The Nifty rose 2% and India's AI stocks fell 20%. Here's what actually happened on 6 November.
A US election started it, and memory chips carried it to India. The stock that fell hardest wasn't one that builds AI hardware. It was a laptop refurbisher, three days after its best quarter
Last Friday, 6 November, the Nifty closed 2% higher at 25,030, its first close above 25,000 since the end of February. Over the week it gained 8.6%.
It was a very bad day to own Indian AI stocks. GNG Electronics, which sells refurbished laptops, touched its 20% lower circuit and closed down 19.9%. Aeroflex fell 10.7%. Netweb fell 6.3%. E2E Networks, ESDS, Sterlite Technologies, HFCL and KRN Heat Exchanger all hit their lower circuits.
The GNG fall is the strange one. Three days earlier, on 3 November, it had reported the best quarter since its listing, with revenue up 42% and profit up 72%, and it raised its guidance for the year. Nothing in that report changed on Friday morning.
What changed was a story about America, and it reached India through memory chips. Once you see how that chain works, you can tell which of Friday’s falls came from the theme and which came from somewhere else. Not all of them came from the theme.
It started with a US election
The US held its midterm elections on 3 November, and the Democrats won control of both houses of Congress. AI data centres were a real issue in this election. They use a lot of electricity and water, many communities have pushed back, and a large number of the winning candidates had promised to slow or pause them.
The clearest case is Texas, the biggest data-centre market in the US. Gina Hinojosa won the governor’s race there. The campaign promise, in Hinojosa’s own words, was “a moratorium, a ban, on all data center approvals until laws can be passed by the Legislature”. The new governor takes office on 19 January. Local voters also decided 41 ballot measures on data centres, 21 of them in Ohio, and most passed.
On Thursday 5 November the US market reacted. The Nasdaq 100 fell 6.3% and the S&P 500 fell 4.4%. The Philadelphia semiconductor index, which tracks chip companies, fell 12.3%, its worst day since March 2020.
The detail that matters for India is which chip stocks fell most. Nvidia, the company most people think of as “the AI stock”, fell 9.2%. The memory makers fell twice as hard: Micron about 17% and SanDisk 22%. In Korea on Friday, the Kospi fell 8% at one point and tripped its circuit breaker before closing 6% lower, and SK Hynix, one of the world’s largest memory makers, fell more than 10%.
It’s worth being precise about what has and hasn’t happened. No law has been passed. Goldman Sachs reportedly estimates that the states likely to tighten their rules hold about 45% of planned US data-centre capacity, with Texas alone about a quarter, but it expects any pauses to be short-lived. Nomura, as quoted by Business Today, thinks a federal moratorium is unlikely and a temporary state-level pause is possible. So what the market priced on Thursday was a fear, not an event.
Why memory is the link
To see why the memory makers fell most, and why that hit a laptop refurbisher in India, it helps to follow the money.
An AI data centre is mostly servers, and every AI server needs a very large amount of memory. Over the past year the big cloud companies bought so much of it that memory became scarce for everyone else. Prices rose fast. By one report, DRAM contract prices rose about 80% in the April to June quarter and about 40% in July to September. Memory makers’ profits rose with them, which is exactly why they were the most exposed when investors started to worry that the buying might slow.
Laptops use the same memory. When the AI buyers take first place in the queue, laptop makers pay more and wait longer, and new laptops get more expensive. That is where GNG comes in.
GNG, the company that the shortage was helping
GNG Electronics buys used laptops, desktops and phones from large companies and banks when they replace their devices. It refurbishes them and sells them with a warranty under the Electronics Bazaar brand, in 55 countries. It refurbished about 7.3 lakh devices in FY26 and listed in July 2025 at an IPO price of ₹237.
Its September quarter (Q2 FY27), from the company’s own investor deck:
- Revenue ₹624.8 crore, up 42% from ₹439.9 crore a year ago. India grew 45% and international about 40%.
- Gross margin 25.1%, up from 19.9%.
- EBITDA, as the company reports it, ₹93.9 crore, roughly double the ₹46.8 crore of a year ago. The EBITDA margin went from 10.6% to 15.0%.
- Profit after tax ₹56.4 crore, up 72% from ₹32.7 crore.
The business is seasonal: the March quarter is the strongest and the June quarter the weakest, so each quarter should be compared with the same quarter a year earlier, as above.
On the call that evening, the managing director, Sharad Khandelwal, spent the opening remarks on exactly the chain above, starting with IDC’s figures: global PC shipments fell 20.1% in the September quarter, from 78.5 million to 62.7 million units, and IDC expects an 11.3% fall for 2026 as a whole. Then came an example. A year ago a typical laptop in India cost ₹35,000 to ₹40,000. The same laptop now costs ₹65,000 to ₹72,000. GNG prices a refurbished device at about 30% of a new one, so in rupees, the customer’s saving has almost doubled.
That explains the demand. The margin had a second source. Management said 125 to 150 basis points of the margin gain came from inventory it bought before prices rose. The rest came from better pricing. It has also been buying a lot more: inventory went from ₹743 crore in March to ₹912 crore in September, and borrowings from ₹406 crore to ₹611 crore. Net of cash, debt is about ₹427 crore, and interest cost for the quarter was ₹22.0 crore against ₹8.4 crore a year ago. Management’s reasoning is that companies are extending laptop leases and delaying replacements because new devices are so expensive, so fewer used devices come to market, and holding stock is the better choice “for at least the next two quarters”.
