Why AI infrastructure has to move ahead of the ask
Henkel strategist Katie Willgoos explains why data-centre thermal management demands early decisions, ecosystem coordination, and calculated risk.
As AI chips become denser and hotter, thermal management and efficient heat dissipation are emerging as major constraints for data centres. Katie Willgoos, Market Strategy Manager for Data & Telecom at Henkel Adhesive Technologies, argues that suppliers cannot wait for a customer order before acting. Technologies need to be ready before the market picture is complete.
Acting before certainty
Willgoos joined Henkel through its 2022 acquisition of a Boston-based thermal-management company, bringing experience developed over a decade at a start-up. She says the advantage of a small organisation is not merely budget or headcount, but the ability to move while some details remain unresolved. Her preference is to be “a little wrong and early” rather than perfectly right and late.
Hyperscalers and equipment manufacturers have compressed their design cycles, so they need partners already developing thermal-interface materials and heat-dissipation solutions. In this account, moving early is not an unsupported gamble; it is a calculated decision based on technical expertise, market knowledge, and close customer dialogue.
An ecosystem rather than a linear chain
Willgoos describes data and telecom as an interconnected ecosystem rather than a conventional value chain. Co-packaged optics, or CPO, is one example: progress requires several parts of the ecosystem to invest and align. Companies may compete for customers while still depending on each other to build and scale complex systems.
A culture of learning and responsibility
The discussion also extends beyond materials. Willgoos says the effect of AI infrastructure growth on jobs—what roles will expand and which may disappear—remains difficult to answer. She frames work on the technology as both a privilege and a responsibility. Her proposed operating culture is to keep asking why established processes exist, avoid organisational silos, and abandon unsuitable technologies without becoming trapped by sunk costs: learn, pivot, and move on.
Source: AI News