Is “Free” going to work in hardware the way it has worked in software? We have been so accustomed to use free software but will hardware be ever free. The distribution costs are almost nil for software products while hardware involves both the distribution and manufacturing costs. But here we are just looking at the final product in the value chain. What about the intermediate stages?
To answer this question, we will have to consider the various levels in the hardware value chain and analyze the scope of “Free” model. We have already started to see some open source movement gaining up in the hardware side. There are some free EDA tools like Spice and GTKwave. There are some open source IPs for Microcontrollers and many other designs on opencores. There are also some open source board designs, schematics etc. Considering the “software” nature of these components in the value chain, we can explore the viability of “free” here.
There is decent amount of stuff free or in open source domain with various licenses. However, there is nothing of equivalent caliber as Apache or Linux from the software world! In software, there is MySQL against Oracle, but there is no free silicon IP equivalent to ARM. To my knowledge, there is no open source silicon IP being deployed in mass production (Are there any? Would be interested to know if any).
Reasons? One of them is that it is very hard to ascertain the quality of a “Free” IP. Another important thing is the much needed support in the integration of the IP. Then there is customizations based on various project requirements like frequency, interfaces, technology etc. With all these problems and considering the fact that the cost of IP is marginal compared to the overall chip cost, it is a risky proposition for a chip vendor to consider a free IP.
So it there anyway of establishing a successful IP model based on open source or “Free”? One of the options is to go the “software way”. The delivery is open source or free but you charge for the support and services. Some IP companies, at a smaller scale, may be already working on this model.
The other “software way” of making things available free by means of advertisement is difficult to replicate in IP model. However, it is worth considering the Freemium model, where a basic version is provided free with additional costs for enhanced features. One thing to note here is that the “Free” need not always be open source. It can be proprietary IPs with no rights for free distribution.
The Freemium can have various flavors. For instance, the use of IP can be free for SoC integration and final implementation, but can involve licensing fees when the production starts. Customers will be happy with this model as they don’t have to pay unless the IP has been proven. On the other hand an IP vendor gets the credibility and a foot in the door of customer more easily. One practical problem can be in terms of cash flow of the IP vendor as it is effectively procrastinating the scheduled cash receivables. However, once the IP cost has been amortized with first few customers, this model can be tried with lesser risk. One de-risking strategy for an IP vendor can be that IP development be started and done along with the free customer so that the IP requirements are based on the true customer feedback.
Freemium model can also be based on feature limitations. Lets say, an IP vendor provides two variations of an IP. One version is a non-optimized version with limited set of features and without support. Another version is full fledged production version IP. Customer can use the free version for bring up and limited production, however, it needs to buy the later version for mass production. The model fails if somehow customer can workaround with the free version for its mass production.
I am sure there are many more options which can be viable for “free” IP business model. Only time will tell if some IP company will get any success with this model.