Export view - By Srishti Mendiratta

 

UPI's Zero-MDR Era May Be Nearing a Turning Point

Merchant Discount Rate, or MDR, is the fee a merchant pays each time a customer pays digitally, whether by card or through a payment app. Banks and payment companies charge it to cover the cost of running the systems that move that money instantly and securely. Debit and credit cards have always carried this fee. UPI never has. Since January 2020, the government made UPI transactions free for merchants, and that decision is a big part of why India's real time payments system grew into the largest in the world by transaction volume.

That zero MDR era may be nearing a turning point. The Lok Sabha recently passed the Taxation and Other Laws Amendment Bill, 2026, creating a legal framework that would allow the government to decide which digital payment methods remain exempt from Merchant Discount Rate (MDR). Once the Bill becomes law, the government can notify a negative list of payment modes that will continue to remain MDR-free, while payment modes outside that list could become eligible for MDR if charges are notified. The Finance Minister has already clarified that any such fee would apply only to merchants and not to end users and that no final decision has been taken yet.

Media reports citing government sources suggest the levy under discussion could fall between 0.25% and 0.4%, likely applied to merchant transactions above ₹2,000. This range hasn't been officially confirmed by the RBI or the Finance Ministry, so it should be read as a reported estimate rather than a settled number, but it gives a useful sense of scale for what's on the table.

Banks have effectively run UPI's merchant side as a cost centre since 2020, absorbing infrastructure and processing expenses without any fee to offset them, meaning every merchant transaction processed has added to their costs without adding to their revenue. Even the lower end of the reported range, 0.25% on transactions above ₹2,000, could generate roughly ₹17,416 crore a year across the sector, going by recent monthly transaction data. At the upper end of 0.4%, that figure would scale to somewhere in the region of ₹27,800 crore. For banks with a sizeable digital payments book, this would turn UPI from a volume heavy, margin light business into one with a real fee income component attached.

Payment aggregators and fintech platforms, the app layer merchants actually transact through, are in much the same boat. An MDR in this range would let them start recovering costs they've carried for years, though how much of that benefit actually reaches them versus banks depends on how any fee eventually gets split and that detail hasn't been worked out yet. If anything, this is the segment most exposed to the outcome, since UPI volumes sit at the core of these platforms' business models in a way they don't for larger, more diversified banks.

Merchants sit on the other side of this. One reported model would apply MDR only to transactions above ₹2,000 made to businesses with annual turnover exceeding ₹1.5 crore, meaning smaller merchants below that threshold could stay exempt altogether. If that structure holds, the real burden falls on larger, high-turnover businesses rather than small shopkeepers and street vendors. Industry bodies have raised a related concern, though: a turnover cutoff draws a hard line where the underlying economics are actually quite similar on both sides of it. A business just above ₹1.5 crore in turnover isn't necessarily better cushioned or more profitable than one just below it. Officials have said the fee itself would likely be small, but for businesses already running on thin margins, even a small new recurring cost changes the math in a way it simply wouldn't for a bigger retailer with more financial cushion.

For retail investors, the point isn't that MDR is coming, it's that this has stopped being a purely hypothetical debate. A reported rate range now exists, even if unconfirmed, giving banks and payment platforms a genuinely plausible path to new fee income, while leaving merchants to absorb the other side of it. Nothing is finalized, and the steering committee's decision is still pending, but the range under discussion explains, in fairly concrete terms, why every part of this ecosystem has real skin in the game.

By Srishti Mendiratta | SEBI-Registered Research Analyst – INH000024295

https://wealthminds.co.in/

investor@wealthminds.co.in

Export view - By Srishti Mendiratta

 

The Limits of Artificial Intelligence and the Value of Human Judgment

Sit through enough earnings calls and you'll hear the same line, said in a dozen different ways: AI is making businesses faster. And it's true. Tasks that took weeks now take days. Teams that needed ten people now run on three. The efficiency gains are showing up in real results, not just in slide decks.

What matters more for an investor is a different question: how much of that speed can you actually trust, and where does someone still need to be watching over it? AI is genuinely good at work that's repetitive and data-heavy, the kind with clear rules. But give it a task that depends on trust, context, or judgment and the risk doesn't go away. It just gets harder to see. Look closely across sectors and the same pattern keeps showing up. The businesses building lasting returns from AI aren't always the ones moving fastest. They're the ones who've figured out where to stop and let a person take over.

Cybersecurity is the clearest place to see this, because AI is helping both sides of the fight. India averaged 3,195 cyberattacks per organization every week in 2025, according to Check Point Software's 2026 report and much of that rise came from AI tools that let attackers scan networks and find weaknesses with barely any human help. Defenders had to respond in kind. Indian firms blocked over 9 billion attack attempts last year, up 27 percent from 2024, because past a certain point, only a machine can keep up with another machine.

Here's the part that should worry boards more, though. AI isn't just a tool that attackers pick up. It's starting to cause damage on its own. In July 2026, OpenAI admitted that one of its experimental models slipped out of a test environment on its own, with no human telling it to and reached a live production system belonging to Hugging Face. The model got in using stolen login details plus a security gap nobody knew about and Hugging Face's own CEO said he had never seen anything quite like it. This wasn't a one-off, either. OpenAI's own hacking test score had already jumped from 27 percent to 76 percent in just three months earlier that year. Put the two together and it's hard to treat AI safety as just an IT problem anymore. It belongs in the boardroom now.

HR shows a quieter but honestly more telling version of the same story. Two out of three Indian companies already use AI somewhere in HR, yet fewer than half have written any usage rules, and a quarter have no framework at all. Barely a third are seeing real productivity gains, even though most expect AI to be central to daily work within a few years. In simple terms, everyone adopted the tool before checking if it actually worked. That tracks, because hiring was never just a data problem. Judging character, fit, and honesty is hard even for a person. It's harder still for a system that has never met anyone.

Marketing runs into the same wall. Content made entirely by AI performs about four times worse than content where a person is genuinely involved in shaping it. Nearly three out of four Indian businesses got no real return from their AI content spend, mostly because they published it without anyone checking it first. The tool did its job fine. The problem was skipping the human review.

Real estate adds one more example. Proptech platforms can lift sales speed by 30 to 50 percent, according to an EY-Parthenon-CREDAI report. But buying a home still comes down to trust, negotiation, and small personal preferences that don't show up cleanly in any data. AI can narrow down the choices. It can't close the deal on its own, at least not yet.

Looking at the bigger picture, the lesson is fairly simple, even if the details change by sector. AI earns its place wherever the work is repetitive and speed matters most. It gets shakier the moment a decision needs judgment, accountability, or a real read on people. That's the signal worth watching if you're allocating capital. The first wave of AI adoption rewarded whoever cut costs the fastest. The next wave will likely reward something quieter: knowing exactly when to hand the decision back to a person.

That's the real test, not how much AI a company has adopted, but how honestly it has admitted what AI still can't do. The businesses that pass that test now are the ones likely to still be standing once the trial runs stop making headlines and the results start getting counted.

By Srishti Mendiratta | SEBI-Registered Research Analyst – INH000024295

https://wealthminds.co.in/

investor@wealthminds.co.in