEO of Mentor, gave a milder response consistent with what he wrote in his book, which is that he believes that EDA has a healthy long-term relationship with the customers: “I’m convinced that salespeople for the EDA industry work with their semiconductor customers to provide them with the software they need each year, even in times of semiconductor recessions, so that the average spending can stay within the budget” (Predicting Semiconductor Business Trends After Moore’s Law, Rhines, 2019). After a short conversation with Wally, I am convinced that he is a strong market believer, believing that we are close to a good equilibrium (and if not, the market can always sort itself out with any temporary deviation). Except that, first, there can be multiple equilibria, and we don’t know whether we are in a good one that sustains EDA’s innovation, and second, even if the invisible hand can work its magic bringing us to a better point, who knows how long it would take. In the long run, market will correct itself, but “in the long run we are all dead” (Keynes, 1923).

So, it seems that EDA people are thinking a lot about value creation but little about value capture, expecting customers to automatically appreciate their products through paying a fair amount. It turns out I wasn’t imagining this. Charles Shi, senior analyst at Needham, pointed out at DAC 2024 that many EDA folks believe that if they create value, then they will automatically get value, and unfortunately, that is not true. So, I set the goal to understand why EDA got a constant 2% share of semiconductor revenue and how this number was reached. We know in theory this outcome is due to a combination of macro-level market structure and micro-level firm practices, but I would like to know what those practices are and how they contribute to value capture. More specifically, I want to know how firms are selling to maximize their gain. The next question, then, is whom to talk to and what to ask, which is essentially the entire process. The list includes salespeople, engineers, CAD teams, and buyers.

Jumping to the end—a preview of findings

After talking to more than a dozen decision makers, formally or informally, answers quickly emerged. Even if the current number of participants may not be enough to guarantee an academic paper and the project is ongoing, I thought it would be interesting to share some key findings. Most of them should not come as a surprise. Think of it as a mirror (perhaps a slightly distorted one) – from time to time we do need to look into one, though the observations there should not be terribly unexpected.

 

A quick preview of some key findings:

1. Professional negotiation teams are often used by the buyers, but not by the vendors.
2. Given the long contract terms, vendors cannot shop customers easily. That means, at contract renewal times, customers have the leverage of alternatives, whereas vendors do not.
3. Negotiations are almost always done by the end of fiscal year or fiscal quarter, even for private vendors. (Buyers insist this is always the case, while some vendors deny it.)
4. Heavy quota pressure for sales potentially contributes to heavy discounts.
5. Customers are only willing to pay a small fraction to small vendors and startups for products that are similar to or even better than those of big vendors.
6. Bundling practice erodes monopoly rents, giving away value from the better products.
7. A small number of large customers often account for the majority of a vendor’s revenue.

The above points are mostly around contracting and negotiating, with some around structures and others around practice, and none are in favor of EDA vendors. Then there are some more positive factors:

8. Historically sticky users.
9. Low competition pressure from startups.
10. Customers’ product search is only triggered by pain, never by price/cost.

The seemingly good news for EDA, of system companies as new customers, may not be too good after all. We tend to think that they have bigger budgets and more generous, therefore increasing EDA’s profit margin. Well, they also have shorter contracts in general, training younger engineers who are more comfortable with switching tools. This increases their bargaining power. Additionally, less experienced users require more technical support, increasing costs for the vendors.

In all, industry structure, business model, and contracting practice combine as driving factors for the value capture, which I unpack in the next part.

Part 2 What I Have Learned

What business is Electronic Design Automation (EDA) in?

One can categorize a business through various lenses. To outsiders, one can explain that EDA is in the software industry. To someone who is interested in technologies, one can say EDA is in the semiconductor industry. If I were to explain it to business or economics researchers, I would say that EDA supplies process tools for chip design.

Why does it matter what business EDA is in? First, research studies suggest that stakeholders evaluate products and businesses through the lens of category, in plain words, putting them in a box first so they can be more easily understood and compared to similar offerings and players. If a business cannot be understood, i.e., put in a box, then it risks being devalued. And in this case, if investors and analysts do not know the proper reference group for EDA, then they would not cover it (or provide buy recommendations), and a mere “no coverage” from analysts can negatively affect stock market evaluation. At least that’s what existing studies say.

