Nvidia’s share price has fallen 47.5% from its 6 October close of $239.24 to $125.53 on 20 November, per Nasdaq data. Meanwhile, AI server makers are demanding up to 45% more on deals already signed, the Financial Times reported on 22 November.
Those two numbers sum up the argument over whether the AI bubble is deflating: share prices and cash flows show strain, while the physical side of the build keeps getting more expensive. Live prices are on our Nvidia stock page.
We checked eight measurable signals, each with an official or primary number and a date. Five point to stress and three to strong demand, by our count. Elon Musk added a ninth worry on Saturday, writing that the “cash shortage is also going to be nuts next year,” as we reported. This is not investment advice.
Business Pill · Free Cash Flow
An evergreen Business Engineer explainer. It teaches the general idea only and does not discuss this report.
The key insight: Of eight AI market signals with official numbers, five point to stress (Nvidia -47.5% since 6 October, a data-centre basket -39.4%, Anthropic near its IPO price, negative cloud cash flow, a 16% wider credit spread) and three to strong demand (Nvidia’s $115.8 billion quarter, server prices up to 45%, Micron’s 2027 supply 75% sold), by our count.
The Five Stress Signals
1. Nvidia’s share price. Down 47.5% from 6 October to 20 November, per Nasdaq. On its 24.1 billion reported shares that is about $2.74 trillion of market value, by our arithmetic. What it shows: investors are paying much less for the same company. What it doesn’t: any fall in Nvidia’s sales.
2. The data-centre suppliers. An equal-weighted basket of the nine US companies in an FT chart fell 39.4% from its 22 June peak to 13 November, by our arithmetic. The FT put the global loss for these ‘picks and shovels’ companies at $4.2 trillion. What it doesn’t show: cancelled projects.
3. Anthropic’s new shares. Anthropic fell as low as $83.05 on Friday, 5 cents above its $83 IPO price, before closing at $88.91, per Nasdaq. That close was 8.5% below its $97.17 debut close. What it doesn’t show: a break below the IPO price, which has not happened.
4. Free cash flow. Five of six AI clouds had negative free cash flow in their latest quarter, meaning they spent more on buildings and chips than their operations brought in, by our arithmetic on their 10-Qs. Amazon was at minus $34.8 billion; only Microsoft had cash left over, $18.6 billion.
5. Credit. The extra yield investors demand to hold riskier US company bonds rose from 3.03 percentage points on 6 October to 3.52 on 19 November, per the ICE BofA index published by the St. Louis Fed. That is 16% wider, by our arithmetic. What it doesn’t show: AI alone; the index covers all riskier US borrowers.


The Three Demand Signals
6. Nvidia’s sales. Nvidia reported $115.8 billion of revenue for its fiscal third quarter, up 103% from a year earlier, and guided to $130.0 billion for the next, per its results release on 17 November. The shares fell anyway, as we covered.
7. Server prices. Dell, HPE, Lenovo and Supermicro are demanding up to 45% more for AI servers months after deals were signed, the FT reported on 22 November. Buyers have largely accepted the increases because they still need the equipment, according to write-ups of the FT report, as our piece explains.
8. Memory. Micron’s chief executive said 75% of the company’s 2027 memory supply is already sold, as reported by Seeking Alpha. Memory stocks fell anyway that week, as we reported.
What the Scoreboard Says
The split is clean. Every stress signal comes from prices of shares and bonds, or from cash flow. Every demand signal comes from the goods themselves: chips, servers and memory.
That gap is the question Musk raised: companies pay for compute now and earn on it later. As we read it, share and bond prices are reflecting the risk that the payback takes longer than planned.
Gartner said in October that more than $3.5 trillion of AI infrastructure spending could be exposed to impairment by 2030 if the current pace of spending continues, as we reported. That is a forecast, not a measurement.
The Structural Read
As we read it, the stress sits in what investors pay for AI assets, and the demand sits in what buyers pay for AI hardware.
As we read it, the gap between the two is the working-capital risk Musk described: money goes out months before the compute it buys earns revenue.
Elon Musk on X, 21 Nov 2026
“Even cumulative industry delays of a few months between paying for compute and getting it reliable enough to generate revenue will mean hundreds of billions of incremental dollars just for working capital.”
Three Implications
FOR INVESTORS Five of eight signals point to stress, all from share prices, bond spreads and cash flow, by our count.
FOR AI CLOUDS Five of six had negative free cash flow in their latest quarter; only Microsoft had cash left over.
FOR CHIP AND SERVER MAKERS Server prices are up to 45% higher after signing, per the FT, and Micron says 75% of 2027 supply is sold.
What Is Not Established
None of these signals shows AI demand falling. Nvidia’s sales, server prices and memory orders all point the other way.
The credit index is not AI-specific, and its latest reading is from 19 November. Anthropic has traded for two days, too short to read much into.
The Bottom Line
Eight signals, five pointing to stress and three to strong demand: share prices, bond spreads and cash flow are weakening, while Nvidia’s sales are up, server makers are asking more and Micron’s 2027 memory is mostly sold. What to watch is Nvidia’s next quarter and whether the credit index keeps widening.
97,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. Share prices are Nasdaq closes; Nvidia’s revenue is from its 17 November results release; free cash flow is operating cash flow minus capital spending in each company’s latest 10-Q; the credit spread is the ICE BofA US High Yield index as published by the St. Louis Fed’s FRED service. The server and memory figures are as reported by the Financial Times and Seeking Alpha. The 47.5%, $2.74 trillion, 39.4%, 8.5%, 16% and indexed chart values, and the five-to-three count, are our arithmetic. Not investment advice; for information only.
Sources: Nasdaq, Nvidia historical closing prices (6 Oct and 20 Nov 2026) · NVIDIA, Q3 fiscal 2027 results release, 17 Nov 2026 (SEC exhibit) · ICE BofA US High Yield Index Option-Adjusted Spread (BAMLH0A0HYM2), Federal Reserve Bank of St. Louis (FRED), daily to 19 Nov 2026 · Financial Times, AI server suppliers raise prices up to 45% after deals are signed, 22 Nov 2026 (headline and standfirst) · Financial Times, data-centre supplier stocks shed $4.2 trillion since June, 15 Nov 2026 · Seeking Alpha via TradingView, Micron expects even tighter supply in 2027-28, Nov 2026 · Anthropic, IPO closing announcement, 20 Nov 2026 · FourWeekMBA, AI capex vs free cash flow (company 10-Qs), 22 Nov 2026









