Is the AI bubble bursting? What the accounts with track records say

Is the AI bubble bursting? The prices are falling: all 38 AI stocks on our list closed lower on October 16 than a week earlier, and half of them fell 17% or more. The posts about them barely use the word. In the month to October 18, "bubble" came up in 45 of the 5,255 mentions of those stocks from the finance accounts we track, fewer than 1 in 100.

The accounts with a graded track record used it to argue the other way: 11 of their 13 bubble mentions were bullish. Of the 17 bearish ones, 15 came from accounts without a graded record. In the week of the drop, graded accounts stayed three to one bullish on AI names while everyone else came close to even.

What people mean by the AI bubble

The AI bubble is the argument that AI stock prices, and the money pouring into data centres, have run ahead of the revenue AI will earn. The argument rests on four things:

• Capex. Alphabet raised its 2026 capital spending plan to $195 billion to $205 billion in July (W.Media). Meta guided $130 billion to $145 billion and spent $31.08 billion in the second quarter, which left $784 million of free cash flow (Meta).
• Circular deals. Nvidia said in September 2025 it would invest up to $100 billion in OpenAI as OpenAI deploys Nvidia systems (Nvidia). Two weeks later AMD gave OpenAI a warrant for up to 160 million AMD shares alongside a GPU supply deal (AMD). Critics say money that moves in a loop makes demand look bigger than it is.
• Valuations. Nvidia became the first company worth $5 trillion on October 29, 2025 (Forbes Australia).
• Debt. More of the buildout runs on borrowed money. Meta's $27 billion Hyperion data centre sits in a joint venture with Blue Owl, funded in large part with bonds that stay off Meta's balance sheet (Yahoo Finance). The GPU clouds borrow to build too.

Professional investors are split. In Bank of America's October fund manager survey, a net 37% were overweight tech, the most since August 2020 (Crypto Briefing), and an AI bubble was the second most-named tail risk (the unlikely event that would do the most damage), after a disorderly rise in bond yields (ChainCatcher).

How much bubble talk there is, week by week

Our AI list is 38 stocks: the 30 names in our AI infrastructure ranking, from GPU clouds to power and cooling, plus Nvidia, AMD, Broadcom, Microsoft, Alphabet, Meta, Amazon and Palantir. Bubble talk here means a mention of one of them that uses the word bubble, about the stock or about AI, whether the post argues for a bubble or against one.

The week of October 12 had the most bubble talk since late August: 18 mentions, 1.3% of the week's AI mentions. Across the month the accounts we track gave AI stocks a third of their attention (5,255 of 15,701 mentions) and more than half of their bubble talk (45 of 82).

Most of the bullish bars are people pushing back on the bubble idea. The bearish ones bunched up from October 13 to 16, when 9 of the month's 17 bearish bubble mentions landed, most of them on Nvidia and CoreWeave. Reddit carried 21 of the 45, X 22 and YouTube 2.

The graded accounts are more bullish on AI than everyone else

Graded accounts are the ones with a real track record: more than ten resolved calls, more than half of them right. Over the month to October 18 they ran 941 bullish to 151 bearish on the 38 AI stocks, 86% bullish. Everyone else ran 2,610 to 1,113, 70% bullish.

The gap widened when prices fell. In the week of October 5, graded accounts were 87% bullish and everyone else 76%. In the week of October 12, graded accounts slipped to 75% and everyone else dropped to 55%.

Part of that gap is where people post. Most graded accounts are on X, and Reddit did most of the turning: Reddit ran 180 bullish to 293 bearish on AI stocks in the week of October 12, after 250 to 135 the week before. On X alone the gap is smaller but it holds: graded accounts 87% bullish over the month, other X accounts 81%.

Graded accounts got more cautious in the drop without calling a bubble, and stayed three to one bullish. Accounts without a record moved further, from 76% bullish to 55%.

Which AI stocks draw bubble worry, and which draw conviction

The eight AI stocks with the most to say about the bubble question, over the month to October 18:

	Mentions (30d)	Bull : Bear, all	Bull : Bear, graded	"Bubble" mentions	Score	Price (1 month)
NvidiaNVDA	606	402 : 138	91 : 12	16	80	-3.7%
MetaMETA	782	517 : 181	108 : 9	6	59	+6.5%
NebiusNBIS	413	332 : 57	133 : 14	5	77	-26.0%
AMDAMD	437	315 : 88	42 : 7	3	72	-5.5%
CoreWeaveCRWV	281	98 : 164	18 : 19	3	41	-24.5%
BroadcomAVGO	159	110 : 42	29 : 14	0	81	-8.1%
OracleORCL	130	37 : 69	10 : 5	1	27	-13.1%
IRENIREN	208	88 : 96	19 : 13	1	66	-42.2%

Most bubble worry: Nvidia. It drew 16 of the 45 bubble mentions, 8 of them bearish, and all 8 bearish ones came from accounts without a graded record. The worry was about circular financing and what happens to the supplier when the buyers stop. Graded accounts ran 91 to 12 bullish on Nvidia, and it scores 80. The stock is 3.7% lower over the month and 10.5% below its October 6 close.

