• • • AI • • •


WHY PCHES ARE CRITICAL TO AI’S NEXT PHASE


BY BEN KITSON,


HEAD OF BUSINESS DEVELOPMENT, PRECISION MICRO


I


f it feels like the semiconductor market is suddenly back in the headlines, that’s because it is.


ASML, the Supplier of photolithography systems,


recently reported that its shares have risen around 97 per cent over the past six months, reflecting renewed investment in chip production. Yet behind the headlines, there’s a quieter and arguably just as critical story in managing the heat generated by both chip production and the AI hardware that depends on it. Today’s cycle is unusual. Hyperscalers are


pouring investment into AI data centres, driving unprecedented demand for high-performance hardware. What’s more, much of this computer equipment has already been committed. Such a combination is creating a perfect storm


for infrastructure planning, with AI operators facing high power densities and unprecedented cooling requirements in their data centres. Traditional data centers were designed around


racks drawing roughly 5–10 kW, however AI clusters now operate at 30–50 kW per rack. What’s more, advanced GPU and accelerator platforms are already reaching 100–120 kW per rack, meaning air cooling alone is no longer sufficient.


32 ELECTRICAL ENGINEERING • JUNE 2026


Thermal management in focus Thermal constraints are finally hitting the headlines. In May 2025, chip giant Nvidia said hyperscale operators were installing tens of thousands of its latest GPUs every week, a deployment rate expected to accelerate with the rollout of its ‘Blackwell Ultra’ platform. The company’s public roadmap indicates that its


next ‘Rubin Ultra’ architecture could place more than 500 GPUs in a single rack drawing up to 600 kW, highlighting the scale of the cooling challenge now facing AI infrastructure. Across AI infrastructure, thermal stability has


become a defining constraint not only in chip design, but also in the infrastructure required to power and cool high-density compute environments. High-performance liquid cooling and


micro-channel heat exchangers have shifted from niche solutions to essential components. The same engineering principles, precise control of fluid flow, maximising heat transfer and producing compact components with tight tolerances, now apply across multiple applications. Engineering expertise developed in high-


precision semiconductor environments is now being applied to printed circuit heat exchanger (PCHE) technology for AI data centres, a convergence between electronics manufacturing and energy infrastructure.


Why PCHEs matter PCHEs aren’t just a fancier version of conventional designs like shell-and-tube or plate-and-frame.


They’re smaller, lighter and more efficient, making them ideal where space is tight and density is high. In data centres, this translates to more racks per


square metre without compromising reliability, while reducing the energy needed to keep compute hardware cool. Energy efficiency is another factor, with AI


workloads expected to drive a significant jump in global electricity demand. Goldman Sachs projects up to a 165 per cent increase by 2030, meaning that every watt of cooling matters. Compact, high-performing PCHEs not only save


floor space, but they also help manage energy costs and improve overall power usage effectiveness, making them a critical component in high-density, hyperscale AI infrastructure.


Scaling chemical etching The very features that make PCHEs effective, the micro-channels, high surface area designs and tight tolerances, also make them difficult to produce. Conventional machining can create prototypes but is slow, introduces burrs and is commercially unviable at scale. Chemical etching, on the other hand, overcomes


these challenges by forming all channels simultaneously across the plate. This produces stress-free, precise features, with diffusion bonding then forming the final heat exchanger plate. Chemical etcher, Precision Micro, has produced


PCHE plates since the technology’s early commercialisation in the 1990s, operating a dedicated 44,000 ft² facility capable of processing


electricalengineeringmagazine.co.uk


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44