GSTN's annual statistical report puts the active taxpayer base at 1.67 crore as on 30 June 2026, 8.9% higher than a year earlier.[1]
Those businesses paid ₹23.32 lakh crore in FY 2025-26, against ₹22.09 lakh crore the year before. A rise of 5.6%.
As on 30 June 2026. The two figures move at different speeds, and the gap is the first thing worth explaining.
Compensation cess stopped applying from the September 2025 return period, which is why the top band thins in the final year.
Source: GSTN, 9 Years of GST
Two numbers moving at different speeds. The base grew 8.9%, collections 5.6%. Why would collections lag a base that is widening?
The explanation sits in the rate structure rather than in demand. Compensation cess stopped applying from the September 2025 return period, and cess collections reduced 31.7% over the year, from ₹1.53 lakh crore to ₹1.05 lakh crore. Strip cess out and the remaining heads grew 8.35%, which tracks the taxpayer base closely. CGST rose 7.4%, SGST 6.7%, IGST 9.4%.
The rate rationalisation that removed cess also lowered rates across a wide base. So collections grew 5.6% in a year when much of what GST applies to became cheaper to tax, and the base it applies to widened. The headline figure understates what the system actually did.
Read against the previous year's report, some patterns in this one have moved, some have held almost exactly still, and a few run against what you would expect.
The rate structure was reworked. On the 56th GST Council's recommendation, the four-tier arrangement of 5, 12, 18 and 28% gave way to two main rates of 5 and 18%, with a special 40% rate for select goods and services replacing the earlier 28%-plus-cess treatment.[4] The 12 and 28% slabs were largely retired, moving consumer electronics, small cars, cement and air conditioners to 18%, and much packaged food and household goods to 5%.
On the administrative side, Rule 14A introduced a simplified registration granting approval within three working days subject to Aadhaar authentication, with Form REG-32 available to opt back out of it. A three-year limit now closes filing for GSTR-1, 3B, 9 and 8 once that period has passed from the due date. The Invoice Management System was extended, with the pending option opened to credit notes, a remarks facility added, and Bill of Entry data from ICEGATE populated into it directly.
These are a few picked from a longer list, chosen because they bear on what follows.
Active registrations stood at 1.67 crore on 30 June 2026. Within that, regular taxpayers numbered 1.49 crore, up 10.5% over the year, while composition registrations reduced 8.4%, from 14.80 lakh to 13.55 lakh.[1][2]
As shares of the base, regular registrations moved from 87.6% to 89.0%, and composition from 9.6% to 8.1%.
Source: GSTN, 9 Years of GST
This is not a one-year movement. Composition returns have declined almost every quarter since April 2020, from 13.97 lakh in the quarter to June 2020 to 8.73 lakh in the quarter to March 2026.
Down 37.5% from the quarter to June 2020 to the quarter to March 2026, through a period when the overall taxpayer base kept growing.
Source: GSTN, 9 Years of GST
The reports give the stock at two dates, not the flows between them, so the cause cannot be isolated. Three mechanisms are consistent with the movement, and they are not the same thing. Businesses already in the scheme may be moving to the regular one to claim input tax credit. Others may be crossing the turnover ceiling and leaving because they have to. And newly registered businesses eligible for composition may simply not be electing it.
The third would be the most consequential, if it is happening widely. Are businesses leaving the composition scheme, or has it stopped being chosen in the first place? The reports cannot tell us, and the two would mean quite different things.
What follows from any of the three is the same, and it is not a tax consequence. A composition dealer files CMP-08, a quarterly summary of turnover and tax. It files no GSTR-1 and reports no invoices. A business that moves to the regular scheme begins reporting line by line, with counterparties and values attached. It acquires a granular record where it previously had none, which makes it noticeable in a way it was not before.
By legal form, revenue is concentrated in companies. Public and private limited companies together account for 63% of cumulative cash collected since 2017, public limited at 33.5% and private limited at 29.4%. Proprietorships, the typical form of a small business, account for 13.4%. Public sector undertakings contribute 9.1%, partnerships 7.3%.[1]
Those figures accumulate from 2017 onward, and exclude IGST on imports.
Going deeper into a single year, where the report breaks taxpayers by turnover, the same pattern holds in sharper form, and turnover allows the MSME limits to be applied. Those limits were revised from 1 April 2025, so FY 2025-26 falls entirely under the new ones: micro up to ₹10 crore, small up to ₹100 crore, medium up to ₹500 crore.[3]
FY 2025-26, against the MSME limits revised from 1 April 2025. Micro businesses pay 16.8% of cash tax. The 16,194 registrations above the ₹500 crore medium ceiling pay 50.3%.
