On 5 September 2026, one tracker recorded the grey market premium on the NSE issue at ₹310. Nineteen days later the shares listed at ₹1,800 against an issue price of ₹1,785. A gain of 0.84%.
Had that September premium held, the listing would have been near ₹2,095, a gain of about 17%. The number was not slightly off. It was off by a factor of twenty.
So what is GMP in IPO terms actually telling you? Quite a lot about direction, as it turns out, and very little about size. The NSE issue is the cleanest test of that available right now, because the whole sequence is dated, public and finished.
What the number is, and how it is built
The grey market is an informal market where IPO applications and allotted shares change hands before they list. It runs on dealer networks, mostly over the phone and on messaging apps. No exchange sits in the middle, and it operates outside SEBI's regulatory framework.
The grey market premium is simply what those dealers are paying above the issue price. The arithmetic every tracker uses is one line.
Issue price plus GMP equals the indicative listing price.
So a ₹1,785 issue price with a GMP of ₹310 implies a listing near ₹2,095. That is the whole calculation. No model sits behind it, no earnings input, no cash flow. It is a sentiment reading, quoted in rupees.
Two related quotes sit alongside it. Kostak is a fixed price paid for a whole application, allotted or not. Subject to sauda only stands if allotment comes through. Both are the same kind of informal arrangement.
One thing follows from all of this and is worth saying plainly. A grey market trade has no exchange, no clearing corporation and no regulator behind it. If the other side walks away, there is nobody to complain to.
The NSE issue, from ₹310 to 0.84%
The ₹22,561.57 crore NSE issue was priced in a band of ₹1,700 to ₹1,785, with a lot of 8 shares. Bidding ran from 17 to 21 September 2026 and the stock listed on BSE on 24 September.
Here is what the premium did across that window, as recorded by trackers that publish dated histories.
Read the shape of that rather than the rows. The premium was highest when the least was known, twelve days before bidding opened. It fell as real information arrived. By listing morning it had given up about 87% of its peak.
The issue itself was not weak. It was subscribed 5.71 times. The grey market was not wrong about direction, either. It said the stock would list above ₹1,785, and it did.
It was wrong about how much, every single day, by a widening margin the further back you go.
Direction, yes. Magnitude, no.
This is not a one-off, and there is data on it.
One tracker studied 90 IPOs over twelve months. It found GMP called the direction of the listing correctly in about 83% of cases. The average gap between the GMP-implied price and the actual listing price was around ₹25. A separate 2026 study of mainboard issues put directional accuracy at 72.73% where a clear grey market signal existed.
Now look back at the NSE table. The final morning's GMP of ₹40 implied ₹1,825 against an actual ₹1,800. A gap of ₹25.
That is less a coincidence than a description of the instrument. On the last day, with subscription data published and the anchor book known, GMP lands close. Weeks earlier, with none of that available, it is a mood.
The same pattern shows up in issues that missed entirely. Scoda Tubes listed at ₹140 against grey market estimates nearer ₹160. NR Vandana Textile debuted flat at ₹45 despite a solid subscription, also below what the grey market had indicated.
If you want a sense of demand before listing, the subscription figures and the anchor investors in IPO allocations are published, dated and verifiable. GMP is none of those things.
There is no single GMP, and today proves it
With no exchange, there is no one price. There are only quotes, and they disagree.
Take the NSE issue. InvestorGain records the peak premium as ₹310 on 5 September. IPO Watch records the high as ₹285 on 7 September. Same issue, same fortnight, two different peaks on two different dates.
The same thing is visible on any live issue. On 24 September 2026, trackers were quoting the A-One Steels premium anywhere between ₹34 and ₹62, a spread of nearly twice the lower figure, on the same day.
There is a quieter point in the data too. One tracker noted on 27 September that of 42 mainboard and SME issues it follows, only 17 had any recorded premium at all. For roughly six in ten issues, the number people go looking for does not exist.
So when a headline says "GMP today is X", the honest reading is that one dealer network was quoted at X, by one tracker, at one moment. It is a data point with a source, not a market price.
Conclusion
GMP is not useless. It is a real signal about how a small, informed, unregulated group feels on a given day, and on the last day before listing it has been reasonably close.
What it cannot do is tell you what a business is worth. It is also at its least reliable exactly when people pay most attention to it, in the weeks before bidding opens.
The NSE sequence is worth keeping. A premium of ₹310 in early September, ₹48 when bidding closed, ₹40 on the morning of listing, and an actual gain of 0.84%. Every one of those numbers was reported as news at the time.
If you want something dated and verifiable to work from instead, the subscription figures and the IPO prospectus are both public before you have to decide. Start with what is an IPO if the process itself is new, and how to analyse an IPO before investing for the part GMP was never built to answer.
