AI Bubble or Not? Why AI and Data Centre Stocks Are Crashing in 2026
The Nasdaq fell 10% in a week, Nvidia lost its CEO and Trump threatened to tax overseas data centres. What's real, what's hype, and what it means for Indian AI stocks.
Microcap Specialist, Research Analyst and 7+ Years in Markets
Short answer: parts of the AI stock market were priced like a bubble. The AI business itself mostly is not. This selloff is about share prices and borrowing, not demand. Before you sell, or buy the dip, it helps to tell those two things apart.
November 2026 has been the worst month for the AI trade in years. On November 5, about $2.5 trillion was wiped off the Nasdaq in a single day. A week later Nvidia's founder-CEO Jensen Huang resigned. Then the US tightened AI chip exports, and on November 20 Trump threatened to tax US companies that build data centres abroad. That day the Nasdaq fell 5.28% and erased its gain for the year. In India, E2E Networks, ESDS, Netweb, Anant Raj, Sterlite Tech and HFCL fell in step with Wall Street.
The odd part: Nvidia then reported its biggest quarter ever, with revenue up 103%. Prices are falling while the AI business keeps growing. This article explains why, with real examples from the US and Indian markets, and what could happen next.
[IMAGE 1: Bubble or reset? (01_hero.png)]
The current situation: five shocks in three weeks
Date (2026) | What happened | Market impact |
|---|---|---|
Nov 20 | Trump says US firms will be barred from building data centres abroad or face a "tremendous tax". Anthropic gives back much of its debut gain. Reuters reports the SEC is examining Nvidia's accounting | Nasdaq -5.28%, wiping out its 2026 gain. Nasdaq 100 touches bear-market territory (-20% from high) intraday. S&P 500 and Dow close in correction |
Nov 16 | US Commerce Department rule: licences now needed for advanced AI chips going to Singapore, Malaysia, Thailand, Indonesia and Brazil | Hits chipmakers and overseas data centres that rent compute to Chinese firms |
Nov 12 | Jensen Huang resigns as Nvidia CEO amid a board probe into reports he was targeted by a foreign intelligence operation. CFO Colette Kress becomes interim CEO | Nvidia and linked stocks slide for days |
Nov 5 | Democrats sweep the US midterms, including key data-centre states. Investors fear moratoriums and tougher AI rules | Nasdaq -5.5%, S&P 500 -4.4%, about $2.5 trillion of Nasdaq value wiped out |
Oct-Nov | US inflation runs hot (October producer prices +6.5% year on year); the Fed is expected to hike again in December | Debt-funded AI builders squeezed by higher borrowing costs |
[IMAGE 2: Timeline of the November AI selloff (03_timeline.png)]
The damage adds up. In the week to November 20, the Nasdaq Composite fell 10.0%, the S&P 500 8.6% and the Dow 8.0%. The S&P 500 has now fallen seven sessions in a row. Nvidia fell from $186.64 on November 6 to $155.46 on November 17, about 17%.
Then came the twist. On November 17 Nvidia reported fiscal Q3 FY27 (Aug-Oct 2026) revenue of $115.8 billion, up 103%, with data-centre revenue up 110% to $107.6 billion, and guided for about $130 billion next quarter. A day later Anthropic priced its IPO at $83, above its range, valuing it near $1.5 trillion. Demand looks real even while prices fall.
Breaking: Trump's "tremendous tax" on overseas data centres
Yes, this news is real. At a White House press conference on Friday, November 20, Trump said US companies would not be allowed to build data centres in other countries, and those that do would face a "tremendous tax". Reuters, CNBC, AP and Axios all reported it. He argued China would end up with those data centres, and singled out Google's investment in Finland as money that should be spent in the US.
What we don't know yet: how the tax would work, how big it would be, and whether he would use executive powers or need Congress. No rule or bill has been published.
Impact on the Nasdaq and AI stocks
Index (Nov 20 close) | Day | Week |
|---|---|---|
Nasdaq Composite (22,617.73) | -5.28% | -10.0% |
S&P 500 (6,910.39) | -3.70% | -8.6% |
Dow Jones (48,147.33) | -1.93% | -8.0% |
The Nasdaq erased its 2026 gain, and the Nasdaq 100 fell 5.6%, briefly entering bear-market territory (20% below its high). Cloud giants slid after the remarks: Amazon was down about 3.6%, Microsoft 2.7% and Alphabet 1.1% that afternoon.
