Unidentified Information

The export finished at two in the morning. The bill turned up three weeks later, and the line that stung was not storage or compute. It was the charge for moving your own files off the platform. That fee is what the EU Data Act cloud switching rules were built to abolish, and there is finally a date on it.

The date is 12 January 2027. From that day a provider covered by the Act cannot bill you for leaving, and the European Commission says so in its own Data Act explainer. Everything before that day sits in a halfway house where the charge is still legal, but only at cost, and only for costs the provider can actually point to.

Why does leaving a cloud provider still cost money?

Because the meter runs on data leaving the building rather than data sitting in it, storage looks cheap right up to the moment you try to move out, and the exit charge is the rent nobody quoted you. Providers price ingress at zero and egress by the gigabyte, which is a perfectly rational way to make a bucket sticky.

That asymmetry is the whole game. You are never charged to hand over your data, only to take it back, and the bill scales with how well the platform has served you. The bigger your archive grew, the more it costs to stop trusting the company that holds it. This is the same arithmetic that decides renting cloud storage against owning a NAS at home, except at company scale it arrives as one invoice.

The conventional advice, that you should architect for portability from day one, is mostly a way of spending money now to avoid spending it later. Sometimes that pays. Often the abstraction layer costs more than the exit fee ever would, and the cleaner move is to accept the lock-in and know the price of the door. Anyone who has tried to back up an iPhone straight to an external drive knows the shape of this problem already: the data is yours, the route out is the vendor's.

How much does it cost to move 2 TB out of AWS today?

EgressCost checked AWS list pricing for US East transfer to the internet in September 2026, and that rate card decides most of these arguments. It is the figure finance will see. Four numbers set the calculation.

Work the archive figure through yourself, because no invoice shows you the sum in advance: 2 TB is 2,048 GB, the first 100 of them ride free, and the remaining 1,948 at the first-tier rate land you just under the price of a mid-range phone. That is our arithmetic on a published rate, not a quoted number. The notice ceiling matters just as much, because it is the one clause a provider cannot draft around, and it means no exit conversation can legally be stretched past a quarter.

EU Data Act cloud switching: wait or move now?

Wait if the move can slip a quarter without costing you anything, and move now if a contract renewal or a price rise is already forcing the issue. The table below splits the two windows apart.

Two things change on that date and one thing does not, which is why a straight before-and-after reading is the only honest way to look at it. The switching mechanics below follow the Data Act as read by Pinsent Masons, and the prices are AWS's own.

Read the break-even row slowly, because you compute that one yourself. Divide your expected exit bill by what the old provider costs you each month, and you get the number of months of double-running that the wait can absorb before it stops paying. A shop paying $300 a month for a 10 TB estate can afford roughly three months of overlap. A shop paying $3,000 cannot afford one.

Does the EU Data Act apply to SaaS?

Yes, a SaaS subscription is in scope wherever it meets the Act's definition of a data processing service, which pulls your project tracker and your accounting tool into the same switching regime as a raw storage bucket. Greenberg Traurig set this out in September 2025.

That is the part most coverage misses, because the story gets written as a hyperscaler story. The small tools are where lock-in actually hurts, since a team of nine can leave a storage provider in an afternoon and still be trapped in a task tracker holding six years of comment history. On-premise systems and private cloud sit outside, along with the heavily individualised builds that were never really a product.

Friction points before you pull the plug

Plenty still bites after the free-exit date arrives, starting with the charges the law deliberately leaves standing, the export formats nobody will put in writing, and a transition clock your provider is allowed to stretch when the work is genuinely hard.

The exemption that will catch most teams is the multi-cloud one. If you keep both providers alive and shuttle data between them, that traffic is ordinary usage and it bills like ordinary usage, whatever the calendar says. Only the act of leaving goes free. Or rather, only the act of leaving completely. Plenty of architectures that call themselves portable are permanent two-provider setups the Act does nothing for.

The second trap is behavioural rather than legal. A rule landing on a date does not change a company's habits on that date, as India's crackdown on dark patterns at checkout showed at consumer scale. Expect the first quarter of the free-exit regime to be full of technically compliant friction, just as the lock-in trade-off between an eSIM and a physical SIM survived every rule written about number portability.

• A renewal that auto-extends before January quietly buys the provider another year of your data.
• Export formats that no rival can import satisfy the letter of a right and none of its purpose.
• Support you asked for during the move stays billable, so agree its scope before the work starts.
• A delay that arrives without notice is not a technical extension, it is a negotiating position.

Open last month's invoice this week and find the data transfer line. If it is small, or if your move is a clean break from a provider that already waived the fee, start now and stop paying to postpone a decision you have made. If it is large and nothing else is forcing your timing, put the migration in the January column, write the notice letter now, and let the deadline argue on your behalf.

Your mother's phone does two things. It rings, and it takes an OTP. She has never opened a browser on it and she is not going to. Yet every 84 days somebody in the house recharges that phone with a pack built around a daily data allowance nobody there will ever open. The TRAI new recharge rules being fought over right now are aimed at exactly that bill, and at the near identical one paid in millions of Indian households every quarter.