EDA is really one of its kind, and a tiny market. So, what reference group can an analyst use? What other stocks is the same analyst covering? CAD? The customers and downstream structure can vary a lot. Semiconductors? Yes, they are often covered by the same analyst, but the semiconductor industry is really not a good reference, except to the extent that their revenues are co-influenced by the downstream applications. Who wants to cover it? Nobody, unless you have an EE-related background. So what box do stock analysts put EDA into? The Black Box.

Willingness-to-pay is the second reason one should ask what business EDA is in. Value is in the eye of the beholder, and we should really ask how hardware designers feel about EDA.

The value of a business is often categorized into gain creators and pain relievers. EDA is certainly not the former, which is often driven by individual consumptions that chase temporary joys. People say businesses that leverage the seven deadly sins are the most profitable. Amen! Pride, envy, and lust – social media and cosmetics. Gluttony – fast food, alcohol. Sloth – delivery, gaming, sports watching, etc. Rest assure that EDA satisfies none of these criteria. EDA is not consumable. EDA is not perishable. And EDA is not going to make the user addicted.

So, EDA is more like a pain reliever. Well, if not careful, some engineers may even assert that it is in the “pain producing” business: “Ugly!” “Frustrating!” “Outdated.” “Stupid.” You can hear the pain. But when I pointed out that perhaps the alternative of no tools is more painful, there wasn’t much of an argument. One issue is, we don’t know the counterfactual enough to develop better appreciation. You never know what you have until it’s gone.

All of the above suggest, it is naturally difficult for EDA tools to price up, even without challenges in competition, business model and business practices, which we will dive into next.

But perhaps the discussion should first start with some positive notes. There are a few factors that work in favor of the big EDA vendors, which is 90% of the market:

Sticky users. IC designers are constantly under time pressure. Changing tools means adjustment and adjustment means loss of time. No one wants to change a routine unless it is absolutely necessary, which means EDA vendors can get by as long as they are not doing a terrible job.

Little competition from startups and small players. There were many startups that outdid incumbents and won market share in the past. That time has gone. A combination of lack of investors, increased complexity of problems, and fully exercised oligopoly power has led to a decline of EDA startups. Those few small or new players have been taking the scraps: providing solutions to peripheral problems, taking only a fraction of the price that a big vendor would get, or earning nothing at all if the big vendor decides to provide the competing tool for free in a bundle with other products.

Customers’ product search is only triggered by pain. Customers do not initiate their search for alternative tools just because the existing ones are expensive. That means, there isn’t much price competition pressure once you’ve got the account. But this also means, the pressure is on the initial pricing. Once you lock the customers in, as long as the pain is manageable, the accounts stay.

To the best of my understanding, EDA vendors are leveraging the above factors fairly well, especially factor No. 2, squeezing the small players (which is not without costs). However, there are even more factors that negatively affect the industry’s value capture, among which I rank the incentive design and pricing the top culprits.

Quota

Here goes a story:

“It was 2010, Sep 3^rd, just one month away from Sep 30^th, the fiscal year end of Company X. The corporate management team sets a goal to book $300 million for next year, and that translates to many sales quotas for its foot soldiers. John works in sales, and his job is to sell all types of tools, to either new or existing customers. John has a yearly quota of $2 mil. He has managed to book $1.4 mil so far, and he has a chance of booking a few new customers for $850k. But he missed the quota last quarter, and if he does not deliver this time, he will be let go. John has a mortgage with a $4k monthly payment and two kids, one is 3^rd grade and the other just entering school. John’s wife has not been particularly happy about his recent working hours and heavy travels.

10 Days later, John closed a deal on Sep 13^th, bringing his balance down to $300k from $600k. John expected to close the next new customer Company Elppa for $450 – $550k for 20 licenses. The negotiation went on for a week, and the customer stood firm at $350k, claiming that’s their budget. By Sep 21^st, John was stressed and asked his manager if it is possible to lower the price to 350. The manager nodded yes to 400, as he was trying to meet his own quota. The eventual deal was $400k with promised additional support to the new customers. John was relieved. The customer was happy. The management was ok with this season’s performance. The licensing price for Elppa effectively dropped from the target $550k to $400k, by a percentage of 27%. Hopefully this gap can be closed in the future.

Two years later, John moved to Company Y. His accounts were left with Ron. Ron couldn’t find a way to increase the price by more than 5% with Elppa since Elppa was also trying to buy more licenses. Ron ended up charging a 5% price increase for 10 more licenses. The gap was never closed.”

