Newcomer Pod · 2 min · Jun 8, 2026

Data center expansion depends on an enormous debt market

Ed Zitron · Founder & CEO, EZPR

Ed Zitron estimates that Nvidia’s growth projections could require $800 billion in debt and argues the needed data center capacity is not being built.

The episode’s best minutes

12:01 AI subscriptions hide the real cost of a unit of workPro pick 7 min
33:06Data center expansion depends on an enormous debt marketPro pick 2 min
41:37 Executive demos can make AI look more capable than it is7 min
4 more moments in the full episode

“we're going to need like 800 billion dollars worth of debt”

Ed Zitronat 33:36 · Newcomer Pod

Takeaways Pro pick

Nvidia’s growth targets imply a huge buildout of data centers.
Ed estimates the expansion could require $800 billion in debt.
More compute may expand rate limits without changing what models can do.

About this segment

Ed Zitron estimates that Nvidia’s growth projections could require $800 billion in debt and argues the needed data center capacity is not being built. The scale of financing raises questions about what the new capacity will actually enable.

Ed Zitron is a writer and podcast host.

Hosts Better Offline and writes Where's Your Ed At, a prominent critic of AI industry hype.

EN
Eric Newcomer, Host · Newcomer

From “Ed Zitron Unfiltered on OpenAI, Anthropic & Why the Whole Thing Is a Con”, Jun 8, 2026. Watch the full episode

Answers “Is there still a GPU shortage?” See all 28 answers

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

33:12
like that's what's happening here and that's one of the big things that people really don't want to talk about which is let's say NVIDIA does the thing they're meant to which is they're meant to sell a trillion dollars of gpus by the end of 2027 yeah sell by which i mean put in a warehouse that would i think that's like 15 or something gigawatts of capacity data centers aren't being built at that scale we are watching one of the biggest kickstarters a pre-order campaign of all time for there to be a trillion dollars of sales we're going to need like 800 billion dollars worth of debt we're going to putting aside cynicism or anything just dollars to donuts we're going to need 800 billion dollars worth of debt and for NVIDIA to keep growing at their current rate in a couple years they're going to have to be selling quarter of a trillion dollars worth of
33:55
gpus in a quarter in a 250 billion dollars in a quarter and again the cash flow required to do this is so severe that microsoft google amazon and meta all having to race there all having to do dodgy little um what's called like off balance sheet deals putting aside however we feel about ai that is more money than i think is out there people will say all the money's on the sideline there was a story in the ft saying banks are worried about choking on data center debt for this to continue requires an astronomical amount of money and then there's the messy little question of why what does building more capacity do does open if open ai had 10 gigawatts more capacity today what would they do what did anthropic do when they got access to colossus one from elon musk oh they are up the rate limits on claude yay like i mean the model
34:47
the consensus is certainly that the models have improved yeah most people think the scaling laws are holding true that the more the larger training runs you run the better output you get i mean severely pulled back on pre-training isn't it more the train i mean there are scaling still the training runs obviously they're scaling on inference with chain of thought reasoning and also pre-training is because it's still happening right it's all of the above they are coming up with new techniques to improve the models in addition to trying to run right larger models