It has been tried before: An EDA company, which usually licenses software as a design tool, believes it can create a circuit design in its services group for one company and then repackage and sell it as IP.
In general, that approach has failed. But what if we take the EDA company out of the equation. Can a design services company be a successful IP company? eSilicon believes it is possible and is being quite successful at it. I recently had the opportunity to speak to Patrick Soheili, vice president of marketing and general manager of eSilicon's IP business, and Gino Skulick, VP and GM of its SDMS business.
I asked about the business that eSilicon is in. Skulick said eSilicon basically has three product lines. The first is its traditional ASIC business, which the company started 13 years ago. It taps into third-party foundries and IP cores and brings it all together as a service for a client.
There is a second product line called SMS, which leverages eSilicon's operational service capability. It brings the supply chain together and is offered to other semiconductor companies to help with their internal infrastructures. It is both a tool and a service. Finally, the company has its IP business.
In a nutshell, eSilicon can do everything involved in the design, implementation, and test of a chip and the IP that goes into it, or it knows somebody who can, and its client can decide just how much it wants to do itself. For example, a product company may not have a packaging or test guy available in house, so eSilicon can supply the necessary experts.
The only things that eSilicon definitely doesn't do are:
- Come up with the specification itself. The client basically takes on all of the product risk.
- Manufacture the parts, although the firm has close relationships with the foundries and can track the whole process for the client.
In some way, I liken this to cloud computing. eSilicon acts as an expandable resource that supplies engineers with various skill sets, the EDA tools they need to get the job done, IP, and other design services as and when needed.
This saves the design company from having to grow its employee count and keeps costs down. At the same time it can lower EDA tool costs. Not only are the costs of the tools aggregated over several designs, but they come with experts in their usage. As a company, eSilicon will also see lots of designs per year and can impart that collective knowledge on each design problem, whereas a small design house may only do one or two 40- or 28-nanometer designs in total.
eSilicon has three levels of engineers within its fold. It has a core group of design experts, it has domain experts, and then it has a slew of designers who are led by the other two groups of people. This gives it the ability to expand and contract quickly based on client needs.
About two or three years ago, eSilicon bought an IP company that specialized in memory and memory compilation. All chips need memory, so that IP is used internally, and eSilicon is its own biggest customer, but it also markets the IP externally. This is a different way around from what was tried in the EDA companies.
Rather than trying to reuse customer designs, eSilicon wants to upsell additional services to its IP customers after they see the value that they may be able to provide. At the same time, it will not develop a new IP block until a design requires it, so it always does have a customer that is helping defray the cost of development.
Today eSilicon has a 45/45/10 percentage split of its staff among ASIC services, supply-chain management, and IP, but does see the IP portion expanding over time as it diversifies into areas other than memory IP.
I will leave it to the market to decide if eSilicon is successful, but I wonder whether you think a company doing IP, design, and tools is a good idea?