Google versus Synopsys (EDA)
This Blog is a follow-up to my ever popular declaration that EDA is DEAD. I know comparing Google to Synopsys is apples and oranges, or more appropriately comparing apples and prunes, but the business model contrast is relevant. Googled: The End of the World As We Know It is absolutely the best book on Google to date. I also read Planet Google with much less interest.
Chapter #1 of Googled says it all with the visit of traditional media mogul Mel Karmazin to Google HQ in June of 2003. As CEO of Viacom, he was by his own admission paranoid of the competition. When Time Warner and AOL merged creating the largest media outlet, Karmazin went in search of online business partners. After spending a day with Google founders Larry Page, Sergey Brin, and CEO Eric Schmidt, and understanding the Google engineering approach to advertising, Karmazin uttered the now infamous phrase” You’re f***ing with the magic!”
At the time, Viacom was responsible for $25 billion worth of advertising sold the traditional way which is best understood via Karmazin quotes:
“I know half of my advertising works, I just don’t know which half.”
“You buy a commercial in the Super Bowl, you’re going to pay $2.5M for the spot. I have no idea if it’s going to work, you pay your money, you take your chances.”
“I want a sales person in the process, taking that buyer out to drinks, taking an order they shouldn’t have gotten.”
“That’s the worst kind of business model in the world (Google’s), you don’t want people to know what works. When you know what works, or not, you tend to charge less money than when you have this aura and you are selling this mystique.”
“Advertisers don’t know what works and what doesn’t. That’s a great business model.”
Which is a stark contrast to Google’s customer centric advertising business model. Instead of charging up-front licensing fees with no accountability for the performance of the product, Google takes a success based approach with the Adwords and Adsense advertising business model.
Pay per click (PPC) is an Internet advertising model used on websites, in which advertisers pay their host only when their ad is clicked. With search engines, advertisers typically bid on keyword phrases relevant to their target market.
Business in general is highly measurable if you have the right tools. It’s simple math, if a business invests X dollars in a product, they will get Y return on that investment, the result being a documented value proposition for that product. Google has a literal data mine for advertisers and provides free tools (Google Analytics) to measure success. So yes, the advertising business magic got GOOGLED, by making it more efficient and accountable.
Now contrast that with the Electronic Design Automation Industry and you will see a similar opportunity. Up front licensing fees will be replaced by customer centric, success based models with documented value propositions. Companies that lead this transition will flourish, followers and/or late adopters will get GOOGLED. Either way EDA as we once knew it is DEAD. Believe it.
Hello Danni,
I would love to see success-based models in EDA.
But then EDA vendors are anchored to their current models by their ‘dependency’ on Wall Street. multi-year predictable backlog you can only have with multi-year commitment based contracts (upfront licensing).
We need to break that link first before there are chances to really approach success-based models.
Thomas
Hi Thomas,
I agree, Google did this pre IPO, but at some point in time Wall Street will turn its back on EDA unless there are signs of growth.
The interesting thing about Google is that they went into the advertising business with their eyes and minds wide open. Google didn’t even have a box to think outside of.
SNPS is the most successful EDA company to date and is flat lining. Future EDA growth will be increased market share at the expense of others, not increased market. That is clearly the focus of SNPS.
Take away the EDA monitization box, open your eyes/mind, and what you will see is a more than $1 trillion industry. Life as we know it depends on this industry and this industry depends on EDA. Now monetize that.
The good news is that once SNPS kills off the competition it can change business models by brute force.
My preference however would be to partner with customers, integrate into the value proposition of the end product, and fairly monetize the mutual success. That is what Google is doing, someone needs to f*** with the EDA magic.
D.A.N.
But…but…that would mean big users could no longer extort free engineering from EDA companies, that smaller competitors could flourish by having access to EDA tools without paying through the nose upfront before they actually started selling chips, and big users would have to share profits with EDA companies. No, no, we can’t do that!! (Can we?)
Great topic, Dan. The book sounds interesting and I’m gonna buy it. Sounds like the problem is that EDA needs to better align interests with opportunities for mutual success.
One part of the problem with EDA is that a relatively small number of customers represents the larger part of the EDA revenues. Those customers are very successful at making many multiples more than their costs. They know how to successfully build a chip with the tools at hand from multiple vendors and even in-house tools. If the model changed these companies might need to provide more of their profits to the EDA cos., which they might not like. On the other hand small customers or customers who are new to chip development (perhaps in a developing country) might be more willing to partner up to achieve better chances of success and risk sharing.
>>…no accountability for the performance of the product
There is some accountability with warranties, I suppose. Also, subscription based licensing reduces the cost of switching between vendors.
While I agree that some portion of future EDA growth is in taking share from others, there are also ops for expansion: IP, FPGA, platform based design, embedded systems, and ESL.