To its credit, the company doesn’t hide how much of the current quarter is the shortage. The investor deck says rising memory prices “provide a near-term catalyst”, while “the underlying opportunity is driven by a structural shift towards affordability, reliability and sustainability”. It raised its FY27 revenue growth guidance from 30% to almost 40%, and its margin guidance from 100 to 150 basis points above last year. For FY28 it said it would “hold back on the guidance for now”.
The question an analyst asked three days early
Towards the end of the 3 November call, an analyst asked the question the market asked on Friday. Paraphrasing only slightly, the question was what would happen if “the AI infra pulls out” and “the memory prices probably come crashing down”.
Management gave three answers. First, before the shortage, GNG had grown “north of 30%” a year for three years. Second, it sees the price rise as “a permanent reset”, and if it needs to, it can move its own price from about 30% of a new device’s price towards 35 to 40%. Third, it called a crash in memory prices an “unlikely scenario” that it doesn’t “believe for a moment will happen”, and said that buying ahead of the curve gives it “a significant cushion”.
Whether that answer is right will only show up in later quarters. But on Friday morning the market chose not to wait.
Why the stock fell 20% anyway
This is my reading of it, in order of weight. Management didn’t say any of this.
1. The quarter was explained by the shortage, so the stock traded as a bet on the shortage. When memory stocks fell 17 to 22% in the US, GNG was the closest thing India had to a listed memory-price story. That’s a strange position for a laptop refurbisher, but it’s the one its own results put it in.
2. Inventory and debt make the reaction sharper. If used-device prices soften, stock bought at today’s prices earns less, and the interest bill keeps running. I’d put that next to management’s point about buying ahead, which works in its favour only if prices don’t fall before the stock is sold.
Who else fell, and why it wasn’t all one story
Here are the official NSE closes for 6 November:
- GNG Electronics −19.9%
- Thermax −17.3%
- Rashi Peripherals −14.8%
- TD Power Systems −13.9%, after −8.1% the day before
- Aeroflex −10.7%
- Netweb −6.3%
- Black Box −5.4%
- E2E Networks, ESDS, Sterlite Technologies, HFCL and KRN Heat Exchanger −5.0%, the lower circuit for each
- For comparison: Kaynes −1.8%, Nifty IT +0.43%, Nifty 50 +2.01%
Two things stand out before anything else. The Nifty IT index, which is Infosys, TCS and the services companies, rose. So the fear was about AI hardware spending, not about technology in general. And the companies that actually build AI systems fell less than GNG did. Netweb, which sells AI servers, had just reported revenue up 336% to ₹1,323 crore, with ₹773 crore of it from AI systems. Sterlite, which sells fibre to the cloud companies, reported revenue up 145% and an order book of ₹35,274 crore, including two hyperscaler contracts of $1.2 billion and $1.11 billion.
Then there’s the part that most “AI crash” lists got wrong. At least two of the biggest falls had their own news.
Thermax reported its Q2 results on the evening of 5 November, a few hours before the fall. Revenue rose 11.4%, but profit after tax fell 4.3%. A 17% fall the next morning is mostly a reaction to that, whatever the lists say.
Rashi Peripherals, which distributes computers and components, also reported on the evening of 5 November. Its numbers were strong: revenue up 48.8% and profit up 114.5%. But at the same meeting, its board approved raising up to ₹900 crore. New shares dilute existing shareholders, and markets often react to that. So Rashi’s fall was partly the theme and partly its own news, and I can’t separate the two.
Two more things change how Friday’s list should be read. First, not every lower circuit is the same size. E2E, Sterlite and HFCL trade with 5% limits, so their lower circuit is a quarter of the fall GNG’s band allows. Second, several of these stocks were falling long before the election. From 14 October to 6 November, E2E fell about 35% and Netweb about 30%. KRN fell 5% on each of the three sessions before the election, after its own results. For them Friday was one more step, not the start.
Both sides of the AI slowdown fear
The case that the fear is overdone. Memory supply can’t be added quickly, because a new fab takes years. TrendForce’s latest official forecast, from 30 September, has conventional DRAM contract prices rising another 10 to 15% in the December quarter, and it expects PC memory supply could fall in 2027 as more capacity moves to servers. That’s the line that matters most for GNG. The policy threat is narrower than the headlines: Goldman expects pauses to be short-lived, and Nomura sees no federal ban. US memory stocks bounced 8 to 10% on Friday morning, after Indian markets had closed. And the Indian companies that fell had just reported strong quarters. GNG says it grew more than 30% a year before any of this started.
The case that the market is right to worry. Memory prices are still rising, but more slowly each quarter: about 80%, then 40%, then an expected 10 to 15%. Spot prices, which usually turn before contract prices, have stopped rising. In the week to 3 November the main DDR4 spot price slipped 0.09% and NAND wafer prices fell 4.33%, and TrendForce said buyers were waiting for a correction. GNG’s margin includes 125 to 150 basis points from cheaper old inventory that will run out, and it is carrying more stock, more debt and more than double last year’s interest cost. Netweb’s growth is being funded by inventory too: its operating cash flow for the first half was negative ₹639 crore. If orders slow, that inventory becomes the problem.
My research desk has detailed notes on the electronics makers and on the AI data-centre suppliers, which cover most of the companies above, and a new note on data-centre power and cooling follows on 16 November. Seven notes are free to read at dhruvmadiaresearch.com/desk/free.
This is not investment advice. I am not a SEBI-registered adviser. Do your own work before acting on anything here.
10 min read