To be fair, the comments landed on a market that was already falling. The Nasdaq was down 3.6% before the press conference, dragged by Anthropic giving back much of its debut gain and a Reuters report that the SEC is examining Nvidia's accounting. The tax threat added to the selling; it did not start it.
Why it matters for AI stocks: hyperscalers grow by building close to customers worldwide. Now they are squeezed from both sides. Abroad, a tax could make new sites more expensive. At home, local opposition, power shortages and stricter oversight after the midterms are already slowing projects, with about half of proposed US AI data-centre projects possibly facing tighter local review.
What it could mean for India
This matters more for India than most headlines suggest. India's data-centre boom leans heavily on US cloud giants: hyperscaler-backed co-location makes up 85-90% of revenue for major Indian data-centre developers (ICRA). Google's planned $15 billion AI hub in Visakhapatnam is exactly the kind of project the policy targets.
Indian markets were closed when the remarks came, so the first reaction will show on Monday, November 23. Watch data-centre builders and suppliers like Anant Raj, Sterlite Tech, HFCL and Black Box. Domestic players building for Indian clients, such as TCS with its 1 GW plan, are less exposed.
The reality check: this is a spoken threat, not a law. It may soften in talks, carry exemptions, or face legal challenges. But until details arrive, it adds uncertainty to every overseas data-centre plan, including India's.
Global AI stocks that crashed
The hardest falls came from chips, memory and neoclouds: companies that sell picks and shovels to the AI build-out, or borrow heavily to build it.
Stock | What it does | The hit |
|---|---|---|
Nvidia (NVDA) | AI GPUs, over 90% of revenue from data centres | -9.2% on Nov 5. $186.64 to $155.46 in 11 days (-17%) after the CEO exit |
AMD (AMD) | AI GPUs and CPUs | -17.3% on Nov 5. Another -11% on Nov 13 |
Sandisk (SNDK) | Flash storage for data centres | About -20% on Nov 5 |
Broadcom (AVGO), Micron (MU) | Custom AI chips, HBM memory | Double-digit falls on Nov 5. Broadcom -6% again on Nov 13 |
CoreWeave (CRWV) | Neocloud: rents Nvidia GPUs, funded with debt | -6.9% on Nov 12, -6.1% on Nov 13. Down 37.8% over the past year |
iShares Semiconductor ETF (SOXX) | Basket of chip stocks | -11% on Nov 5, its worst day in six years |
The pattern: the more a company depends on borrowed money or on future AI spending, the harder it fell. CoreWeave funds its growth with loans tied to customer contracts, so every rate scare and every hint of slower spending hits it twice.
The moves cut both ways. CoreWeave jumped 24.7% on November 11 after earnings, then gave it all back. After Nvidia's results, memory names bounced hard: Sandisk rose about 13% and Micron and Arm about 11% in a single session. That kind of whipsaw is typical of a market unsure what things are worth.
Indian AI and data centre stocks that crashed
India has no Nvidia, but it has an "AI proxy" basket: GPU cloud providers, server makers, data-centre builders, and the cable and power-equipment firms that supply them. That basket fell almost tick-for-tick with Wall Street.
Stock | AI / data-centre link | The hit |
|---|---|---|
ESDS Software Solution | Cloud and data-centre services | Rs 1,864.35 (Sep 24 high) to about Rs 1,147 (Nov 9): -38% |
E2E Networks | GPU cloud (India's closest neocloud) | Rs 769 (Oct 14 high) to about Rs 490.50 (Nov 9): -36%. Seven straight lower circuits in October, another on Nov 6 |
TD Power, HFCL, Sterlite Tech | Power equipment, optical fibre | Fell up to 15% on Nov 6 |
CG Power, Hitachi Energy, MTAR Tech | Power and electrical gear for data centres | Fell up to 8% on Nov 6 |
Black Box | Data-centre networking and IT services | Down more than 6% on Nov 6 |
Anant Raj | Real estate developer building data centres (Anant Raj Cloud) | Down more than 5% on Nov 6 |
Netweb Technologies | AI servers and HPC systems | Repeated 5-7% one-day falls during each global AI selloff |
[IMAGE 3: Drawdown from recent highs (02_drawdowns.png)]
What makes India's case interesting: the wider market did not crash. The Nifty and Sensex gained roughly 10% in November through Nov 17, and FIIs bought Rs 33,340 crore in a single week. Money left the AI proxies and went into banks and large caps. That is a valuation reset in one pocket, not a market-wide panic.