I’m looking forward to seeing how the EDA magic get’s shaken up (or should I say f***ed up).
Well said Daniel!! The real question is who is prepared to bell the cat?
The band aid that is EDA will come off slowly.
The fundamental problem for EDA is that the NRE cost for chip design keeps going up so fewer companies can afford to do it. Companies like Synopsys whose revenue stream is tied to tape-outs are chasing fewer and fewer customers, and inevitably they have to hike prices to maintain revenue if they can’t increase market share (a death spiral). The user cost for Google keeps dropping.
EDA might be saved if the NRE costs were headed in the other direction, e.g. some new mask generation technology using nanotech that costs orders of magnitude less, otherwise it will probably see major consolidation.
While NRE costs keep increasing the FPGA option begins to look more and more attractive. Todays FPGA is what could only have been done in an ASIC yesterday. Synopsys and Mentor seem well positioned for continued growth in FPGA design starts as ASIC design starts dwindle.
I call “baloney!” (yes, I said it, “baloney!”) on the whole basis of the article: the notion that Google has a results-based business model. If Google truly had a results-based business model, it would be paid on sales, not clicks. Who cares about clicks? Customers want to make sales. But customers pay for clicks whether or not they make sales.
Google can’t control the sales process, you say? Exactly. Just as EDA companies cannot control project success. AEs in EDA watch engineers consistently follow known failure paths on projects that go far beyond tools. We see things like specifications that change in the middle of the project, a lack of complete test plans, and a bizarre unwillingness to simulate FPGA RTL code (We’ll debug it in the lab.)
There are so many factors outside of EDA tools that a business model like the one proposed here is unfair and cannot work. It’s like paying the nail gun manufacturer after the house sells.
—-
Huge disclaimer. These are my personal rantings have have nothing to do with my employer.
Have you used Google Analytics? With the Google tools a business can derive sales-per-click from the different marketing activities they participate in. I do the same with my blog.
Semi Ip companies have a success based business model, it’s called royalties. TSMC is success based, it’s called wafer pricing. Fabless ASIC companies are success based, it’s called packaged parts. FPGA’s are success based, charge $1M for FPGA tools and see how many design starts you get.
EDA sells point tools that have no clear value proposition. If EDA partnered with customers and properly documented the value proposition EDA could also be success based.
EDA must get rid of the $5B box it has created or EDA will crumble. Foundries and the top semiconductor companies will develop internal tools and we will be back where we started.
But thank you for the baloney sandwich.
D.A.N.
Great debate. I agree to a degree with Dan that we have created a $5B (and shrinking) box. But I agree with Ray that it’s not as simple as “success-based.” His analogy about the nail gun is apt; we’re selling tools, not materials. All of the examples that D.A.N. gave are for companies that sell materials — stuff that goes into the customers end product, so the volume is directly tied to the success of the customer’s product. EDA is not that way at all. It is much more like the nail gun, where a builder buys the best he can afford and then uses it to build as many houses as possible to maximize his ROI on the purchase.
I agree completely with the notion that most EDA companies have not done a good job of describing their value proposition. Many of the tools in the flow are now commoditized, much like nail guns! But just as in nail guns, there are some well differentiated products in the market, and companies that have focused on being able to articulate the value of those products.
But no matter whether we differentiate our products and figure out our value propositions, the rapid maturation of the semiconductor market will drive EDA in new directions. With fewer customer companies doing fewer chips, it is hard to see the traditional EDA market continuing to operate the way it has for a generation. I think it is possible we will see more internal development of tools, and we may see new licensing models where EDA companies partner with the largest semi companies to build custom tools. We may also see a trend at the other end toward more shrink-wrap style tools that actually work out-of-the-box without AE intervention.
Of course none of that eliminates the need to build and deliver real value; that’s just fundamental good business. But how to sell hammers to builders based on the success of their houses in the market is not clear to me, or anyone else.
On selling hammers:
Maybe you sell the hammer with a guarantee (insurance against the house’s failure). Just factor the house-failure risk into the hammer price.
I’d say you probably want to be selling the nails instead, with insurance against the nail’s failure.
If you buy enough nails I’ll give you the hammer for free
IP goes in the end product, verification correctness goes in the product. I think it’s hard to compare house development with chip development.
As David Packard said, “profit is the best single measure of our contribution to society and the ultimate source of our corporate strength. We should attempt to achieve the maximum possible profit consistent with our other objectives.” EDA needs to figure out how to contribute more with less. There are some areas where I think EDA could improve cost structures and enhance performance. For me Google disrupted the incumbents by messing with the cost structure. Heck they build their own proprietary servers to reduce the cost per search.
