Of the three Artificial Intelligence (AI) Executive Orders (EOs) released by the White House on July 23rd, this is the most complex and in many ways the most immediately damaging (for a variety of reasons). So, what’s wrong with it? Well, it’s not just the environmental implications that are problematic, there’s a bigger issue in relation to the vision (or lack thereof) implied by this EO for AI in America.
Problem #1 - (Lack of) Vision
I’m not sure whether this is worse on some type of rating scale than the other issues, in fact, I think it might be comparing Apples and Oranges. But this problem stands out to me as the most immediate because it’s the trigger for all of the other problems. This EO envisions that growth and evolution of AI in the US must necessarily follow the current path of “scaling up and out” ever-larger data centers to ‘grow the models.’ What’s wrong with that you ask? Here’s a just a few reasons:
It seems to forfeit or dismiss other, better evolutionary paths (none of which are even mentioned in this or the other EOs and the National AI Action Plan).
This implies that we’re placing all (or most) of our bets on a technology that we don’t actually understand and that’s likely to be left in the dust by other, better approaches to AI design and architecture. We’ve talked about some of that here before…
The current generation of LLMs will never achieve AGI - this EO seems to leave us stuck on and with LLMs as the end-all, be-all of AI development and adoption. This is beyond short-sided.
Add all of this up and we’ve got a pretty good formula for losing the AI Race; and losing it while encountering the all of the harmful impacts of the associated problems. To be fair, this lack of vision didn’t start with the current Administration, but they have totally bought into it. The Groupthink associated with this stunning lack of imagination is going to be the number one reason why the US doesn’t win the AI race as it is will continue to redirect money and efforts into ever bigger data centers as the AI market further consolidates, locking out all of the innovators.
Problem #2 - Environment
Facilitating the rapid build-out of mega AI data centers (the AI only references data centers costing more than $500 million) is a recipe for disaster. The EO also calls for building these on public lands (and curiously it also mentions superfund sites as well). Most public lands in the US are out West, where we are already suffering enormous water shortages that are only predicted to get worse. Why are we choosing to accelerate the problem - this kind of action combined with other massive water wasting (like Arab countries using the last aquifers out West to grow Alfalfa) could turn much of the American West into Ghost Towns in 25 years or less. Not convinced, check out this study on global fresh water loss. Much of this EO is crafted to ensure that any & all environmental reviews are bypassed, especially those that would look at water issues.
Problem #3 - Energy
The EO also calls out data centers that are using at least 100 Megawatts MW. For comparison, 100 MW could power up to 50,000 homes. Many of the data centers being planned will consume much more power than this. In order to meet this growing energy demand, coal plants that were scheduled to close are being kept open (adding more carbon to the atmosphere) and new nuclear plants are in the process of being built (when we still haven’t figured out what to do with waste from the old ones).
BTW - they went out of there way here - not to mention - renewable energy sources such as Wind and Solar (although they might qualify as other dispatchable baseload energy sources, whatever that means):
(ii) natural gas turbines, coal power equipment, nuclear power equipment, geothermal power equipment, and any other dispatchable baseload energy sources, including electrical infrastructure (including backup power supply) constructed or otherwise used principally to serve a Data Center Project.
Other Problems
There are a few other issues associated with this EO that need to be highlighted; they include:
This isn’t just about streamlining approval, the EO actually includes a provision for Federal subsidies of companies (some of the wealthiest in the world) to pay for data center development. Sec. 3. Encouraging Qualifying Projects.
Revocation of a similar AI EO from the Biden Administration because it emphasized “Clean Power.” Sec. 4. Revocation of Executive Order 14141.
A careful review of all of the ‘streamlining’ provisions in the EO shows a comprehensive elimination of any and all oversight for the approval (and later management) of such data centers including tossing out provisions about toxic waste and endangered species.
What’s perhaps the worst thing here with this EO is that all of the negative outcomes associated with continuing down this path are completely avoidable. In other words, not only could we win the AI race, by not doing this - that is in fact the only way we’ll win it. Instead, we’re going to lose the ‘race’ while getting stuck with an environmental nightmare and a failing national energy grid.
Copyright 2025, Digital Perspectives