Iri Trashanski, Chief Strategy Officer at Ceva, is shaping the future of the Smart Edge with extensive experience across tech sectors.
Back in the day, a well-known mantra in the semiconductor industry was “Real Men Own Fabs,” a reaction to the emerging “fabless” model that allowed chip design companies to bypass owning costly manufacturing facilities and rely on foundries instead. Despite some in the old guard trying to downplay it, fabless became the standard model for the industry, and the outsourced model—rent vs. buy—has become the de facto approach for most.
Over time, other aspects of the semiconductor ecosystem became available as outsourced services, tools or IP that could be bought, licensed or rented. Owing largely to the complexity and cost of each individual step that requires specialized expertise, it often proved a better economic model to not do it yourself.
A fledgling British startup took the outsourcing approach further still when it introduced a model that offered what was then considered virtually a non-starter to outsourcing—critical design capabilities that powered the key processing functions in chips. The model allowed these critical blueprints, what we call intellectual property (IP), to be licensed by any company that saw value in not designing that particular aspect of their devices on their own.
Arm, as the company became known, broke new ground at the time with its RISC processor IP, and it eventually emerged as a standard alternative to designing your own central processing unit (CPU) core or buying a pre-packaged processor as a self-contained chip (Arm is an ecosystem partner of Ceva).
The semiconductor IP market quickly became a flourishing element in the chip ecosystem and gave rise to other successful providers. Today, the overall semiconductor IP market represents a $7 billion market and is growing at roughly the same clip as the overall chip industry. It is, in fact, the silent partner to many chip companies and device makers.
Why The Shift To Smart Edge Makes The IP Model Even More Attractive
The emergence of AI at the edge—where devices process data locally rather than rely on cloud computing—has made the efficiency of IP licensing even more valuable.
We call it the Smart Edge, and it brings many benefits to the high-tech gadgets and systems that populate our homes, workplaces, cars, public spaces and even our bodies, including faster response times, more diverse and personalized user experiences and better security and privacy, to name a few.
Just as important, the market opportunities for the forecasted $127 billion market for intelligent edge computing are diverse and fast-growing—driven in large part by the proliferation of smart devices across sectors where efficient and real-time decision making is crucial. Product developers, large and small, are rushing to provide devices that improve enjoyment, convenience and safety for customers.
Market research firm ABI Research forecasts that by the end of this year, the installed base of connected devices will exceed 55 billion units, spanning a wide array of end markets, including mobile devices, personal computers (PCs), wearables, home entertainment, smart home, automotive, commercial building, industrial and other consumer and Internet of Things (IoT) applications.
The speed, dynamics and fragmented nature of edge AI products place a premium on time-to-market, which in turn dictates the most efficient product development strategies. Time-to-market is not merely about efficiency—it is a determinant of market leadership. The ability to deliver new products faster and offer differentiated features can significantly influence market position, revenue and brand perception.
Companies developing chips and products in this space prefer to focus their resources on making their products unique in some way. Spending time on necessary commodity features or addressing gaps in areas where expertise is lacking can have a significant impact on market success.
Leveraging the proven semiconductor IP model optimizes internal engineering resources, allowing companies to focus on areas of true differentiation. For companies of all sizes trying to differentiate in Smart Edge markets, licensing functional blocks from reliable IP providers speeds time to market, ensures compatibility with standards and is far more economical than doing it themselves.
Core Requirements Of The Smart Edge
For Smart Edge devices, several core functions are required.
• Connecting is essential for the Smart Edge, allowing devices to link via wireless spectrums or standards such as Wi-Fi, Bluetooth, UWB, 5G or other proprietary protocols for updates and data. As wireless standards evolve, having access to the latest, most robust connectivity IP is crucial to keep up with performance and complexity.
• Sensing is the ability for devices to detect what is happening around them, collect relevant data and trigger AI to run when something has changed. It enables humans to interact with the devices via multiple modes—movement, gesture, voice, sound and touch.
• Inferencing runs trained AI models locally on the device using neural processing units (NPUs) for energy-efficient data processing. Using trained models on the device (versus connecting to a central server) reduces latency, lowers cost per query and enhances privacy.
All of these functions are available as trusted, licensable IPs from suppliers. Instead of spending months or even years developing and validating proprietary IP in-house, companies can seamlessly integrate silicon-proven IP into their designs, reducing development cycles. This agility is critical in the fast-evolving AI landscape.
Employing third-party IP offers significant economic advantages. Companies avoid the high costs associated with in-house R&D of IP blocks, such as embedded processors, neural network accelerators and wireless connectivity modules, and they benefit from economies of scale and the technical advancement provided by specialized IP vendors.
The Smart Edge is poised for significant growth, enabled by advancements in wireless connectivity, multi-model sensing and low-power inferencing. These are critical components for enabling real-time data processing and communication at the edge, where efficiency and speed are paramount. By leveraging silicon-proven IP, companies can capitalize more effectively on the market opportunities at hand.
Published on Forbes Technology Council
Iri Trashanski