Where the bears were: CoreWeave and Oracle. These two drew more bearish than bullish mentions overall, on debt and on how they will pay for the build. Graded accounts split evenly on CoreWeave, 19 bearish to 18 bullish, and the score is 41. Oracle scores 27, the lowest of the eight, though the few graded accounts on it leaned bullish, 10 to 5. Both fell more than the AI list's median of 11.8% over the month.

Most conviction: Nebius and Meta. Graded accounts ran 133 to 14 on Nebius through a 26.0% fall, and 108 to 9 on Meta, which rose 6.5%. On Nebius, graded accounts used the word "bubble" to mock the people calling one. CKCapitalxx put it this way: "the bubble call stops being a thesis and starts being a personality."

The exception: Broadcom. It is the one name here where graded accounts were warier than everyone else, 29 to 14 against the crowd's 110 to 42. Its score is 81.

The bear case, in two sentences

The bear case is that the spending has outrun the revenue: Meta kept $784 million of free cash flow from a quarter in which it spent $31.08 billion on capex, and chip suppliers such as Nvidia and AMD now help finance their own customers. If the revenue comes late, lenders feel it first, and the bearish posts we tracked in October were about CoreWeave's bonds, neocloud debt and circular deals.

Common questions

Is the AI bubble bursting?

We do not predict tops. AI stocks fell hard in the week to October 16, 2026: all 38 on our list closed lower, and half fell 17% or more. But bubble talk stayed under 1% of AI stock mentions from the accounts we track, and graded accounts stayed three to one bullish on AI stocks in that week.

When will the AI bubble burst?

A date for the burst is a guess, and we do not make one. You can watch the three things the bear case rests on: the big tech companies' capex plans, the cost of debt for the GPU clouds, and whether the accounts with good track records start turning bearish. In October 2026 they had not.

What happens if the AI bubble pops?

If it follows the dot-com pattern, it is slow and deep: the Nasdaq Composite fell 78% from March 2000 to October 2002 (Wikipedia). The bearish posts we track expect the debt-funded names, the GPU clouds and small caps, to fall first and hardest, and the profitable giants to follow later.

Is Nvidia in a bubble?

Nvidia draws more bubble talk than any other AI stock we track: 16 of 45 bubble mentions in the month to October 18. All eight bearish ones came from accounts without a graded record. Graded accounts ran 91 bullish to 12 bearish on Nvidia over the same month, and its score was 80 on October 21.

What is the AI bubble?

The AI bubble is the argument that AI stock prices and AI data centre spending have run ahead of the revenue AI will earn. It points to record capex, circular deals between chipmakers and AI labs, high valuations and rising debt.

Where to read the room yourself

A price chart shows you the selling. To see who changed their mind, sort the posts by track record: in October it was the accounts without one, while graded accounts got more careful and stayed bullish. For the names on the infrastructure side, start with the AI infrastructure stocks ranking.

The Quantral app shows this reading for every stock, as it changes: the score, every bull and bear call behind it, and the record of the account that made each one.

Mention windows are by US Eastern day: weekly Monday to Sunday, August 24 to October 18, 2026, and the month September 19 to October 18. Mentions are posts from the accounts Quantral tracks in which the stock is the subject; bullish and bearish are by the direction of each post. "Bubble" mentions use the word bubble about the stock or about AI. Graded accounts have more than ten resolved calls, most of them right. The AI list is our own grouping of 38 stocks: the 30 names in our AI infrastructure ranking plus Nvidia, AMD, Broadcom, Microsoft, Alphabet, Meta, Amazon and Palantir. Price moves use daily closes: 1 week is October 9 to October 16, 1 month is September 18 to October 16, and "from the high" is from the highest close since July 1. Scores are the seven-day reading as of October 21, not historical values. Company figures are as published by the companies and the outlets linked. Positions attributed to accounts are their views, not Quantral's. Nothing here is financial advice or a recommendation to buy or sell.