Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/eda-ip-business-model-debate-daniel-nenni-versus-aart-de-geus-discussion.378/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

EDA / IP Business Model Debate: Daniel Nenni versus Aart de Geus (Discussion)

new EDA business model needed

Too bad Aart does not like success-based business models. In an industry where many variable engineering/ sales salary components are based on achieving success on certain goals, why should this not apply to entire business models as well?

It is a shared risk/ reward model where partners work together to jointly achieve success. If positive it is a win-win. If negative: well, each partner suffers some loss.
Especially if EDA sees itself so instrumental in enabling the entire IC industry, they should not be shy to demonstrating their capabilities by engaging in success based models.
 
It is very easy to see the flaws in the current models. The larger public EDA companies do not take a long term view. They simply seek to glean what the customer's current budget is and then give whatever tools the customer wants in order to get the whole budget. Much like the customers learned in the early 1990s to wait until the end of the quarter for a better discount, they have now learned to keep shrinking their EDA budget because their big EDA supplier will take it, whatever it is, anyway. It is comical to think that at the same time the big EDA companies adopted this model they were training their sales teams on "value-based selling". The "all you can eat" model doesn't focus on value, it focuses on taking all of the current year budget.

Success based models can sell on value, but their are several drawbacks. It takes years to reach commercial success on a semiconductor project. It is also hard to measure the contribution of an EDA tool to the ultimate success of a project. Really, how much does the P&R tool contribute to the success of an iPod Nano? When IP companies switched to this model they needed many years to get paid on the royalty side. This would be a very difficult transition for a public EDA company. Artisan Components did this by selling for value and taking a significant upfront license fee in addition to the royalty component. Artisan also stubbornly stuck to this model, did not discount much on the upfront license fee, and was eventually rewarded for it. It also helped that eventually Virage Logic went to similar models as well. This type of change is perceived as very dangerous for public EDA companies, in part because there is so little differentiation in the middle of the IC physical design flow. Customers can be ambivalent as to which tool to use. When you are in that situation all you can compete on is price, and support. The problem is the all-you-can-eat deals apply this logic across all the tools in the salesperson's bag, not just the ones that have little value.

Whatever model is chosen, more effort and discipline needs to be applied. On average, I think that the big companies have tried to improve margins by squeezing the costs, and indirectly the talent, out of their sales channels. In real terms, EDA sales people are typically paid less now than they were 15 years ago. No one is writing huge commission checks anymore. You don't see as much marketing effort on technology of individual tools anymore because the big companies don't sell that value like they used to. If the industry simply did less bundling, per license pricing would increase, as would sales costs. I think margins and revenues would in fact grow. Is the industry willing to compete on a tool by tool basis or are Synopsys and Cadence trying to see if they can build the Microsoft Office model of EDA? You might have noticed that MS Office prices aren't going up either.
 
The reasons industry-leading companies reject new business models are captured in The Innovator's Dilemma by Clayton Christensen. And, per the book (and my experience), the executives are doing the same things: 1) few intelligent questions, 2) confidence that they know all business models, 3) rationalization that the industry is "mature" (read: no significant growth is possible).

The irony these days is that the iPod, iPhone, iPad, AppStore, Google, Android, Amazon, Kindle, and Facebook (to name a few) are all shaking up their respective industries while the EDA industry is sure that there is no significant growth potential. So, I guess, we have found the first industry in the last 60 years to be immune to innovation? No.

What the EDA industry's new business model should be:
1) Time-to-market value. Get the customers their chips/systems quickly and efficiently. This can be done.
2) If designs can be done quickly and efficiently, then IP will become both less expensive and more profitable. Look at the applications for iPhone/iPad and Android. People can write them in a couple of weeks and have them on sale in a day. Since they are written in an object-oriented framework, they work across various similar platforms and are very easy to migrate between disparate technologies (iPhone -> iPad -> Android -> x86 -> ARM). The IP industry will quickly change to this model if #1 is addressed.
3) Software "blocks" == hardware blocks. The choice of SW vs HW is should only be a matter of speed, power, time-to-market and cost.
4) Growth will come from an end-to-end value: Platform hardware through system and software integration.

By the way, the large fabs should REALLY be driving this because 32nm with 450mm wafers will be quite a challenge if most companies only need a couple thousand chips. If this new EDA business model creates 3000 new design starts, each with a 2 month design time, then the shuttles (multi-die wafers) with 10 or more designs on each one will make a LOT of sense and $$.

So, let's see if the industry will actually start asking questions or if it will continue to work hard at sustaining the status quo.
 
From my perspective a problem is that there is a lack of open frameworks and good standards for the tools so mixing and matching is hard. Anyone who is innovative outside of the big EDA companies will find it very difficult to sell a point tool and will therefore be unlikely to get funding for the effort. A second problem I've seen inside big EDA companies is that rather than extend tools to cover more markets they'll buy up point tools that cover a market and sell them under an extra license - e.g. the difference between an "analog" P&R tool and a "digital" P&R tool is probably minimal at the code level, and they could be integrated at a fairly low cost, but generally that won't happen.

I'll also agree with meneboe's 3rd point 'Software "blocks" == hardware blocks' - but maybe not yet. I would also take Android as an example of how fast you can grow a market if you are prepared (like Google) to embrace open-source platforms. Maybe if some of the EDA customers started investing in open-source tools we could make some faster progress towards an IP based time-to-market driven design flow.
 
The business of ESL is difficult.

The SW development team cannot pay a large amount of cost to Virtual Platform.
HLS does not need a lot of licenses. Because run time is fast.

Will they continue selling an expensive ESL tool as it is?
 
Selling a hardware/ESL design methodology to software engineers is difficult, in my opinion it's easier to sell a software methodology to the hardware engineers. My own project (http://parallel.cc) attempts to do that by extending C++ in ways that make it easy to develop fine-grained parallel processing software - it's platform independent and open source, so the cost is low.
 
Hi Hiroshi,

I see FPGA companies adopting ESL type tools and expect them to go full ESL very soon (Xilinx/Altera). Since these tools will use a success based business model that is tied to silicon people will adopt them and FPGA design will have another advantage over ASIC. FPGA design starts are growing and ASIC design starts are shrinking. EDA had better wake up!

If there are FPGA designers reading I would be greatly interested in their opinion on this.

D.A.N.
 
Last edited:
Back
Top