The Real Bottleneck
Artificial intelligence runs on electricity, and electricity is the part of the AI story most people are not watching. Ian King’s Musk pitch opens on this exact point: AI is power-constrained, the grid cannot catch up, and the companies that can deliver electricity in weeks rather than years are the new chokepoint.
The example is the clearest one available. Musk built xAI’s Colossus data center in Memphis in 122 days, then doubled its size in 92 days. The computing was the easy part. Powering it was the hard part.
The Scale of the Problem
Colossus needs roughly 2 gigawatts of power. For context, that is more than Hoover Dam produces in a year. The local utility, Memphis Light Gas and Water, could spare eight megawatts of the 150 Musk wanted for phase one.
Read those two numbers again. A data center wanted 150 megawatts and the utility could offer 8. The gap between what AI needs and what the grid can deliver is not a rounding error. It is the defining constraint of the buildout. Our look at energy infrastructure companies explains who actually closes that gap.
Why the Grid Cannot Catch Up
The grid is slow by design. A new transmission line can take a decade of permitting, land disputes, and construction. A new power plant is not much faster. Data center operators do not have a decade. They want power online in months because the chips are already ordered and the customers are already paying.
That mismatch is why Musk built his own power plant in Memphis. When the utility could not deliver what Colossus needed, the fastest path was to stop waiting for the grid and generate the power on-site. King’s quote, “For the first time in his life, Elon found something his billions couldn’t buy,” is marketing, but it points at a real truth: money can buy chips, but it cannot buy a faster grid.
The Chokepoint Argument
This is the core of King’s investment thesis. If AI is power-constrained, then the scarce asset is not compute, it is delivered electricity. Companies that can stand up generation in weeks, rather than wait for a decade-long grid upgrade, become the bottleneck that everyone else has to pay.
It is a reasonable frame, and it has been playing out across the industry. Hyperscalers have signed long-term gigawatt-scale power deals with on-site power providers precisely because the grid option does not exist on their timeline. Our piece on the Elon Musk power plant details the specific deal at Colossus.
The Caveat
The chokepoint argument is directionally correct, which is why so many newsletters now run some version of it. The caution is that being directionally correct about an industry is not the same as picking the right company at the right price. Power-constrained AI is real. But every company in the sector trades as if it will win, and not all of them will.
The honest way to use this frame is as a filter, not a forecast. Ask which companies can actually deliver power on a data center’s timeline, which have the balance sheet to fund the buildout, and which are already priced as if they have won. The demand story is strong. The individual pick is where the risk lives. For the classic version of this thesis, see our AI energy bottleneck explainer.
How Long the Shortage Lasts
The natural question is whether this is a temporary squeeze or a structural shift. The answer is a bit of both, and the distinction matters for how you invest.
The temporary part is the current scramble. Hyperscalers are building data centers faster than anyone can connect them to the grid, so the next two to three years are a bidding war for on-site power. That is the window King’s thesis is aimed at, and it is real.
The structural part is longer. Even if utilities accelerate, transmission lines take years to permit and build, and the demand from AI is growing faster than the supply of grid capacity. Some forecasters think the gap narrows only toward the end of the decade, when a wave of gas, nuclear, and geothermal projects finally comes online.
That timing matters for the companies in this space. The ones signing 10-to-15-year contracts today are locking in revenue that outlasts the shortage. The ones chasing short-term spot exposure are betting the scramble never ends. Knowing which is which is the difference between a durable infrastructure business and a cyclical one.
For an investor, the useful frame is this: the shortage is real and multi-year, but it is not permanent. Buy the companies with contracts long enough to survive the day the grid finally catches up, not the ones whose entire thesis depends on the shortage never ending.
Ready to see the research? Click here to access Ian King’s report.
NewsletterVetter is an independent publication. We receive compensation from some of the services we review through affiliate links. Nothing on this site is investment advice. Always do your own research.