Money Moves in Quiet Places

Money Moves in Quiet Places
By Jeff Brown, Editor, The Bleeding Edge

Very early on in my career, after earning my degree in aerospace engineering at Purdue, I was deeply frustrated with the lack of jobs in the industry.

It wasn’t a good time to be an aerospace engineer.

The air transportation, space, and defense industries were all in a funk in the early ‘90s.

But I was deeply determined after having worked hard for four years to get my degree. So I packed up my car and drove to Seattle in hopes of finding a job at Boeing.

One of the best things that came out of that move was the excitement of driving across the entire country east to west, coast to coast.

I traveled through Chicago, up through Minneapolis, then across North Dakota, northern Idaho, and then Washington State before ending up in Seattle and the beautiful Puget Sound.

One of the cities I drove through along the way on I-94 was Bismarck, North Dakota, situated on the Missouri River.

It’s a small town in the middle of the state amid the great plains of the U.S. It’s the kind of quiet place that we can drive past in a matter of minutes, surrounded by sweeping vistas of North Dakota – grassy plains and gently rolling, wooded hills.

Bismarck just isn’t the kind of place that we’d think to be associated with bleeding-edge technology and artificial intelligence (AI).

Certainly not a place for a young aerospace engineer to take a passing interest in, at the time.

Yet here we are, all these years later…

Who’s Got The Power?

On August 1, the North Dakota Public Service Commission held a conference – the NDPSC Data Center Technical Conference: Large Load Power Impacts.

It was a full-day event that resulted in about four hours and 42 minutes of discussion around topics like:

• How much demand there is for data centers and data mining
• The power requirements to support new data centers
• How much new electricity production will be needed to meet the new demands
• Any regulatory support needed to support the potential new construction and business opportunities for North Dakota
• The impacts of large data centers on surrounding areas

Bismarck – North Dakota in general, for that matter – isn’t a place that we’d immediately think of when considering developing artificial intelligence or building out massive data centers. But it should be.

Arguably the largest “asset” that North Dakota has for the high-tech industry is that it is the ninth-largest energy producer of all the states in the U.S., producing more than 4,000 trillion BTUs annually.

And because the state is sparsely populated, it’s one of the few states that produces a lot more energy than it consumes.

Which is what makes it such an interesting target for companies building AI factories – hyperscale data centers designed to create artificial general intelligence (AGI).

Access to cheap electricity – and lots of it – is the most critical input.

We’ve been exploring this topic a lot lately. This is a once-in-a-generation trend that will transform our lives in remarkable ways.

And oddly enough, the proceedings of the day-long conference revealed some plans that made even me do a double take.

The Double Take

Here’s what the chair of the Public Service Commission Randy Christmann considers to be a big electricity user, to date…

Our big customers were often 8 to 10 megawatts – that was a big user that probably got a special fee in the rates that the utility provided a special user.

But, as we’ve been tracking here in the very pages of The Bleeding Edge, the energy usage story is changing dramatically – almost by the day – with the increased training and operating requirements AI technology is presently demanding.

Here’s Commissioner Christmann again…

Now we’re plopping in 200-megawatt power [demands] in the countryside – that completely and dramatically changes what the grid is doing to serve that.

To translate the Commissioner: This is what has been required to provide electricity for the handful of data centers that have already been built in North Dakota.

And there are already plans for the building of at least six more.

But that wasn’t the most interesting insight from the discussions…

Buried within the almost five hours of conversation was this gem from the state commerce commissioner Josh Teigen…

Data companies have approached us and said we want to start with 500 to 1,000 megawatts and scale to 5,000 to 10,000 megawatts.

Teigen went further to say that project estimates for at least two of these massive facilities have been around $125 billion.

That’s what made me do a double take. Ten thousand megawatts! $125 billion… each!

Data centers requiring this much electricity will need massive, dedicated power plants to keep them running.

To put things into perspective:

• Today’s largest data centers powering artificial intelligence cost billions, have at least 100,000 powerful GPUs, and at least 100 megawatts.
• The next generation of AI factories will cost tens of billions and have more than 1 million powerful GPUs and more than 1 gigawatt of power – these are being constructed and will come online by late 2025 and 2026 – these are the AI factories that will create artificial general intelligence (AGI).
• And by 2028, we’ll be looking at data centers with more than 10 million powerful GPUs for training, costing more than $100 billion and requiring more than 10 gigawatts of power, the equivalent amount of power for an entire medium-sized state.

It’s this last category that was under discussion last month… and in Bismarck, North Dakota of all places. I wonder if anyone else was paying attention.

High-Stakes Conversations in Quiet Places

Fortunately, the Bismarck forum was public, which allowed us to gain insights into what some of the most powerful tech companies in the world are planning.

While the names of the companies that were evaluating North Dakota for their data center plans were not named, they were identified as having trillion-dollar market capitalizations.

That’s perfectly logical, considering the $125 billion price tags for each AI mega factory. Very few companies can even think of making that level of investment.

We already know that Microsoft and OpenAI have already planned their $100 billion AI factory for AGI. The other obvious suspects are Amazon, Alphabet, Meta, and of course Apple – all are hungry to keep up with the Microsoft/OpenAI duo.

And as we explored in yesterday’s Bleeding Edge – “One AGI to Rule Them All,” Elon Musk and his team at xAI have shocked the entire industry by building and commissioning Colossus – now the world’s most powerful AI training cluster – in a mere 122 days.

Many in the industry are so shocked that they are in disbelief. They are raising questions as to whether or not Musk and his team at xAI actually completed Colossus. One claim that was xAI didn’t have enough power to possibly run the 100,000 GPUs contained within Colossus.

While the utility company can’t yet supply Colossus with all of its power needs, xAI already has about 20 gas turbines of its own to make up for what the local utility can’t provide.

Is it so hard to believe that Musk just found a way to get it done? It’s not hard to just buy your own turbines and natural gas supply to offset what the utility can’t provide in the short term. Musk has the resources to do it.

The reality is that large tech companies are very nervous about how fast xAI is moving. In a few months, Colossus will double in size, growing to 200,000 powerful GPUs.

And that’s precisely why executives from the thus-unnamed “trillion-dollar companies” are having powerful discussions in places like Bismarck, North Dakota. Colossus only fuels this feverish race to invest, build, and create an AGI.

And whether these mega AI factories are built in Texas, New Mexico, Pennsylvania, Wyoming, North Dakota, Louisiana, or Oklahoma – all places that have large amounts of energy production – is beside the point.

The point is that it is happening. Not 10 years from now…

It’s happening right now and in the very near future.

Regards,

Jeff

P.S. Opportunities to profit from this trend are cropping up even faster than these data centers. I’ve written a new playbook explaining how to profit from – and hopefully retire on – this massive, multi-year megatrend in AI. Click here for details.