What Do The TRAI New Recharge Rules Actually Change?

The draft Thirteenth Amendment would require every operator to sell a voice and SMS only voucher for each validity period it offers a bundled pack at, priced with a largely proportional reduction for the removed data. TRAI issued it for comment on 7 April 2026 and closed submissions later that month.

This is the second attempt. A December 2024 order already told operators to carry at least one voice and SMS voucher, and by late January 2025 all three private networks had one on the shelf. They complied with the letter of it. What they did not do was spread those packs across the short validities most low income subscribers buy, or cut the price by anything close to the value of the data removed. One voucher at 84 days and one at a year, and the box was ticked.

TRAI's own reasoning in 2024 put roughly 150 million subscribers on basic or feature phones with no data requirement at all. It is a large enough number that "there is no demand" stops being a credible answer. The same cost of ownership logic runs through India's repairability index and the repair versus replace maths: the question is never whether a product works, it is whether you are paying for capability you will never use.

The operators pushed back hard in June 2026, and their arguments are worth separating. Jio's is technical: 4G and 5G are fully IP based, voice rides the data bearer, so a data free plan is a billing fiction rather than a network reality. Airtel warned about digital exclusion. Vodafone Idea raised the practical one, that background app traffic, OTPs and software updates will quietly tip customers into pay as you go charges they never agreed to. The technical objection is the weakest of these, or at least it is the one that proves too much: if voice cannot be unbundled from data, the voice only packs already on sale in 2025 should not have been possible either.

That last figure is the one people forget. The voice only packs did not arrive because anybody at a telco wanted them. They arrived because a mid 2024 round of price increases pushed a wave of subscribers off the networks, and the regulator moved in response. Read the sequence backwards and the current draft stops looking like fresh regulation. It looks like TRAI closing a gap it left open the first time.

Are Voice And SMS Only Plans Actually Cheaper?

Yes, at the one validity where they exist. On an 84 day cycle the data free pack costs about 44 per cent less on Jio and about 48 per cent less on Airtel than that same operator's cheapest daily data plan, which is a real saving rather than a rounding difference.

Here is what the shelf looked like in April 2026, taken from the operators' own published 84 day and annual tariff listings. Read the rupee gaps first, then look at the row that matters most, which is not a price at all.

Look at the per day row again. The voice packs are already the cheapest way to keep a line alive, so the rate argument was settled before this draft was written. The row that decides everything is the one below it. Three months upfront on Jio or Airtel, nine on Vi, and no 28 day option anywhere. I would push back on the idea that a cheaper 84 day pack helps the person TRAI says it is protecting. Shape matters more than rate here, and a household that recharges in small amounts because that is what the week allows cannot use a quarterly pack however good the arithmetic looks.

Friction Points: Where A Data Free Recharge Plan Bites

A voice and SMS only pack strips data completely, so anything the phone does quietly in the background either fails outright or falls to pay as you go billing, and the full validity has to be paid for in a single upfront amount rather than monthly.

Vodafone Idea's objection is the one worth taking seriously, because it describes a failure mode the customer will not see coming. A smartphone on a voice only pack does not stop trying to reach the network. Play Store updates, WhatsApp sync, a banking app checking in, an OTP that arrives through a data channel instead of SMS. Any of those can trigger default per megabyte charges on a balance the customer topped up for calls. The disclosure burden here belongs entirely to the operator, and it is the same category of quiet default that India's dark patterns crackdown and the checkout tricks it names was written to stop.

There is also a question nobody has answered properly. If voice genuinely runs as an application over the data bearer, what does a proportional price reduction even mean? Removing a customer's data allowance does not remove the operator's cost of carrying their calls. My own view is that TRAI has stated a principle without stating a formula, and until it does, every operator will read proportional as generously as it can get away with. That is not cynicism about telcos, it is what an undefined pricing test always produces.

• Turn mobile data off in settings, not just in the app, before switching a smartphone to a voice only pack.
• Check whether your bank sends OTPs by SMS or through an in app push, because the second one needs data.
• Confirm the pack keeps incoming calls alive for the full validity and does not need a separate top up to stay active.
• On a feature phone the risk is close to zero, which is exactly the handset these packs were designed around.

If somebody in your house is on a feature phone or a smartphone used only for calls, move them to the voice and SMS pack today and stop waiting for the amendment. Open your operator's app, find the 84 day voice voucher, and check what you have been paying instead.

You are two taps from paying. The number you agreed to somewhere back on the product page was ₹412. The number on the screen in front of you now says ₹468, and in the gap between those two screens a handling charge appeared and a membership plan added itself to your basket. You never said yes to either one. You also never said no, and that, precisely, is the design.

Why This Enforcement Round Actually Matters

The Ministry of Consumer Affairs disclosed the action in a written reply to the Rajya Sabha on 4 August 2026. Nine digital platforms penalised, among them IndiGo, Zepto, FirstCry, Physics Wallah, BookMyShow and SpiceJet. Quick commerce, aviation, edtech, ticketing, baby products. That spread is the story. This is not a rogue category behaving badly, it is a default way of building a checkout screen that spread across every sector at once because it works.

