
It’s early to tell whether the recent generation of AI advances will go beyond recommendation engines and fraud detection—the two shining stars of the big data era. The distinction is simple: If I fudge a recommender, I lose a possible sale; if I fail a fraud detection, I lose money I already have.
One might categorize the latter as mission critical—that high-stakes idiom from the Space Race days.
Whether “nice-to-have” or “need-to-have,” the AI boom has placed immense pressure on the data centers running these LLMs and their supporting agentic systems. After all, current investment plans augur for superscale, 24/7 tokenization across the globe.
This raises an increasingly urgent question: How will these data centers ensure continual operation when extreme weather—which we seem to see with increasing frequency—threatens the power supply, cooling systems, and roofing required to keep those chips churning?
Learning experience
I got a chance to dig into this trend recently for Data Center Knowledge. The result was “How Data Centers Are Adapting to Extreme Weather.” Like every story, it was a learning experience.
In this case, I learned a thing or two about the elements. I had previously followed the evolution of predictive analytics in data science; since starting this blog, I’d even had occasion to look into meteorology hero Ted Fujita—the man most responsible for aviation’s adoption of Doppler Radar. But I didn’t know much beyond a few headlines on the topic of climate change. Yet, it seemed highly relevant to how a data center survives a storm.
The backstory is that weather has been a bogeyman for data centers throughout this century. Cooling systems were knocked out and generator fuel ran short during Hurricane Katrina in 2005. Natural gas lines froze during Texas’s historic Winter Storm Uri in 2021. In 2022, exceptional heat in Europe led to cooling failures that forced Google and Oracle to shut down entire cloud regions. That same year, X (formerly Twitter) experienced a similar outage as a heatwave tore through central California.
Vortex Vector Vexes
When I started work on this story, an extraordinarily long cold spell was beginning in Boston. More importantly, this was part of a widespread, long-duration deep freeze across the continental United States.
Scientists and meteorologists suggest a polar vortex has come into play. It normally confines the low-pressure cold air surrounding Earth’s poles, but it can wander. Dramatically increased shifts in climate variability have been laid at the door of the polar vortex and related “Arctic Amplification.”
This Arctic Amplification—in some cases related to Sudden Stratospheric Warming (SSW)—is among the climate phenomena cited in recent years by NOAA, NASA, and academic researchers. Meteorologists have also connected SSW with the January-February winter storm “Fern,” the name given to the deep freeze currently descending upon us.
Also learned during the course of this story: a new word. Jellify. To jellify is to become gelatinous. It’s what happens to fluid-carrying equipment when the weather gets truly frigid. As extreme weather patterns shift, areas carefully designed for extreme heat may suddenly find their infrastructure turning to literal jelly in the cold.
In a pinch, the machinery accompanying today’s great LLMs can go on the fritz just as their buzzing tokens start to detect a rising anomaly. The can be taken down by jellifcation of cooling equipment, other peripherals, or missed deliveries of diesels. Weather happens. J.I. Vaughan
Related
“How Data Centers Are Adapting to Extreme Weather” – DCK
“Fujita – Watcher of the Air Waves”- PG