D.A.N. Google Analytics sales data is no different than EDA saying, “90 M chips have been shipped that were designed with our SW” Google doesn’t extract any more money that the cost of the click, and EDA doesn’t extract any more money than the cost of the tools.
The problem with the EDA industry goes beyond articulating the value, or creating a new business model. After all, a success-based business model is simply a way of getting more money for the same tools. Why would EDA customers sign up to do that? Answer. They wouldn’t (and haven’t).
The basic problem in the EDA industry can be found in the Michael Porter 5-forces analysis of our industry (http://bit.ly/8PGQzc) Here they are for EDA:
1. Customers — The customers in EDA have incredible power in certain markets, such as simple simulators, and less power in other markets such as ASIC synthesis. As the number of customer shrinks, their power will increase.
2. Substitutes — FPGA companies have a model where they give the tools for free. This is a potent substitute force.
3. Entrants — There are high barriers to entry to become a big EDA company (sales channel) but low barriers to bring new technology into the market (no capital expenditures, small, well-understood customer base, etc)
4. Competition — There is fierce competition in EDA because there are no real cost-of-goods-sold and tremendous sunk costs in each tool. Each dollar goes straight to gross profit so $1 is better than no dollars.
5. Suppliers — This is the toughest force in the industry. The primary supplier in EDA is the collection of people who work in EDA. We have very specialized knowledge in a tremendously arcane field writing and selling very complex software. So labor costs are high.
Trying to change the business model won’t affect these five forces. Any business model that extracts more money from the customer will run into the Force 4, competition. Customers simply won’t sign up for a more expensive business model.
Articulating our value will not change the five forces. The problem isn’t that our customers don’t know, or understand our value, it’s that we compete fiercely for every incremental dollar, while also competing for the small pool of people who understand this stuff.
It’s a good thing the Google guys did not read Michael Porter’s 5-forces analysis, otherwise they would have failed. Read the book and you will see what I mean. A 500 word blog cannot do it justice.
Spoiler alert: if you are a personal data privacy freak do not read the book. It will give you nightmares!
D.A.N.
Sorry. I’m confused. The Porter model works for all industries. Google’s as well. Why would they have failed by understanding the forces on their industry?
From an EDA user/ CAD guy/ EDA buyer perspective: I do not need more tools, the top 3 have all the ’same’ stuff. Digital implementation alone are 4 (+1?) suppliers, they all do work. Where is the differentiation?
The big EDA guys do not sell tools, they sell volume (all products included).
And yes, right now they try to steal customers from each other.
Since most EDA vendors insistent on their ‘flavor’ of the same standard, the tools are not compatible creating huge migration cost. Therefore the new tools must not cost a lot so that I have a positive business case despite the migration efforts.
But this is not what I am really after either, it is a short-term measure to reduce cost.
Longer-term I want the EDA vendors to provide design flow/ design methodology solutions which are not just a wrapper for their tools (reference flows/ design kits/ Lynx/ …) but are really based on tools from a variety of EDA partners. The solution in itself is differentiating.
Ray, too many EDA vendors right now talk about ‘value’ but only to argue for higher prices/ higher volume.
But value is a two-way thing, if I do not see the value/ believe in the value there is none and I will not buy at that inflated price. Without value it is just cost.
You also talk about AEs seeing all the mistakes done at IC design teams. I would think that this is partly true and this is why I argue for deeper partnerships which engage EDA vendor personnel into our IC design projects much more and hands-on. As part of the design team they can and need to speak up when they see such things. Therefore you could do make a difference, but only as a partner/ not as a supplier. We are very selective about this.
Furthermore on success-based. Success can not be attributed to a tool, otherwise I am unable to differentiate which tool really contributed to the success, and how much. Here a partner has the upper hand to really contribute through solutions and hands-on project engagements.
This however means something quite different than todays’ engagement models (consulting) and goes much more in the direction that EDA vendors and internal CAD groups partner first to create these solutions for flows and methodologies (educating the partner in the process) and then jointly engage to drive design projects to adopt these methodologies by being on-site/ hands-on and experience the IC design directly/ learn/ fix/ adapt and the loop is closed.
You say, that no EDA vendor may go success-based today. And from a generic point of view this may be true. But it is time to realize that no single EDA vendor will be able to solve our IC design problems (and no single CAD group either). So that means new and different partnerships and these partnerships need to show results/ success. Once you are that far why not incentives on success as we do internally for our design teams?
It is time for a sea-change otherwise the $5B is shrinking since IC companies still need to address their IC design problems and if they can not get help externally in a satisfactory manner I guess it is back to internal development (which today is actually the case for design flows and methodologies anyways, today the EDA guys only provide tools).
