- Electronic System Design Alliance (ESDA) 2018 CEO Outlook (San Jose)
- Global Semiconductor Alliance (GSA) Silicon Summit (San Jose)
- Design and Reuse (D&R) IP-SOC Days (Santa Clara)
- TMSC Symposium (Santa Clara and Shanghai)
- GSA Global Leadership Summit (Shanghai)
In each of these events, I asked our industry leaders what they think of the Silicon-as-a-Service business model. In particular, I wanted to know their outlook for when they plan to implement this business model. Through the discussion and conversation, however, I soon realized that the concept of Silicon-as-a-Service was not well known. There were different interpretations of what it meant and hence different opinions to my question.
This is not surprising as most semiconductor companies are still selling silicon as a product rather than deploying it as a higher revenue and higher margin service platform. However, the concept of “xx-as-a-service” is not new. In fact, using hardware as a service platform has existed for years. According to Wikipedia1:
Platform as a Service (PaaS) or application platform as a Service (aPaaS) or platform base service is a category of cloud computing services that provides a platform allowing customers to develop, run, and manage applications without the complexity of building and maintaining the infrastructure typically associated with developing and launching an app.
The “xx as a service” business model offers great value to users because
- It reduces entry barrier on cost,
- It offers great flexibility on access, and
- It improves quality of service: latest technology availability and 24×7 support.
Moreover, it enables “pay-per-use” revenues or recurring revenue generation for businesses. A clear win-win for all stake holders.
In fact, this win-win business model is perfectly applicable to silicon. Despite lacking of a proper definition in Wikipedia on silicon-as-a-service, industry pioneers have already adopted this innovative business model. The first key player is not a typical member of our semiconductor industry, it is Google2! As stated in a recent article on CNBC3:
Google has finally started renting its new A.I. chips to outside developers
- Google will now allow other companies to rent its artificial intelligence-tailored chips, which it introduced in 2016.
- Google’s tensor processing units are available in “limited quantities” to Google cloud customers.
- The TPUs allow Google to rely less on chipmakers like Nvidia and add another revenue stream to its cloud business.
According to Google’s pricing calculator4, 1 hour of a single TPU instance will generate $6.5 for Google, which means potential revenues of $57k per TPU per year. This is far above the revenue generated by the sales of one single TPU device, a win for Google. At the same time, it allows companies who otherwise cannot afford to build their own AI chips to access latest AI technology through Google, a win for its customers.
Other examples of Silicon-as-a-Service include AMD with their RadeonTM Pro Graphics5 and NVIDIA with their NVIDIA GPU Cloud (NGC)6. Both have started to offer services on their silicon in the Cloud for AI, and Deep Learning applications. AI has become a key driving force of the “Silicon-as-a-Service” model in the Cloud.
In addition to AI applications, the growth and the economics of the IoT also naturally leads to this service oriented business model, as evident by the following two articles:
- In the whitepaper from Comtech Telecommunication, “Silicon as a Service: New Revenue and Differentiation Opportunities for IoT Chip Vendors7”, it explores how current trends such as the proliferation of IoT, rise of cloud-based services and increasingly ubiquitous wireless and wired network connectivity enable IoT chip vendors to re-architect their business model. Ultimately, adding revenue-generating capability on new and existing deployed chipsets.
- An EE Times article by Rick Merritt titled “Qualcomm preps Silicon as a Service8” has the opening statement of “Qualcomm is leaping from OEM silicon to end-user services with a kind of offering that one analyst sees as the start of a broad trend at the intersection of blockchain technology and the Internet of Things. Qualcomm Wireless Edge Services combines security and management features that are already attracting interest from the likes of Alibaba and Baidu.” Qualcomm’s Wireless Edge Service was announced in the last Mobile World Congress.
Last but not the least, Accelize, a French start-up company, has developed a market place where acceleration functions such as GZIP compression or video encoder for FPGAs in the cloud can be accessed in a pay-per-use model. At the heart of this new business model is the silicon licensing scheme that protects the intellectual property as well as metering usage so as to allow pay-per-use subscription. This is where Algodone comes in. Algodone’s Silicon Activation Licensing TechnologyTM (SALT) is the enabling technology to realize the Silicon-as-a-Service business model9.
Above are all the examples of Silicon-as-a-Service at play.
The Algodone’s vision of Silicon-as-a-Service is broad. We believe Silicon-as-a-Service is a silicon licensing and delivery model where silicon innovations generate revenue beyond initial point of sales, either through a subscription licensing model or feature reconfiguration or upgrades later in field. Silicon-as-a-Service can be applied to all vertical market segments, beyond the ones mentioned above. How this is done will deserve a blog on its own.
We welcome industry thinkers to participate in the discussion and definition of Silicon-as-a-Service. To us, there is no mystery. Silicon-as-a-Service is about value generation throughout the device life cycle. The definition is simple, yet the doing requires courage as it will disrupt the current way of business in the semiconductor industry. But as the benefits shown by other industries and the early adopters of our own semiconductor industry, it will be a win-win for all stakeholders. We are proud to be an enabler for this disruption and transition.