Capital Economics has placed the Western economy in the late stage of a bubble. James Reilly, the London firm's senior markets economist, examined eight types of market indicator and found most of them already at, or close to, the levels that have historically preceded a peak in share prices. Equity and debt issuance is rising sharply. Market capitalization is concentrated in a handful of large technology companies. Earnings growth forecasts for the biggest indexes look progressively less sustainable. The closest analogue Reilly finds for the present is the months before a crash, the top of the dot-com boom among them.
The AI boom, according to Fortune, is accompanied by large-scale misallocation of capital, and financial markets are under heavy pressure from the scale of the build-out. Reilly's conclusion from the full set of data is the same: most of the factors he studied have reached or nearly reached levels that in the past came before a market top.
While economists work out how unstable the situation is, the Federal Reserve has to decide whether to apply the emergency brake or hope the euphoria resolves without a crash. Kevin Warsh, the Fed chair, is expected to announce an increase in interest rates later the same day. Higher rates make borrowed money more expensive and, in the intended case, reduce consumer and corporate spending, which helps slow inflation.
Some analysts think the size of the inflation problem is overstated. Core consumer price inflation — the measure that excludes volatile food and energy prices — reached a new post-pandemic low in August, per Fortune.
UBS expects the Fed to lower its long-run inflation forecast and raise interest rates at the same time. The firm says that combination has not happened before; one of its analysts called it highly unusual and unique.
That pairing is the most revealing detail in the whole picture. A central bank that cuts its own inflation outlook while tightening is not fighting the number it publishes. The reading that makes both moves coherent — and this is my inference, not anything Warsh has said — is that the target is asset prices rather than the consumer price index. If so, this would be the first rate decision of the AI cycle aimed at the market itself rather than at the cost of living.
The trouble is the one CNN identifies: a rate rise could cool almost every segment of the market except the segment doing the most damage, the rapidly growing investment in AI. In that case the warning signals keep flashing even with the braking measures fully engaged.
That deserves more weight than it usually gets, because it describes a failure of the instrument rather than a lag in its effect. Rate rises work by making capital expensive. Spending that does not respond to the price of capital is not responding to expected returns either — it is responding to the fear of being absent from the thing everyone has agreed will matter. That is closer to a working definition of a bubble than any single item on Reilly's list of eight.
And nothing in these forecasts names the tool that would work instead. Reilly's categories describe conditions, not timing; markets have sat in the late stage of a bubble for years at a stretch. If the Fed is right about the overheating and its own tightening leaves the overheated part untouched, the discussion has run out of instruments at exactly the point where it needs one.
The uncomfortable conclusion is not that a crash arrives this quarter. It is that the AI build-out may now be large enough, and funded in a way indifferent enough to borrowing costs, that the central bank can raise the price of money and watch the largest trade in the market keep bidding anyway. A regulator discovering the limits of its reach in public is a slower problem than a crash, and a harder one to correct.