The businesses also look healthier than the share prices suggest. Netweb had a Rs 2,629 crore order book as of September 30, 2026, plus Rs 652.5 crore of L1 (lowest-bidder) positions. These stocks fell mainly because they had run far ahead of earnings and because they trade on Nasdaq's mood.
Is AI a bubble? Facts on both sides
The honest answer is "both, in different places." Here is the evidence each side points to.
The bubble case | The reality case |
|---|---|
Debt is rising fast. The five biggest hyperscalers are expected to spend about $690 billion on capex in 2026 and $870 billion in 2027, up from about $480 billion expected at the start of the year. More of it is now borrowed. | The revenue is real. Nvidia earned $65.3 billion of net income last quarter on $115.8 billion of revenue. Dot-com leaders in 2000 mostly had no profits at all. |
Hidden obligations. Hyperscalers carry about $822 billion of future lease commitments not yet on their balance sheets (PIMCO). | Capacity is full. Northern Virginia, the world's largest data-centre hub, had a 0.2% vacancy rate in H1 2026 despite record construction (CBRE). |
Circular money. Chipmakers invest in AI labs that then buy their chips. Nvidia is among seven partners buying $26 billion of Anthropic stock at the IPO price. | Spending is growing. IoT Analytics estimates generative and agentic AI spending at $308 billion in 2025, forecast to reach $700 billion in 2026. |
Fast depreciation. AI chips age quickly. Gartner estimates more than $3.5 trillion of infrastructure spending could be at risk of depreciation by 2030. | Sceptics are converting. A CFA Institute blog author who set out to prove AI was a bubble said the evidence made him more bullish. |
Political risk. Ohio and Illinois paused data-centre tax credits in 2026, a new Democratic Congress may slow approvals, and Trump now threatens to tax overseas builds. | Big IPOs still clear. Anthropic priced above its range, with about six times more orders than shares on offer. |
The verdict
Think of it in two layers.
- The core (Nvidia, hyperscalers, top AI labs): real profits and real demand. Valuations can still fall 20-30%, but this is not a dot-com-style zero.
- The edges (debt-funded neoclouds, small "AI story" stocks, Indian proxies that tripled on narrative): this is where bubble pricing lived and where most of the crash is happening.
For scale: after the dot-com bust, the Nasdaq lost over 75% of its value by late 2002. On November 20 the Nasdaq 100 briefly touched a 20% drop from its high. That is a serious correction, but still far from a dot-com-style collapse. What would turn it into a burst is a real cut in hyperscaler capex. That has not happened yet.
Future outlook for AI and data centre stocks
Global: what decides the next move
- Fed decision in mid-December. Markets expect another quarter-point hike. Higher rates hurt debt-funded builders most.
- Nvidia's CEO search. Strong numbers, but no named successor. A credible permanent CEO would remove a big overhang.
- Hyperscaler Q4 earnings (late Jan to Feb 2027). Capex guidance for 2027 is the single most important number for the whole AI trade.
- Chip export rule. Public comments on the new rule run until about mid-January 2027. Any softening or widening will move chip stocks.
- Anthropic's trading and a possible OpenAI IPO. They test how much appetite is left for AI at trillion-dollar valuations.
- Details of the overseas data-centre tax. Its size, legal basis and exemptions will decide how much it really hurts cloud giants' global plans, and India's pipeline.
Scenario (our view) | What happens | Likely market effect |
|---|---|---|
Bull | Capex guidance holds near $870 billion for 2027, Nvidia names a CEO, Fed signals a pause | Chips and neoclouds rebound sharply from oversold levels |
Base | Spending holds but rates stay high | Choppy range. Leaders recover, weak balance sheets keep lagging |
Bear | A hyperscaler cuts 2027 capex or a neocloud struggles to refinance | Edge stocks fall further; this is when "correction" becomes "burst" |
India: the long-term data centre story is intact
India's data-centre build-out runs on its own drivers: data localisation rules, cloud adoption and AI demand from Indian firms.
- Installed capacity is projected to grow from 1.5 GW in 2025 to 6.5 GW by 2030, backed by a pipeline of about $90 billion.
- India is on track to reach about 2 GW of operating capacity by the end of 2026 (CBRE). About $173 billion of investment commitments have been announced since 2021.
- India generates nearly 20% of the world's data but has only 3-4% of global data-centre capacity. That gap is the long-term opportunity.
