.jpeg)
Could these 6 OpenAI problems spill into real estate?
1. GPT-5's flop
Remember when GPT-5 was supposed to change the world?
Sam Altman said it would embarrass prior models. Scale was everything. More compute meant bigger leaps, which would justify the cash burn.
Then it flopped, and OpenAI restored the prior model within 24 hours.
The news cycle moved on, but so did OpenAI's users and competitors.
2. First mover ≠ moat
Anthropic, Google, and Grok are matching or beating GPT-5 on benchmarks, and DeepSeek (drafting behind ChatGPT's progress) matched it at a fraction of the cost. Is OpenAI beginning to look a lot like Netscape?
3. Last resort
In Oct. 24, Sam Altman said ads would be a last resort to try to make sene of the OpenAI business model. Guess what OpenAI announced last month?
4. Insatiable cash burn
$8B burned on compute annually on $20B in revenue. Negative cash flow through 2030. $1.4T committed to infrastructure. OpenAI needs to hit Google's revenue in four years.
5. Revolving door
The researchers who built the ChatGPT are leaving to start well-funded competitors. Talent concentration was OpenAI's edge. That edge is dispersing fast. And several not-so-flattering perspectives on Altman have emerged, but that's another story.
6. Microsoft cuts both ways
OpenAI is 45% of Microsoft's $625B revenue backlog. If OpenAI stumbles, Microsoft's AI narrative stumbles with it, and $100B+ in annual infrastructure spend becomes a harder story to tell.
In sum, OpenAI has some very real problems.
----- CRE spillover? -----
Data center real estate has had no bad news for three years, and the amounts investing in this space are so large they're hard to comprehend.
Perhaps no better example than OpenAI's Abilene, Texas deal...
Crusoe and Blue Owl put in $5B in equity behind $10B from JPMorgan in construction financing. The debt closed because Oracle signed a 15-year NNN lease before a shovel hit the ground. OpenAI is Oracle's subtenant.
Strip away the AI narrative and this is a net lease deal. Fortune 100 credit tenant, 65% LTV, JV promote. Similar to a large industrial transaction. Just 100x the size.
Data center investments reportedly accounted for more than 30% of all real estate private equity capital last year.
If OpenAI's problems exacerbate, it could go away just as fast as it showed up but would take lots of capital with it, which could send meaningful jolts through capital markets.
On the other hand, data center and infrastructure investing seem to be cramming out real estate capital. Anyone want to take the opposite side of the thesis and suggest that a data center shock could be good for real estate capital?
PS -- We're running the first AI benchmarking study across CRE's 4M workers. 5 minutes. All respondents get results before public release. Would appreciate your feedback. See comments.
COMMENTS