This is absolutely a made-up story. Except that there was a time an employee must be let go if he has not met the quota for two consecutive quarters, and that customers do always want to negotiate by the end of a quarter to leverage this quota pressure, and that it is difficult to increase price for an established account. And the mortgage probably is four thousand a month. Rumor has it some tools are given for free in the bundle to attract clients, and the units for those tools have become unsustainable as a result. Rumor also has it there was once a 98% discount from the listing price at a desperate time (compared to the usual 60-80% discount rate). While the discontinuous incentive design, the quota system that are used by all vendors, can increase sales on some occasions, it can also point effort in the wrong direction, especially when the quota is always the total dollar amount.

The story depicts some issues that are essential to EDA’s value capture problem, given the current business model.

Negotiation

Most customers make multi-year orders. This type of contracts brings comfort to both sides. Customers can now focus on their work and develop routines, and vendors have the confidence that they will not starve in the next season. But this also means, customers’ budget is mostly locked with their existing contracts. In addition to limited customer turnover, EDA vendors can also expect few new clients. This puts vendors in a weak position in contract negotiation. With potential new customers locked into their existing deal, vendors have few alternatives equivalent to any existing customer at the time of negotiation. In contrast, customers can always switch vendors, despite the difficulty in executing a switch. This results in imbalanced negotiation power. This imbalance is particularly true when: (1) the customers have already been using competing tools; (2) the customers are young and adaptable; (3) the customers are big.

Customers negotiate with a few EDA vendors; vendors negotiate with hundreds of customers. In these repeated negotiations, big customers largely use professional negotiators who do nothing but negotiate contracts, whereas vendors have their street-smart people-orientated engineer-turned-sales. No matter how awesome these salespersons are, it is hard to argue there is no skill difference, not to mention the quota pressure. And yet, these are the only moments that any value created by EDA is cemented into revenue.

Bundling

Bundling is often considered a highly effective tool for price discrimination, able to maximize value capture by hitting each customer’s willingness-to-pay (WTP) for a set of products. It works because each customer has different levels of WTP for any specific product, but the variance is largely cancelled out when a whole bundle of products is offered together. The price for the bundle is usually fixed.

But bundling in EDA is nothing like the instrument used in typical pricing strategy, despite sharing the same name. In fact, there is no point in using the word “bundle” at all; a more accurate description would be “there is no fixed price, you can buy anything you want, we hope you try our other products, we will give you a discount when you buy more, and we just charge them all together.”

So how exactly this practice is harming EDA businesses?

One case is when the client wants to purchase a competing product of A, possibly superior, the sales may say, if you buy our B and C, we can give you A for free!

This sort of “bundle” is essentially bad pricing competition that squeezes out the small players, and big vendors themselves have to bear the consequences of low pricing for a long time due to locked-in contracts, as we discussed earlier.

The bigger vendors also did not develop mutual forbearance against each other to avoid competing on the same product features. Business school classrooms like to use the Coca-Cola vs. PepsiCo example through which students develop the understanding that both companies are so profitable largely because they do not compete on price, but on perceived differentiation through branding. It is possible to have different strengths instead of all striving for a complete flow.

The “bundle” practice also obscures the value of each tool. Instead of using A1 to compete with A2, a company can use A1, B1, C1 together to compete with any combination of competing tools. When you offer A1 for a low price, you are effectively eroding the profit from B1 or C1, even if perhaps one of them has monopoly power. As for how much value is given away, only these vendors can tell with their data.

Industry structure

One determining factor here is the industry structure of EDA and their downstream. Just the big three EDA firms account for 90% of market share. That market concentration should come with high market power. Well, if you look at the customers of any big vendors, the semiconductor companies, perhaps two or three can make up 70% of their revenue. Not so much power for the vendors after all.

Business model?

Many think the business model is the problem. I am not so sure. The common argument is that even though EDA provides critical tools to enable any downstream chip design and applications, the current business model does not allow it to get a fixed share of the final value created. Some others say, unless EDA can find a way like Silicon IP, charging fee per production unit, the business won’t be sustainable.

Let’s take a look at the underlying logic of these arguments. This is equivalent of saying, the university should charge students a fixed percentage of their future incomes; otherwise it is not fair, since the university degrees enable their careers. Or, power tool manufacturers should charge customers based on the value of house they build. Even better, the coffee you bought this morning woke you up and help you close a $2 million deal, so pay the coffee shop 1%. It does not make sense.

But this is how people are thinking, and the pricing logic for EDA vendors follows: We guess the customers’ revenue and use that to price discriminate; we believe we should charge customers with simple designs or low value application lower fees and increase the number when the design complexity and downstream revenue increase. Whether consciously or not, many vendors are on the same page in this pricing logic.

