Nvidia is down more than 25% and 59 of 64 brokers still say buy. Test the 2027 spending case instead

Two brokers changed their minds about Nvidia on 11 November. Neither said demand is shrinking.

Baird cut the stock to neutral, citing among other risks limited visibility into 2027 data-center spending. Morgan Stanley took Nvidia off its Top Pick list but kept an overweight rating. Reuters reported both moves.

Both are shifts in analyst conviction. Neither move reports a cut to customers' capital budgets or quantifies a revision to Nvidia's revenue outlook.

CTOL's view is that 2027 AI infrastructure spending now belongs in a model as an explicit risk. A forecast built on another uninterrupted spending cycle should survive a slower one. CTOL's confidence is moderate that the two moves justify testing dependence on 2027 spending before treating the bullish consensus as comfort. They support little more: no demand collapse, no lower base-case revenue estimate and no basis for a numerical valuation haircut.

The number everyone quotes is 59 of 64

Reuters reported that 59 of 64 brokerages still rated Nvidia buy or strong buy, or 92.2% of the sample. It also reported that the shares had fallen more than 25% since Morgan Stanley named Nvidia a Top Pick the month before.

Both facts hold at once because they measure different things. The ratings count tallies opinions at one moment. The share price records what investors lost over a period. A large buy majority sat alongside the loss, so it protected nobody.

The split of that fall among 2027 expectations, export restrictions, leadership risk and a lower multiple stays open. The drop also leaves the stock's cheapness unsettled, since that claim needs a bridge from expected sales and cash generation to the price paid.

The number that matters is 2027 customer spending

A spending worry reaches the share price through a chain of four stages. Customers' 2027 capital budgets and commitments drive equipment orders. Orders, their timing and Nvidia's share of them drive its data-center sales. Expected sales and cash generation then set what an investor pays today.

Each stage has its own measure: customer spending, Nvidia revenue and the share price. Any one of them can move while the others hold, so none stands in for another.

That makes a conditional question the most useful one. If 2027 customer spending grows more slowly than the valuation assumes, does the Nvidia thesis still hold? The answer is a downside sensitivity rather than a forecast that spending will contract.

A numerical "visibility discount" would be worse. It would float free of any customer-spending series, earnings bridge or valuation multiple.

Two warnings sit inside one headline

Baird's reported worries come in three parts: limited visibility into 2027 spending, possible demand softening after years of AI investment and uncertainty over how long Jensen Huang stays beyond next year. Customers spending less, a supplier losing access to a market and a change of leadership can each hurt Nvidia. They are separate stories, and a reader should keep them separate.

Morgan Stanley did less. It dropped a relative Top Pick label, stayed overweight and cited near-term uncertainty and the importance of Huang's leadership. That positive stance neither confirms Baird's spending concern nor proves it wrong. Neither action puts a number on a 2027 shortfall.

The Reuters account also puts China-chip restrictions, regulatory pressure and Huang's future beside the spending concerns. Weaker analyst enthusiasm could therefore reflect Nvidia-specific access or leadership risk rather than a broad retreat in AI investment. Two changes in brokerage preference are too thin to establish an industry downturn.

Customer commitments settle it by 30 April 2027

CTOL will revisit the call by 30 April 2027, using comparable disclosures then available on customers' 2027 data-center spending plans and Nvidia's subsequent data-center performance. The measure is the direction of like-for-like customer spending commitments, and the count of buy ratings stays out of it.

• Firm or rising 2027 commitments weaken the case for spending-led caution.
• Cuts to those commitments, especially if followed by softer Nvidia data-center performance, strengthen it.
• Revenue deceleration alone settles little, because revenue and aggregate customer capital spending are different measures.
• Missing customer disclosures at the checkpoint will count as silence, and silence is no evidence of a cut.

A shareholder should find out how much of the thesis rests on 2027 customer spending and whether the position survives if that spending slows. Operators and lenders need evidence about their own operation or borrower before changing orders or terms. Expose and stress-test the spending assumption today. Change the view when customer commitments show which way the investment cycle is heading, whatever the next ratings tally says.