The gauge fell to 7.4 from 8.1, and with that the sell signal that Bank of America had been running since May ended. Michael Hartnett, the bank’s chief investment strategist, spent much of the note explaining why that should not be read as permission to buy.
The Bank of America Bull and Bear Indicator is a composite of fund manager positioning, flows into and out of asset classes, and credit conditions, scored from zero to 10. Readings above 8 have preceded weak returns over the following weeks and months often enough that the bank treats them as a sell signal. Readings below 2 have done the opposite. It is a measure of what investors have already done with their money, not a forecast of what the economy will do.
That is why the move from 8.1 to 7.4 carries less information than the headline implies. The signal ended because positioning came down, not because assets became cheap. A reading of 7.4 is still high on a scale whose midpoint is 5, which means the average investor is still leaning into risk after a period in which leaning into risk has been expensive.
The composition of the move matters as much as its size. A gauge that falls because cash allocations rose is a different signal from one that falls because credit spreads widened, and the second kind is the one that has preceded trouble in past cycles. Hartnett does not treat the two as equivalent, and neither should anyone reading the number from a headline.
Hartnett names two conditions for a genuine return to risk assets, and he is explicit that neither is present. The first is a Federal Reserve whose guidance the market believes again, a credibility problem that a single meeting does not solve. The second is a reduction in the anxiety surrounding how artificial intelligence is being financed, which is not a sentiment question but a question about the borrowers.
Then he narrows it to one variable that can be observed daily. Semiconductor prices. If they keep rising, he argues, technology shares can be bought back from the middle of November. If they fall, the meaning is different and worse: it says credit for AI is tightening and capital spending is topping out.
The logic is worth following, because chips sit on both sides of the trade. They are a cost for the companies building models, a revenue line for the companies selling them, and a quoted price for everyone in between. Memory in particular is the cleanest read available on whether the buildout is being driven by demand that exists or by capacity that has been ordered. When DRAM and high-bandwidth memory prices climb, the buyers are paying up because they cannot get enough, and the spending plans that support a great deal of equity valuation survive another quarter. When those prices turn, the order book is filling more slowly than the fabs can produce, and the next step in that sequence is a capital budget cut.
A capital budget cut is where an equity story becomes a credit story. The companies that borrowed against contracts to buy equipment need the equipment to earn. If the buyers of computing slow their orders, the resale value of a graphics processor falls, the collateral behind the borrowing weakens, and the spread demanded by lenders widens. That chain has already begun in the lower-rated part of the market, where bonds issued by companies that rent out computing trade at yields that imply distress rather than optimism.
The mid-November window has a practical basis. Third-quarter earnings are through, the guidance has been given, and the de-risking that follows a bad stretch of positioning has usually run its course by then. What Hartnett is describing is a tactical entry, on a gauge built for a horizon of weeks to months, conditional on a price series that has nothing to do with valuation.
His record argues for taking the framework seriously and the timing loosely. The same indicator produced a sell signal in May that was early, in the sense that equities kept climbing for a while, and correct, in the sense that the return over the following months was poor. Gauges of this kind are usually early in both directions. They tell you the crowd is on one side of the boat. They do not tell you when the boat tips.
The other caution is that a tactical call can be overtaken by a cycle. If a credit event arrives in the lower-rated AI complex, the question stops being whether technology shares are a few percent too expensive and becomes whether the financing structure of the buildout survives. Hartnett’s conditions are a way of saying that the market is not yet pricing that outcome. If it starts to, the indicator will fall much further than 7.4, and the signal it produces will be a different one.
What the note really describes is a market in a holding pattern, with investors unwilling to add and unwilling to leave. The gauge measures that reluctance accurately. It cannot say how it resolves. For that, watch the price of memory, which is the only number in this story that is quoted every day and reflects what somebody actually paid.