There are two parts to this logic. One line of reasoning is that once a tool is provided, it can be used for 100 hours, or 1000 hours, and of course they should charge more for the ones used more. This part seems somewhat reasonable because the vendor is essentially providing more tools for the heavy users, even though there is little or no additional cost incurred by the vendor. A solution is cloud and usage monitoring, which could be implemented with time.

The other argument is that some customers use the tool to produce 1 million chips whereas some others only produce 10 thousand. Shouldn’t one charge the former more per license, given that the tools enable them to achieve a higher revenue? I do not believe so. The tool should charge a customer up to the added value – in this case, how much cost it saves the customer compared to the alternatives, which also decides the customer’s maximum willingness-to-pay – and charge at least its own production costs (in this case, its own costs for maintenance). As for where exactly the price lands in this range, it depends on competition and negotiation, which are discussed above in the industry structure and negotiation sections.

So, what are the possible remedies that can improve EDA’s value capture? 

Part 3 Remedies

Three remedies are proposed:

1. Better incentive design
2. Smaller accounts (but more of them)
3. Stay in different lanes

The first proposed remedy focuses on the incentive issues in negotiation, the second on negotiation power, and the third on the willingness-to-pay and negotiation power resulting from market structure.

Better incentive design

The only opportunity EDA has to capture its value is the moment a deal is made. Who makes the deal and how it is made determine two years of revenue from that contract. I am not an expert in contracting but it does not take an expert to see that the total-dollar-amount based quota distorts incentives.

Here is what big vendors could do without changing much of their existing business model:

Hire one or two experts in sales incentives design. Ideally, they have a PhD in Economics or Finance. They can do data work and some simple formal modeling. They have experience or at least appreciation for experimental and behavioral economics. They could be just out of graduate school, or currently working in Amazon dealing with pricing models that never leverage much of their real training. Currently, the big vendors employ hundreds of engineers with PhDs, but only hire a few staff with a BS or MS for performance analytics and pricing. No, let an Econ PhD work, and they will be worth more than 4 bachelor’s.

It is best to hire them directly instead of sourcing from consulting firms, as the sales performance data needs to be reviewed constantly, and incentive schemes may need to be adjusted. However, it would be reasonable to first evaluate whether there is a real need through economic consulting groups.

I imagine EDA firms could use the same type of people for pricing.

In any case, any incentive scheme should consider a quota not just based on the total dollar amount. Realized price per license needs to be incorporated into incentive formulas as well.

Smaller accounts

This suggestion is not about getting new and small contracts. It is changing the situation where two or three customers make up 70% of an EDA vendor’s revenue, so it has little negotiation power in each contract. Obviously, EDA cannot change the semiconductor industry structure, but it can change the concentration of its contracts. That is, breaking down bigger contracts to smaller ones so each one is not as critical. Essentially you are treating one customer as twenty customers. This is also aligned with EDA’s price discrimination strategy based on complexity and the total value created. Different projects of a customer can have different levels of complexity and produce at different scales.

What’s the benefit for customers when this can increase their contracting and negotiation costs, not to mention the vacancy time of each license (though we can expect that the pricing can eventually be based on actual run time)? Clean separation for budgets allocated to different products. Accountability at the project or business unit level.

Stay in different lanes

Stay in different lanes, or at least have separate strengths. Save the effort in working on one’s shortcomings and divert those resources into innovation around one’s unique strengths. This also extends to vendors’ recent diversification into other areas, such as IoTs and Automotives. With new realms open for innovation, this could be a time to a reset the mode of competition.

The above remedies are brief by design. They are not comprehensive solutions, but practical ideas meant to prompt a rethink of how EDA approaches value capture. EDA’s non-stop innovation is vital, but sustaining the field and keeping it attractive to talent requires taking value capture just as seriously.

Professor Liyue Yan is a researcher of strategy and entrepreneurship at Boston University’s Questrom School of Business. Her work examines strategic decision-making and entrepreneurial entry, with ongoing projects focusing on the Electronic Design Automation (EDA) industry.

Also Read:

Lessons from the DeepChip Wars: What a Decade-old Debate Teaches Us About Tech Evolution

AI RTL Generation versus AI RTL Verification

PDF Solutions Charts a Course for the Future at Its User Conference and Analyst Day

 

 

Share this post via:

• 
• 
• 
•