Businesses within the micro limit, up to ₹10 crore turnover, are 95.2% of taxpayers and pay 16.8% of cash tax. Everything above the medium ceiling of ₹500 crore is 16,194 registrations, roughly one in a thousand, and pays 50.3% of all cash GST.[1]
Put the other way: the entire MSME range, 99.9% of the filing population, pays a shade under half of India's cash GST. The other half comes from 16,194 businesses.
Four tiers, under the MSME limits in force for FY 2025-26. The three MSME tiers together pay 49.7% of cash GST; the 16,194 businesses above them pay the other half.
Source: GSTN, 9 Years of GST. Tiers follow the MSME turnover limits revised with effect from 1 April 2025. Counts are at GSTIN level, so this is the population within MSME turnover limits, not a count of registered MSMEs.
What makes the imbalance worth reporting is not that it exists but that it does not move. Between the two reports every tier paid more in rupees, and the largest shift in any tier's share of cash tax was 0.12 percentage points. The slab above ₹500 crore went from 50.30 to 50.29%.
The base has widened for nine years. The distribution of who funds GST has not noticed.
Filing discipline has improved on both returns. The gap between returns due and returns filed on the due date has narrowed steadily across the period for GSTR-1 and GSTR-3B alike.[1]
They are also agreeing with each other more often. GSTR-1 declares outward supplies invoice by invoice, GSTR-3B is the summary on which tax is paid, and when the two disagree on liability the system raises an automated notice. The number of those notices is a direct measure of how often the mismatch arises.
1,14,148 notices in the twelve months to May 2025, against 92,456 in the twelve months to May 2026, while the filer base grew about 9%.
Source: GSTN, 8 and 9 Years of GST. Both totals cover twelve months. May 2026 is left off the chart because its resolution rate was still incomplete at the reporting date, though its notices are counted in the total.
In the year covered by the previous report, 1,14,148 notices were issued. In the year covered by this one, 92,456, a reduction of 19% against a filer base that grew about 9%. Per filer, roughly a quarter fewer mismatches.[1][2]
That is what you would expect as accounting software and e-invoicing push more returns toward reconciling by construction rather than by effort.
One counter-current belongs alongside it. Where a notice is raised, the share resolved by the business correcting itself has slipped, from around 94% in mid-2024 to between 88 and 91% through 2026, so while fewer disagreements arise, the ones that do are proving marginally harder to close.
In March 2026, 1.13 crore businesses reported 43.91 crore invoices through GSTR-1.[1]
Log scale. From 0.6 invoices a month at the smallest slab to 6,883 at the largest, under one set of filing rules.
Source: GSTN, 9 Years of GST. FY 2025-26, at GSTIN level.
They are not reporting at anything like the same rate. Businesses above ₹500 crore turnover average 6,882.7 invoices a month each. Those under ₹5 lakh, excluding nil filers, average 0.6. Across everyone, averaged over the three months the report tabulates, it is 34.6. Four orders of magnitude, under one set of rules, with the same return, deadline and penalty applying across the range.
Mapping the slabs to the MSME limits puts the volume in a different light, and the mapping is exact: the revised ceilings of ₹10 crore, ₹100 crore and ₹500 crore each fall on a slab boundary the report already uses.[3]
Businesses within MSME turnover limits are 99.9% of filers and generate 70.5% of all invoices, while paying 49.7% of cash tax. Close to 31 crore invoices in March 2026 alone.
One qualifier. MSME status is a composite of investment and turnover, assessed at PAN level, while these tables count at GSTIN level. A large company's branch registration in one state can sit in a low slab. So this is the population within MSME turnover limits, not a count of registered MSMEs.
A further 39.5 lakh of those filers report turnover below ₹20 lakh, which places them under both the ₹20 lakh services threshold and the ₹40 lakh goods threshold. Just under 35% of everyone filing sits below the point at which registration becomes compulsory on turnover alone, and that group grew 6% over the year. Not all of them had the choice: inter-state suppliers of goods, e-commerce sellers and those liable under reverse charge must register whatever they turn over, and special category states set lower thresholds. For most of the group, though, registration was a commercial decision. The lowest band excludes nil filers, so these are businesses actually filing, not dormant registrations. The reason is ordinary enough: a registered supplier can issue an invoice its buyer claims credit on, and an unregistered one cannot, so staying in a supply chain means staying registered.
