AI operators could literally follow Bitcoin miners to power sources

The big picture: Bitcoin miners have already had to get creative about finding sources of energy and using it efficiently.

• Axios spoke to three different miners, Riot, Terawulf and Iren (the latter two are also dabbling in AI to diversify their income streams), who illuminated some lessons Bitcoiners have learned that could serve AI.

Geographical flexibility

The big companies of tech are accustomed to acting like 800-lb gorillas (they sit down wherever they want to). Terawulf's Nazar Khan, chief operating officer and a co-founder, told Axios that that is increasingly not going to be an option.

• Favorite places for data centers, such as Northern Virginia, he said, are hitting capacity.
• On the other hand, parts of the country not previously seen as desirable locations look a lot better today. He pointed to Columbus, Ohio, as just such a place.

"The map of where you can go is expanding," Khan explained. Terawulf's founders have found that a 100 GB fiber connection is feasible in a lot more places than it used to be.

• Bitcoin miners scramble to reach any spot with consistent, inexpensive power. AI operators will need more room than miners do, but they will still need to spread out more than they've been accustomed to.

Advanced facilities

Chasing abundant (therefore, cheap) power often means operating in out-of-the-way places, Iren Energy's (formerly: Iris) Lincoln Tan tells Axios.

• "Beyond the hardware, the most important thing is actually having really good facilities that can withstand the heat, the cold and the dust of operating in remote environments," he said.
• Unplanned downtime comes from broken machines, usually due to environmental conditions. Data centers that mange the environment lower downtime.

They can take that a step further and consider fundamentally changing the architecture of computing, as miners already have, by cooling computers with liquid rather than air.

• Riot Platforms' Pierre Rochard, its head of research, explained to Axios that traditional data centers are far less energy intensive than bitcoin miners, commanding less than 20% of the power demand per square foot of facility.
• AI facilities, however, will probably be somewhere in the middle, density-wise, of those traditional data centers and bitcoin mining. That could be dense enough to try liquid cooling, as more and more miners do.

Zoom in: "There is great cooling technology out there, and they should not limit their ambitions based on the energy consumption of the system," Rochard said. "Bitcoin mining has really shifted the mindset on the feasibility of developing extremely energy-intensive data centers."

• "It's all about doing it responsibly," which Rochard described as collaborating with power providers to make sure the grid stays stable.

And then there's demand response. Bitcoin miners have found that they can get access to new streams of electricity by agreeing to turn operations off when demand peaks, as Axios has previously reported.

• AI operators are unlikely to be able to flip off that fast, though one segment of the operation right now — the one that crunches data for "learning" to improve AI models — can probably work more like Bitcoin miners.
• Operations supporting "inference," or delivering answers to users, probably need to be closer to end users and will be much more sensitive about downtime.

"As we transfer to inference, that will be interesting, in terms of whether latency is important," Tan said. Though he's skeptical that it will be extremely important for most users.

The intrigue: Whether AI providers join demand response programs or not, insiders expect AI to make demand response itself work better.

What's next: AI is developing much more quickly than the facilities to serve it can possibly get built. New data facilities take years.