The call for “Sputnik moments” far outweighs their actual arrival. But the U.S. might find itself in a Sputnik moment in its race with Communist China over artificial intelligence.

A ‘Sputnik moment’ calls for unified national action toward a goal. And with a new President with a strong penchant for industrial policy, it is a good bet that the A.I. race with China will galvanize national action. The challenge will be to the institutions the U.S. has to rise to this Sputnik moment. And American institutions will be challenged because A.I. takes power. Gobs and gobs of electricity will need to be built quickly without putting existing customers on the hook.

On one side, we institutionally have what amorphously but accurately can be called “Big Tech.” Big Tech – the Googles, Microsoft, OpenAI, Metas and Amazons of the world – lives by the creed “move fast and break things.”  On the other side, we have the U.S. utility industry. Utilities are something of the opposite of tech. “Move deliberately and don’t break anything” could be utilities’ obverse slogan to Silicon Valley’s.The question then becomes how do we goose the ability to add a lot of new electric generation quickly, without bogging down in the plodding nature of utility regulation? But this also must be done without breaking things – the affordability and reliability of the electric grid are non-negotiable in the quest to build more power more quickly.

Co-location has been at the center of the institutional debate to date. Big Tech, to its credit, has been moving fast. Co-location, which involves contracting the entire output of an existing merchant plant and removing it functionally from the capacity of the RTO, looks to be a feasible solution to meet immediate electricity needs. Nevertheless, co-location is a finite pool of existing resources, and so long as the co-locator agrees to pick up its fair share of additional costs for interconnecting to the grid, there is an agreement to be had. The threat to the RTO regulatory construct comes from co-location’s removal of existing capacity from the system. An RTO with less system capacity induces higher prices, and those higher prices are paid by other customers than those who co-locate large loads. Big Tech, then, becomes the enemy of the regular customer and a thorn in the side of the RTO. Co-location is a short-term solution with follow-on effects that bother other customers and create a dilemma for regulators.

Further, the co-location move is a once-and-done game. Once the existing co-location capacity is gobbled up, things become more difficult. RTOs with capacity markets like PJM have shown themselves capable of imposing very high capacity auction prices, but not so great at actually inducing new capacity onto the system. Thus, the voracious electricity needs of AI are not going to be served by this type of RTO regulatory model.

The RTO model is particularly challenged in meeting Big Tech’s need for speed. In RTO-centered regions, generation planning has been cleaved from transmission planning. In fact, in some RTO regions, there is not even siloed generation planning because that job has been re-assigned to the invisible hand of the marketplace. In all cases, the central authority (the RTO) has no authority to build generation or treat transmission and generation as the complementary products that they are.

PJM presents as the poster child. PJM faces tremendous load growth from data center demand in the mid-Atlantic. PJM responds to the demand for new generation capacity by…wait for it…ordering $6 billion of transmission! There can be no question that if PJM’s engineers had authority to plan generation along with transmission, they would have met their data center challenge with a combination of local generation and a lot less transmission. That is not a choice PJM’s engineers can make. Transmission can be planned; generation is left to “the market,” which is a labyrinthine set of energy and capacity auction rules that, so far, has failed to induce large quantities of generation into the ersatz market. So even though the heart of the problem is energy supply, not delivery – the solution (because it’s the only tool in PJM’s box) is a big, controversial and suboptimal transmission solution that will take many years to realize.

If your institutional model struggles to bring on generation – co-location looks good. ERCOT is another region where co-location is touted. Not coincidentally, Texas is also a state that challenged with insufficient margins, and one that has turned to heavy state intervention to fix what its markets are not fixing. And ERCOT, through the political guidance of Texas, appears to be willing to cede its attachment to “competitive generation is the only good generation” for outright regulatory dictates to build generation capacity, which is one way to solve a capacity shortfall.

To meet the institutional Sputnik challenge, however, it would be a mistake to take from PJM and ERCOT the lesson that co-location is essential if we are to realize our national ambitions to power a data center-centered economy. Co-location is a necessary (but second best) response to a particular set of institutional limitations where we no longer have holistic and integrated system planning.

Perhaps surprisingly, the AI electricity demand is being better met by the proven model of vertically integrated planning. Data centers looking for new sources of generation and more diversified power pools with holistic transmission planning can access that needed capacity through a planned system, such as existing in the Interior West and Southeast. To be sure, ERCOT can and has been adding capacity by regulatory fiat, but this is an exception to its market, not the rule. And the integrated planning models of the West and Southeast will need to adapt to become faster and more flexible. The U.S. evolved its utility industry to be deliberate and cautious, the AI challenge will require some verve and impatience with past practices.

U.S. utilities and regulatory institutions will have to move faster, will have to be willing to entertain new contracting models, but will also have to get the foundational questions of utility regulation right. Cost allocation and fairness questions abound, and admit to no definitive answers. Interconnecting data centers with the right integrated mix of generation, transmission and distribution, and arriving in a timely fashion at a result that optimizes availability affordability and fairness among customer classes means a traditional grid connection will likely be superior to co-location. Integrated resource planning in a faster and more flexible form is back. Big Tech’s need for speed and big utilities’ need for deliberate regulatory certainty will have to meet somewhere in the middle. U.S. tech companies, U.S. utilities and U.S. regulators are all going to need to rise to the occasion.

Ray Gifford is the Managing Partner of Wilkinson Barker Knauer’s Denver Office and former Chairman of the Colorado Public Utilities Commission. By day, he practices regulatory law before state utility commissions and FERC.