- Big-ticket projects include Google's planned AI hub in Visakhapatnam (about $15 billion over 2026-2030) and TCS's $6.5 billion plan for a 1 GW network.
- The new US chip curbs name Singapore, Malaysia, Thailand, Indonesia and Brazil, not India. Some of that demand could move here over time, though this is a possibility, not a given.
The risks for India: power and water are the bottlenecks, and data-centre electricity demand is projected to rise to 13.56 GW by 2031-32. Trump's overseas data-centre tax, if it becomes law, could slow US cloud giants' projects here. And in the short run, Indian AI stocks will keep trading on Nasdaq's mood, whatever their own order books say.
How to think about it as an investor
This is education, not a buy or sell call. Use these questions to judge any AI stock you hold or are watching:
- Is it core or edge? Does the company earn real profits from AI today, or does its price rest on a story about the future?
- How much debt? Companies that borrow to build data centres suffer most when rates rise. Check debt-to-equity and interest cover.
- What is the order book? For Indian names like Netweb, a growing order book is evidence. A rising share price alone is not.
- How much did it run up? A stock that tripled in a year can fall 40% with the business unchanged.
- Size your positions. These stocks routinely move 5-20% in a day. Use stop-losses and don't over-leverage, especially in F&O.
A correction separates companies with real earnings from companies that only had a good story. The next few months will show which is which.
Sources
- Nvidia Q3 FY27 results (Stock Analysis)
- Jensen Huang resigns (Bernama/Anadolu)
- Nov 5 Nasdaq selloff (Political.org)
- Trump on overseas data-centre tax (CNBC)
- Trump threatens tax on overseas data centres (Axios)
- Nov 20 market close (investingLive)
- US AI chip export rule (Techlicious)
- Indian AI stocks fall on Nov 6 (Kotak Neo)
- ESDS and E2E prices (Inc42)
- AI credit expansion (PIMCO)
- Indian data centres (ICRA)
- India Data Centre Market Monitor H1 2026 (CBRE)
Disclaimer: This article is for education only and is not investment advice or a recommendation to buy or sell any security. Prices are as reported on the dates shown and change quickly. Do your own research or consult a SEBI-registered adviser before investing.
Questions and answers
Is the AI bubble bursting in 2026?
Not yet. The Nasdaq 100 is close to 20% below its high and smaller AI stocks have fallen much more, but AI revenue is still growing fast. A real burst would need Big Tech to cut data centre capex, which has not happened.
Why did AI stocks fall when Nvidia's results were so strong?
Prices had already assumed great results. The selloff was driven by other things: the US midterm result, Nvidia's CEO exit, new chip export curbs, Trump's threatened tax on overseas data centres, and rising interest rates that make debt-funded data centres costlier.
Why are Indian AI and data centre stocks falling with the US?
Global funds treat names like E2E Networks, Netweb, ESDS and Anant Raj as India's version of the AI trade. When the Nasdaq sells off, they get sold too, regardless of their own order books.
Will Trump's overseas data centre tax hurt Indian stocks?
Possibly, if it becomes law. Indian data centre developers earn most of their revenue from US cloud giants, so a tax on building abroad could slow new projects here. For now it is only a statement with no published details.
Which AI stocks are most at risk?
Companies with high debt, little profit and big recent run-ups, such as debt-funded neoclouds and small caps priced on AI stories. Cash-rich leaders with real earnings have fallen less and tend to recover first.
Is this a good time to buy the AI dip?
That depends on your time horizon and risk appetite, and this article cannot tell you that. A safer approach is to separate core from edge names, buy in stages rather than all at once, and keep position sizes small while volatility is high.
What should investors watch next?
The Fed's December decision, Nvidia's CEO appointment, details of the overseas data centre tax, hyperscaler capex guidance in early 2027, and FII flows into Indian data centre names.
This page is education, not advice. It explains how things work. It does not recommend any security or any action.
- 16Defence Electronics: India's Next Big Theme? Future Outlook to 2030Record defence production, a 22% jump in capex and exports up 63% have put radars, jammers and air defence electronics at the centre of India's defence story. Here is a simple guide to the outlook.
- 01How stock tip scams on messaging apps trap tradersFree groups, screenshots of profits, a "VIP" plan and an app that will not let you withdraw. Here is how the trap works, step by step, and how to check anyone before you pay.
Saras is coming soon. Join the waitlist and we will let you know when it opens.