Zepto Marketplace drew the largest single penalty, ₹7 lakh, and the reasoning is worth reading closely because it is the exact thing that happened to your ₹412. The regulator found that showing a lower price first and appending handling charges later is drip pricing. Automatically including a membership plan without clear consent is basket sneaking. Two separate named offences in one checkout flow, on one order, in about four seconds of your attention. The government says Zepto has since removed both.

And the legal footing is older than the fines. The Prevention and Regulation of Dark Patterns Guidelines, 2023 already name thirteen specific practices: false urgency, basket sneaking, confirm shaming, forced action, subscription trap, interface interference, bait and switch, drip pricing, disguised advertisements, nagging, trick questions, SaaS billing and rogue malware. In June 2025 the CCPA told e-commerce platforms to self-audit and strip these out. What changed in August 2026 is only that somebody finally started collecting money. Naming a thing is what makes it arguable, which is the same reason India's repairability index for phones mattered before a single score was printed on a box.

Here is the scale of the thing, in the four figures that decide whether you should care.

That last figure is the one that should bother you, because it moved in the wrong direction across two years of the guidelines already being in force. Whatever the self-audit regime achieved between 2024 and now, it did not stop hidden charges reaching more people. And the per-person cost is deliberately small: a 2026 Datum Intelligence report, Dark Patterns in India's Online Marketplaces, puts the average loss at ₹78 to ₹87 a month per shopper, which aggregates to somewhere between ₹25,000 crore and ₹28,000 crore a year. Nobody disputes a ₹56 handling fee. That is the whole business model. It is engineered to sit just under the amount worth arguing about, in the same quiet way that a subscription storage bill creeps upward while you are not looking.

The Thirteen Tricks, In Plain Language

Most coverage lists the thirteen names and stops there, which is useless at the moment you actually need it. What follows is what each family of tricks looks like on a real screen, and the tell that gives it away before you have paid. Seven of the thirteen cover the overwhelming majority of what you will meet in an Indian checkout this year.

Read down the Tell column and one thing links all seven. Every single one of them relies on you being in a hurry. None of these survive a reader who stops for ten seconds and compares the total on screen against the number they had in their head when they opened the app. That is not a moral failing on your part, it is arithmetic: the platform has spent years testing that screen and you have spent four seconds on it. The lopsidedness shows up clearly in how widespread each family has become.

Across the platforms Datum Intelligence studied, roughly seven in ten pushed users into choices through forced action, close to seven in ten held charges back until the final step, and more than half advertised one thing and delivered another.

Where This Gets Messy

Now the part the press release will not tell you. The total collected, spread across nine companies, several of which turn over that much before lunch, is not a penalty. It is a filing fee. Zepto's fine costs about what a mid-sized outdoor hoarding costs, levied for a practice that ran on millions of orders. If the arithmetic of a penalty is smaller than the revenue from the behaviour, it becomes a line item, and any finance team can do that sum faster than the regulator can. The pattern is familiar to anyone who has watched a service centre quote against a car's actual value: the number is set at whatever the other side will absorb without complaining.

The self-audit design has the same problem, only politer. Platforms were asked in June 2025 to inspect themselves and remove what they found. Marking your own homework is a reasonable opening move, and I would rather have it than nothing, but the 63% figure earlier suggests how far goodwill gets you. Most of the current enforcement is suo motu or complaint-driven, which means it depends on somebody noticing and somebody escalating. Well, that somebody is you, and there is currently no version of this system where it is not.

There is also a wider point that gets lost in the outrage. Not every pre-ticked box is malice; some are genuine product decisions made badly by a team optimising a conversion metric nobody sanity-checked. That distinction matters for how you respond, because the fix for the first is a complaint and the fix for the second is switching platforms. Either way you are being asked to stay alert during the one moment you are least inclined to, which is the same trap as agreeing to an eSIM activation you cannot easily reverse while a shop assistant waits. The practical watch-outs hold in both cases:

• Guidelines, not statute: the 2023 framework is guidance under consumer law, so remedies stay soft and penalties stay small relative to the money involved.
• Enforcement is reactive: action largely follows complaints or the regulator's own notice, meaning most instances are never seen at all.
• Compliance can be cosmetic: a platform can rewrite one button, keep the underlying flow, and technically satisfy the objection raised.
• The window is closing: the self-audit advisory runs out at the end of this year, and what replaces it has not been set out.

So do the boring thing that actually works. Before you open the app, decide the number you are willing to pay, and check it against the final total before you authorise anything, every time, including on the platforms you trust. The standard advice to read the terms is useless here because the terms are not where the money moves. And if the total has grown between the product page and the payment screen, that is a complaint to the National Consumer Helpline with a screenshot attached, not a shrug. The burden of noticing should not sit entirely on the person spending the money. Until it stops doing so, notice anyway.

Related: What the TRAI new recharge rules mean for the mobile data you never use