Sacks says hyperscalers plan colocation power generation that will lower residential rates through grid donation and economies of scale.
“When I've talked to the hyperscalers, they also tell me that their plan is not to draw from the grid, it's to basically set up their own power generation behind the meter. It's called colocation.”
Sacks cites $350 billion hyperscaler capex this year, heading to $1 trillion by 2030 per Jensen.
“Again, just the top four hyperscalers, $350,000,000,000 of expected capex this year. On its way, I think Jensen said 1,000,000,000,000 by 2030.”
Sacks says hyperscaler capex will exceed $700 billion this year, up from $350 billion last year, over 2% of GDP.
“This year, I think we were supposed to have $660,000,000,000 of hyperscaler CapEx up from $3.50 last year. I think there's now the new estimate is is gonna be over 700.”
Sacks identifies GPU oversupply as the primary AI infrastructure risk, comparing it to dark fiber after the dotcom crash, particularly if compute was built expecting $30-50 per watt pricing.
“The biggest risk, to me, is not on the demand side, the biggest risk is that you get a glut of compute and you get an overbuild. And, in the same way we had dark fiber after the dotcom crash, if you have dark GPUs that would be a disaster for everyone, especially if you built out your compute infrastructure expecting a spot price of $30-$50/watt, as, you know, Elon said they were expecting, right.”
Sacks details that Elon plans to add six to eight gigawatts requiring $300-400 billion capex, and suggests NVIDIA is creating credit lines through banks and private equity for downstream purchasers.
“So, just to take one example. Elon plans to add somewhere around six to eight gigawatts next year. We know that that would cost three to four hundred billion dollars of Capex. The company just raised a hundred billion in its equity and debt offerings. So obviously, they would have to go out and finance that somehow. And as we talked about on our previous episode, the simplest way to finance it would be to get seller financing from NVIDIA. . .so now, Jensen is creating, you could say, the line of credit using these big banks, using these big private equity shops. And he's making that available and that's going to now benefit all these downstream purchasers.”
Sacks suggests that political obstacles to building data centers paradoxically prevent infrastructure overbuilding.
“in a weird way all the political headwinds I think insure against that outcome, because it is so hard to build data centers,”
Sacks warns that overbuilding computing infrastructure could create unused GPU capacity similar to the dark fiber problem after the dot-com crash.
“The biggest risk is that you get a glut of compute and you get an overbuild. And in the same way that we had dark fiber after the dotcom crash, if you had dark GPUs, that'd be a disaster for everyone.”
Sacks argues Trump's requirement that AI companies build their own power generation prevents residential price increases and can lower prices
“President Trump was ahead of the curve requiring AI companies to build their own power generation so new data centers don’t raise electricity prices for residential ratepayers. Done right, data centers actually lower prices by producing excess power and funding grid upgrades.”