Taken with the composition shift, the invoice-level base is filling from both directions at once, from above as businesses leave the simplified scheme and from below as businesses register without having to.
What all of it produces is a verified receivable. Every one of those invoices carries a named buyer, a value and a date, confirmed in the buyer's own filings. That is the asset supply chain finance and TReDS are built to discount, and seven in ten of the country's invoices now originate below the ₹500 crore line.
Seven in ten of the country’s invoices now come from businesses inside the MSME turnover range. TReDS is the rail built to turn those invoices into working capital.
E-invoicing requires an invoice to be registered with any authorised Invoice Registration Portal, and assigned a reference number, before it is issued.
The mandate covers B2B supplies and exports. Sales to consumers sit outside it. The report confirms this by omission, listing B2B, deemed exports, exports with and without payment, and SEZ supplies, with no B2C line because no B2C invoice is reported. B2B alone is 87.5% of e-invoices generated.
The two lines hold about 15 percentage points apart, every month. Invoices inside the e-invoice net are much larger than those outside it.
Source: GSTN, 9 Years of GST
Across the year to June 2026, e-invoices were 64 to 75% of GSTR-1 invoices by count, and 80 to 91% by value.[1] The distance between those two lines held near 15 percentage points every month.
Invoices inside the net are much larger than those outside it. About a third of reported invoice count sits outside the strictest layer of reporting, against roughly a sixth of value. That is the signature of consumer-facing sales and sub-threshold suppliers, many transactions each of them small.
It leaves a question the report does not raise. Multiple registration portals now absorb national volume between them, clearing 25 to 28 crore e-invoices a month, and GSTR filing has stabilised around them.
So is the system ready for a B2C orbit? The capacity question looks different now than it did in 2020. What has not been answered is whether it would be worth the compliance burden on small retailers, or the breadth of consumer transaction detail it would place on record. This data settles neither.
Through the year to June 2026, between 9.3 and 10.0 lakh businesses generated e-invoices each month. Between 82 and 88 lakh received them.[1]
Jul 2025 to Jun 2026. The ratio held near 1:9 for eleven of the twelve months, so the verified record reaches far beyond the firms that create it.
The ratio held near 1:9 for eleven of those twelve months. Every e-invoice has a counterparty, and the receiving side is an order of magnitude wider. A business well below any e-invoicing threshold still accumulates a verified, machine-readable record of its purchases, because its suppliers are above it.
That is the link to earlier work here. The Invoice Trail looked at how such a record becomes an asset a small business can borrow against. The 1:9 ratio is why the argument reaches past the firms generating the documents.
Of goods movements recorded on e-way bills, 98.73% travel by road, against 0.56% by rail, 0.65% by air and 0.06% by ship. The previous year's report shows 98.71%, so the split is structural.[1][2] E-way bill generation ran to 1,55,96,08,676 in FY 2025-26.
There is a possibility here that the data cannot establish. A tax system in which virtually all goods move by road shares a physical network with infrastructure that already records movement. A FASTag read places a vehicle at a plaza at a given moment. An e-way bill declares a consignment travelling between two points within a window. Should two independent records of the same journey be read against each other? The question may be a better one for fintechs and solution providers than for an administration. A consignment that can be shown to have moved is a different credit proposition from one that has only been invoiced.
Nine years in, 1.67 crore businesses file into a single system, and seven in ten of the invoices flowing through it come from firms whose turnover falls inside the MSME range. Nearly 40 lakh of them sit below the threshold at which any of this becomes compulsory, and most register anyway, because their buyers need an invoice that carries credit.
That is formalisation arriving through commercial necessity rather than enforcement, and it has a consequence the tax system was not designed to deliver. A business that files invoices has a verifiable record of what it sold, to whom, and when. It can be assessed by a lender who has never met it. For a small firm in India, where the absence of exactly that record has been the binding constraint on credit for decades, this is the more durable outcome than any collection figure.
The direction is already set for the years ahead. At its 57th meeting this October, the GST Council recommended a further round of process reforms: registration amendments accepted automatically, cancellation moving to a system-driven route, refunds of excess balances released without an officer seeing them, and new mechanisms to align what a business declares with what its suppliers reported. Much of it applies from the April 2027 return period.[4] Each is administrative taken alone. Together they continue the same movement, which is the system doing more of the checking itself, on the strength of detail the businesses have already supplied.
The edge of all this is where consumer sales begin. A third of invoices by count stay outside the strictest layer, and that share has not narrowed in a year. Whether it should is the question the next several years will answer.