Thus sayeth a PhD materials scientist and entrpreneur via EE Times, July 11:
For decades, materials were background players in semiconductor design. Now they are at the center stage.
Earlier this summer, Meta penciled a 20-year deal to extend the life of an Illinois nuclear plant, which will provide 1.1 GW to power the company’s operations in the region.
The deal is a lifeline for nuclear operators. Still, more importantly, it’s a watershed for energy-hungry tech firms as a deployed example of how big tech can underwrite the viability of nuclear energy.
Elsewhere, xAI has been widely criticized for ignoring environmental regulations while using unlicensed methane gas generators to power its Colossus supercomputer in Memphis, Tennessee.
The demand-side numbers don’t lie: Goldman Sachs Research forecasts that global power demand from data centers will increase by as much as 165% by the end of the decade, with AI expected to account for 19% of that by 2028.
Meta’s deal looks like a bold solution to AI’s energy appetite, but it’s just kicking the can further down the road. The real problem isn’t supply. It’s demand.
As AI workloads continue to grow exponentially, the only sustainable path forward is radical energy efficiency —a challenge the semiconductor industry is now being forced to confront head-on.
The answer isn’t just coming from traditional chip architectures, but from a fundamental reimagining of the materials that connect and insulate our computing systems.
For decades, the industry was guided, you might say insulated, by the predictability of Moore’s Law. Every two years, we could count on faster, cheaper, denser chips. But that promise has broken down, coinciding with a second tectonic shift: the AI explosion that’s driving compute and bandwidth demand at a pace that far outstrips what traditional architectures were built to handle.
Triple threat: energy, heat, and physics
We’re pouring more energy into systems to achieve increasingly smaller performance gains. That’s not a growth curve; it’s a red flag.
Every watt of compute power becomes a watt of thermal load, forcing the deployment of ever more elaborate cooling systems, such as liquid immersion, cold plates, and phase-change materials. In compact environments where fans are impractical, thermal design is already the limiting factor....
....MUCH MORE
Here is the author, Stefan Pastine, at everybodyWiki.
And at MIT's Technology Review: "This US startup makes a crucial chip material and is taking on a Japanese giant"