If you look at it: there are ca. 6.000 R&D engineers in EDA plus 4.000 AEs. Mostly they are busy to develop their own version of the same tool the other guy already has, little is going into internal R&D to develop new features. They all partner initially with start-ups to link to these tools. Once one of the big guys acquires the company the other companies then again have to dedicate R&D engineers to create their own version…..
On the other side there are at least 10.000 CAD engineers (central, local, embedded) within the IC companies. They are all working on the same thing: design flows/ design methodologies. This is essentially the same everywhere, absolute duplication.
Partnering these tool communities up right could provide a tremendous leverage to the much needed advancements in EDA/ CAD to address todays and future IC design problems.
As Winston Churchill said: It’s not enough that we do our best; sometimes we have to do what’s required”.
Or to paraphrase: it’s not enough that we do what is best for an individual, sometimes we have to what’s required for the community.
I suppose this “duplication” is by design. Big purchasing companies like to have multiple sources to reduce risk and cost.
Based on what you’ve said, EDA could better align with your interests by: reducing the huge migration costs, enabling differentiated multi-vendor solutions with a vibrant ecosystem, more project specific support, communication, and community. Seems that the EDA companies know this and are trying to but it’s hard to stop a moving train and get it running on a different track. Above all else the heavyweight EDA purchasers have the power to mold the industry.
A very interesting article from Mike Fazeli of Atrenta, formerly a heavyweight EDA purchaser from TI is on the Chip Design Magazine site:
What EDA Needs to Do to Start Growing Again
http://chipdesignmag.com/display.php?articleId=3746&issueId=36
My feelings exactly.
D.A.N.
I think the very basic premise of “success based biz model” should be viewed in another perspective. Sure, we buy an EDA tool by paying an upfront license fee irrespective of the project outcome – for the simple reason that in the present form, it is just a tool – and not a solution completely serving a biz objective from spec to manufacturing – as cited by Mike Fazeli/Atrenta. If we do not get results from a tool from one vendor, we switch to another – albeit not a zero cost transition.
The issue here is not a success based model or not but rather on the evolving value-proposition perceived by the chip designing company from an EDA vendor’s offering.
I see the following path for the EDA vendors
- EDA vendors’ offering to be more service oriented and I do not mean tool based trouble shooting here.
- Point tools across EDA vendors are losing their differentiation but are essential for basic design flow. I do not see chip vendors getting into developing these.
- EDA vendors will partner for point tools. These tools will evolve as per market requirements and be available on a pay-per-use model and bolstered by cloud computing.
- As a service, EDA vendors will closely work with the chip developers to see the project from specs to manufacturing. This is the service part of the “tool+ service” offering of the EDA vendors. The major chunk of the EDA vendor’s revenue will come from this service part and this will be the deciding criteria for the EDA vendors’ ranking in the chip designing world.
Meenu (www.asic-vlsi.com/blog)
Also recommend this article where John Bruggeman, CMO at Cadence, talks about two-dimensional strategy, where the other dimension is profitability. http://www.edn.com/blog/920000692/post/1590051159.html
EDA companies LOVE doing the kind of work Meenu describes, though we are often rebuffed by companies that don’t want to be “locked in” to any EDA vendor. (It seems that partnering doesn’t include commitment.)
Now, if EDA customers want what Meenu describes, and EDA companies want to deliver what Meenu describes, why isn’t this happening all over the place?
I contend that the five forces transcend us all, especially the force of internal competition in the EDA industry. EDA customers are not willing to give up the strong negotiating position they hold due to competition (hence their desire to lower switching costs) and so while they like the nice idea of partnership, they really can’t resist knocking tool prices as low as possible when the bill comes due.
I call your Five Forces and raise you Six Sigma!
Foundries will drive new EDA flows and business models. The TSMC Open Initiative Platform ( http://danielnenni.com/2009/11/02/tsmc-oip-explained/ ) is but the tip of the iceberg.
How do you truly partner with a 100% success based company using an up-front licensing model? Do you really think foundries and fabless semiconductor companies will continue to assume the risk while EDA sits in the bleachers begging for more pie?
TSMC is the 6th and greatest force…..
Remember that we really don’t have an upfront licensing model. N-year term deals ensure that the EDA company cannot walk away after the contract is signed. We need to make sure the customer is successful so we can grow the contract at renewal.
The idea of EDA companies selling licenses and walking away is so 1980’s.
The idea of EDA scaling the N-year term business model while your domestic partner (foundries) and customers are success based is laughable.
***Reference MENT’s financial highlights.
Flat growth rate companies/industries are also called life-style companies, as they support the life-styles of the executives. I’m okay with that, just quit whining about it and be honest with the stakeholders.
http://danielnenni.com/2009/08/10/semiconductor-design-enablement-ceo-panel/
Great, I didn’t know about this topic up to the present. Thanx!!