Ken Griffin Calls AI a Bubble While Pending Home Sales Hit a Record Low
AI is supposed to be the one thing that can lift everything else.
AI is supposed to be the one thing that can lift everything else. Then Ken Griffin says it is “almost certainly in a bubble,” Jon Gray says it is not, and pending home sales fall to the lowest level in the National Association of Realtors’ 25-year tracking history.
That is not a tidy market story. It is three warnings arriving in the same week, wearing different clothes.
Griffin, the Citadel chief executive, did not say AI was worthless. He said investors were “almost certainly in a bubble.” He also said that radical economic change often comes with a bubble attached. That distinction matters. A useful technology and an overpriced market can occupy the same address at the same time.
Gray, Blackstone’s president, disagreed. His argument was straightforward: past bubbles built capacity and hoped demand would appear. In AI, he said, the problem is keeping up with demand.
Both men have money riding on the answer. That does not make either wrong. It does mean the debate is not a neutral seminar.
The disagreement is about timing
Griffin’s concern is the gap between spending and proof. Companies are buying models, chips, data centers, consultants, and software. The bills arrive now. The productivity gains are harder to locate on a quarterly income statement.
That is a familiar problem. In 2000, the internet was real. The companies were often not. In 2008, housing was real. The financing wrapped around it was not. The lesson was never that the underlying idea was fake. The lesson was that a real idea can still be financed badly.
Gray’s answer is demand. Blackstone sees customers competing for compute, power, chips, permits, and capital. He described AI demand as “almost parabolic.”
Demand is a better defense than a slogan. But demand alone does not settle valuation. A project can have eager customers and still fail to earn enough after depreciation, financing costs, and the enormous upfront bill for infrastructure. Those are the numbers that decide whether an AI buildout is a business or a very expensive belief.
Housing is the quieter number
The housing report did not make headlines like Griffin’s warning. It should have.
Pending home sales fell 7.3% in September from August. They fell 11.3% from a year earlier. The index reached 66.0, the lowest level since the National Association of Realtors began tracking the series in 2001.
Pending sales count signed contracts, not completed closings. They are usually a leading indicator for existing-home sales. When they collapse, the weakness tends to show up in closed sales later.
The National Association of Realtors attributed the drop to fast-rising mortgage rates. Contracts fell in all four U.S. regions.
That is not an AI story on its face. It is a cost-of-money story. Mortgage rates above 7% changed what a household can afford. The housing market is discovering that arithmetic before the stock market finishes deciding what AI is worth.
The anniversary nobody needed
The same week carried the 39th anniversary of Black Monday. On October 19, 1987, the Dow Jones Industrial Average fell 508 points, or 22.6%, in one session. It remains the largest one-day percentage decline in U.S. stock market history.
Anniversaries do not cause crashes. A date on a calendar is not a forecast. But the timing was convenient for anyone already uneasy about concentrated bets, borrowed money, and markets that assume good news will keep arriving.
I remember 1987 for the speed. I remember 2000 for the paperwork: companies with revenues, losses, and stories that did not survive contact with cash flow. I remember 2008 for the leverage. The assets were familiar. The financing was not.
AI has elements of all three, without being identical to any of them. The technology is real. The spending is enormous. The debt and infrastructure commitments are growing. The unresolved question is whether the returns arrive fast enough to support the capital already committed.
What is known, and what is not
What is documented is simple. Griffin warned that AI is “almost certainly in a bubble.” Gray disagreed and pointed to demand outrunning supply. Pending home sales reached a record low in the Realtors’ 25-year history.
What companies claim is broader. Blackstone says its portfolio companies’ spending on AI models has risen sharply, and that bottlenecks in power, chips, permitting, and capital constrain supply. Those are claims from a firm with a large stake in data centers and energy assets. They deserve attention, not automatic belief.
What remains unknown is the important part. No one yet knows which AI projects will earn their capital costs. No one knows how quickly corporate spending turns into measurable productivity. No one knows how much of the current enthusiasm is already reflected in prices.
The part that concerns me
The concern is not that Griffin used the word “bubble.” People use that word when markets are expensive, and sometimes they are wrong.
The concern is the pattern. Housing demand is breaking under financing costs. Credit markets are showing strain around AI-related borrowing. Two senior Wall Street figures cannot agree on whether the largest investment theme of the moment is sustainable.
When the people closest to the money begin using different words for the same set of facts, the market is no longer arguing about technology. It is arguing about price.
Griffin may be early. Gray may be right that demand is extraordinary. Both can be true for a while. Markets can stay expensive longer than skeptics expect, and useful technologies can still produce painful losses along the way.
The housing number is the reminder. Buyers did not stop wanting homes because homes stopped mattering. They stopped signing contracts because the monthly payment stopped making sense.
AI investors should hope the same arithmetic does not eventually apply to data centers, chips, and models. If it does, the argument will not be about whether AI changes the economy.
It will be about who paid too much to own the change.