Estimated reading time: 3 minutes

The Pulse: Ultra-high Stakes—Will UHDI Enter the Mainstream for Low- and Mid-volume?

The casual observer might think it is easier to make a low-volume product than a high-volume product. However, making anything in low volume means forgoing the extensive ramp-ups and process fine-tuning that are more easily amortized where high volumes are concerned. Why do I note this?

Well, UHDI is capital-intensive to introduce and a challenge to shake down a process with limited volumes. But there are signs that, at least in the arena of designs down to 25-micron line widths, that more fabricators are starting to test the waters. As industry knowledge builds, so should the adoption. UHDI is long established and shaken down in high-volume markets—phones/tablets, etc.—but the production world they inhabit is far removed from the low-volume specialist market, which now dominates U.S. and European fabrication.

Modelling and Measurement
Since last addressing this topic at Polar, we have had the chance to build, model, measure electrically, and SEM measure sectioned traces from a low-volume specialist provider. Initial results look positive, provided you de-embed the DC resistance from the traces, which, at a couple of mils or less, look more like resistors than pure conductors. I mused whether we would uncover an as-yet-undetected electrical effect that would make the measurement/modelling scenario more complex than before, but so far, so good. Papers should follow in due course.

What About AI?
I am fascinated with new technologies and how, over time, they yield benefits, often differently from what was predicted by the research. For example, in the 1960s, the prediction was that by the turn of the century, we would all be in flying cars. How wrong was that? Likewise, the invention of the laser yielded a variety of applications, some predictable, some long-lived, and others less so. Semiconductor lasers for recording data on CDs exploded into the mainstream, then faded as silicon memory progressed to where “recorded music no longer needed to rotate.” Other applications of lasers—eye treatment and medical treatment—outlived the use of lasers for storing music or other data. Likewise, lasers for data transmission in fiber optics seem here to stay.

How will AI play out? In image recognition, it shows promise in medical diagnostics with the use of machine learning. I question how far it will go in the programming environment with all the OEM requirements of traceability and IP protection. Will there be a day when a layout engineer asks ChatGPT to lay out a board for an avionics control system and simply trusts the output? It seems a big leap from now, where even manual intervention is often preferred over simple autorouting.

Likewise, how long until AI can choose materials for a stackup, especially when new materials continue to roll out, even when the AI engine can’t look back on historical data for learning? Interesting times indeed, and it will be fascinating to look back in 10 or 15 years to see which AI predictions came true and which fell into the weeds of “seemed like a good idea.” There is no easy way of predicting other than to observe from the sidelines.

A prediction that seems generally true for new technologies is that they often underperform and are over-hyped in the short term, but in the longer term (decades), can outperform initial predictions. I am interested in where AI gains traction and where the promises fall short. Whatever happens, it will be an interesting ride.

Inspirational Reading for 2026
On a completely different topic, but relatable to our changing industry, I read Paul Waldner’s autobiography, A Constant Immigrant. It’s a page-turner about a survivor who has lived and breathed the PCB industry. For Gen Z readers and others who think life is “stressful,” Paul’s book is a story of triumph over adversity. Older readers will discover a rich history of the PCB industry we know and love.

This column originally appeared in the January 2026 issue of I-Connect007 Magazine.