Trust the process: why demand for on-device processing is growing

Given its potential to change the way we interact with the world around us, it’s no surprise to see that the market value projections for the artificial intelligence of things (AIoT) are astronomical. The market for devices that inhabit the convergence point between AI and the IoT is expected to reach a value of $102bn by 2026 according to IoT Analytics, and/or $144bn by 2028 according to Brand Essence Research.

To guarantee as large a slice of that pie as possible, engineers and brands need as comprehensive a plan to differentiate and elevate their product. That’s reflected in this year’s Edge of Now report – the annual report which has examined the attitudes of engineers towards the AIoT over the past three years.

This year’s report has found that 82% of engineers see the integration of AI into their products as a competitive advantage. With the AIoT closer than ever, that integration is the #1 priority for almost two thirds (63%) of those surveyed in 2022, as they prepare to launch products into a market that will only become more competitive in the coming years.

On the increase

However, engineers are unable to develop and deploy new or increased AI capabilities without the processing power to support it. This is, traditionally, where the balance between cost and power has made meaningful AI integration a logistical impossibility.

That’s not to say that processing power isn’t baked into most devices as standard; 73% of engineers’ on-device processing requirements are already ‘high’ or ‘very high’. But the cost of processors capable of supporting AI functionality has, historically, been prohibitive.

That is changing, with engineers increasingly capable of bridging that cost/power gap, they require more processing power to accommodate their more ambitious, more intelligent designs. Indeed, demand for on-device processing is set to continue for 75% of engineers.

With consumer technologies (43%) and smart home devices (25%) the most common sectors for on-device processing demands – both of which are well-established verticals in their own right – increasing the processing power of a device lays the foundation for AI to flourish in new and innovative products.

The core of the issue

The development of such products requires silicon that can empower the designer or engineer to demonstrate their ingenuity. If it’s now possible to secure power-appropriate processing capabilities at feasible cost, what do I need for my design?

At XMOS, we firmly believe that the answer is our fast and economical platform, xcore®.ai – a means of embedding intelligence within a device itself. The versatility of the chip is such that designers can tweak its structure to rebalance its AI, DSP, I/O, and/or control capabilities, driving down potential costs.

Moreover, the affordability of opens up the potential for mass-produced deployment. If a city council is investing in smart lighting, for example, or if a factory owner decides to integrate voice control across an entire complex, is a financially viable processing option.

Hardware that ticks all of these boxes – cost-effective, mass-producible, and versatile enough to be moulded into the vision of the designer – is fundamental to the future of the AIoT.

Scroll to Top