Something like $800 billion of capital spending is going into artificial intelligence in the United States this year, David Sacks said, and he would not be surprised if the figure reaches $1 trillion.
The usual answer to a boom that size is a new agency to supervise it. Sacks told the G20 the United States should not build one, and said the ministers there were asking how to copy Silicon Valley rather than how to police it.
"You don't change jockeys in the middle of the race when you're winning, the policies are working."
Sacks co-chairs the President's Council of Advisors on Science and Technology, was the White House AI and crypto czar, and came to this interview straight from putting that case to the G20.
I listened to the full segment so you can skip it.
Here are the 4 takeaways that matter.
👤 Guest: David Sacks, co-chair of the President's Council of Advisors on Science and Technology and former White House AI and crypto czar, who had just argued the case against an AI regulator to the G20
🎙️ Host: Carl Quintanilla, a co-anchor of Squawk on the Street on CNBC
📰 Published: 1 September 2026 on CNBC (Squawk on the Street)
🟢 Spotify | 🟣 Apple Podcasts | 🔗 Episode page | ⏱️ 4 min
Key Takeaways
AI capital spending forecasts are being revised up every time they are published He put this year at something like $800 billion and next year's forecasts at $1.4 trillion
His case against a new AI regulator is that the existing statute book already reaches AI Fraud, cybercrime and non-consensual images are already illegal, and he would use those first
An approval agency cannot run at the speed of a product that ships every two months The FDA and the FAA take over five years to clear a new drug or a new airplane
The risk he names is regulatory capture, and the victims are startups rather than incumbents Large companies can afford the process; small ones get boxed out of the market
He credits the boom to policy, not only to the technology He named innovation, infrastructure and energy as last year's priorities that set it up
1. The message to the G20
Quintanilla asked Sacks to tell viewers what his message to the G20 had been.
Sacks said the scale of the investment is what other governments are reacting to, and that the numbers keep moving up. "Something like $800 billion of capex is being invested this year. I've seen forecasts for 1.4 trillion next year." Capex is capital spending — the money going into data centers, chips and power rather than into running costs
He expects this year's own figure to be revised past a round number. "So I wouldn't be surprised if we already hit $1 trillion of new capex investment this year."
He said countries around the world are looking at the United States with envy because the build-out is accelerating the American economy
He credited policy rather than the technology alone. Sacks said the innovation was unleashed by President Trump's policies last year, which he said prioritized innovation, infrastructure and energy, and that those policies paved the way for the boom
2. Don't change jockeys
Quintanilla put it back to him as a question about interference: the point now is not to throw too much sand in the gears, right?
Sacks agreed, and his argument is that the current settings are working and should be left alone. "You don't change jockeys in the middle of the race when you're winning, the policies are working."
He said the companies are policing themselves adequately. "And look, I also think that our AI companies are behaving responsibly. On the whole. I don't think we have a need here for overregulation."
The stronger half of the argument is that AI is not unregulated today. "And let's not act like we are bereft of regulations. We already have a ton of laws and statutes in this country that apply to AI." He named laws against fraud, cybercrimes and non-consensual images, and said the list goes on His position is to make full use of those before creating new agencies
3. Why an FDA for AI fails
Sacks took on the specific proposal by name — an FDA for AI, or as he said it is sometimes described, an FAA for AI — and answered it on timing. He said those agencies take over five years to approve new drugs or new airplanes
The mismatch is with the release cycle. "That is not the right model for an AI industry that is releasing new AI models every two months."
His second objection is that the government could not staff it. "This is a very dynamic industry. It is moving very fast. And frankly, the federal government does not have the technical talent and capability to approve models that are moving that fast."
4. Capture boxes out startups
Quintanilla played devil's advocate: the United States still has an enormously innovative pharmaceutical industry despite the slowdown, so would an FDA-style regime for AI simply need to be global to work, given the need for speed and to stay competitive?
Sacks answered with what the drug industry says about its own regulator. "Well, I hear a lot of people from the biotech industry or the pharmaceutical industry saying that we need to reform the FDA, that it takes too long, it's too costly." The cost of the process, he said, favors only the big companies that can afford to go through it
The failure mode he named is regulatory capture, and it lands on the smallest firms. "And that is the big danger when you create a new regulatory agency, is that you get regulatory capture, the big companies figure out how to comply, but the little startups don't, and they get boxed out."
Against that he set what he calls permissionless innovation — the ability to build and sell a product without asking a government first. "You can have a founder who is a college dropout who can create their product in a dorm room, or you could have two founders who are PhD students create their product in a garage, and they don't need to go to Washington to seek approval, which they can never figure out."
He put himself and one of the largest companies in the world on the same side of that example. "That is how I got started. That is how Google got started."
Bonus Insights
Quintanilla framed the segment before Sacks spoke, telling viewers he had just come from the G20, where he said the United States should not create a regulatory body for AI similar to the FAA or the FDA on the grounds that it would hamper model improvement
The demand at the G20 was for the recipe, not for the rules. "And all these ministers who are at the G20 are trying to understand what is the secret sauce that has made Silicon Valley so successful, because they're trying to emulate that."
Sacks's bottom line is that the United States is winning the AI race on policies already in place, that the laws on fraud and cybercrime already reach the industry, and that a purpose-built approval agency would move too slowly to matter and would shut out the small companies rather than the large ones.
Products, Companies & Tools Mentioned
Google (Named alongside his own start as a company built without asking Washington for approval first)
The Food and Drug Administration and the Federal Aviation Administration (The two agencies proposed as models for an AI regulator; Sacks said both take over five years to approve a new drug or a new airplane)
The G20 (Where he made this argument to finance ministers hours before the interview, and where he says the questions were about copying Silicon Valley rather than regulating it)
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: