AI Is Moving from Hype to Hardware — and from Pilots to Scale

Sep 04, 2026
Key Summary
  • AI investment is accelerating, with major spending flowing into servers, semiconductors, networking and infrastructure.
  • Inference cost and efficiency are creating new demands across memory, advanced packaging, power and cooling.
  • PCB manufacturers are adopting AI, but few have scaled it across factory operations.
  • AI’s long-term impact will depend on electronics infrastructure, data quality, operating discipline and human judgment. 

by Sydney Xiao, president, East Asia, Global Electronics Association

I’ve been reading a few pieces on AI recently, and three of them stayed with me for different reasons.

The first is about where the money is going.

Gartner — Worldwide AI Spending Forecast

Gartner expects worldwide AI spending to grow 47% in 2026, reaching about $2.6 trillion. A significant share of that is going into the infrastructure behind AI — servers, semiconductors, networking and the systems needed to run AI at scale.

What I find interesting is the speed at which investment is moving into real capacity. AI is becoming a major infrastructure build-out as well as a technology cycle.

The second is about where AI itself is going.

McKinsey — Frontiers of compute: The technologies to reduce AI inference costs

McKinsey’s “Frontiers of compute: The technologies to reduce AI inference costs” makes a point that stayed with me: the next breakthrough in AI may come from a cheaper token.

As AI becomes more widely used, inference cost and efficiency matter more. That creates new demands — and opportunities — across memory, networking, advanced packaging, power and cooling.

From an electronics perspective, this is where AI becomes very tangible. The economics of AI increasingly depend on the hardware and systems underneath it.

The third read is closer to home: Global Electronics Association’s “AI in PCB Manufacturing: From Pilots to Scale.”

One number stood out to me: 68% of PCB manufacturers are already using AI, while only 8% have reached real scale.

That gap feels very familiar. Starting a pilot can happen quickly. Scaling it across a factory requires good data, clear processes, the right skills and strong operating discipline.

For companies, the key question is whether AI is genuinely improving quality, productivity and decision-making.

For individuals, AI can help us move faster. Experience, curiosity and judgment still matter, and the real value comes from knowing when and how to use the technology well.

For me, these three reads point in the same direction: AI is attracting serious investment, creating new demands across the electronics value chain, and moving from experimentation into real operating capability.