Quick test: what did the 2G scam cost India?

You said ₹1.76 lakh crore, didn’t you. Everyone does. Now the follow-up: who was convicted for it?

Nobody. Not one person, not one company, in eighteen years. And yet the Supreme Court cancelled all 122 licences and, as recently as 2022, refused to hand one of those companies its entry fee back. Both of those things are true at once, and if that feels like a contradiction, this post is for you.

I went through the primary record for this one: the CAG report, the 2012 and 2022 Supreme Court judgments, all three trial court judgments from December 2017 (yes, three… most people don’t know the third exists), the TRAI recommendations from 2007, and the Delhi High Court’s live docket. Not the coverage of the documents. The documents. What follows is the picture they actually paint, which is sharper and honestly stranger than the one in public memory.

The setup: a 2001 price in 2008

In the licensing regime of the time, you didn’t buy spectrum. You bought a Unified Access Service Licence, and startup 2G spectrum came bundled with it.¹ So the “price of spectrum” was really the licence entry fee, and that fee was ₹1,658 crore for a pan-India licence… a number discovered in a 2001 auction, when the Indian telecom market was a fraction of what it became.¹

In January 2008, the Department of Telecommunications sold licences at that seven-year-old price. To be honest, if the story stopped there it would already be worth a post. It doesn’t stop there.

The mechanics of that January are the part every institution that later looked at this found indefensible. DoT announced on 24 September 2007 that applications would be accepted till 1 October. 575 applications piled in, 408 of them in the final eight days! Then DoT quietly restricted Letters of Intent to applicants who had filed by 25 September, a cut-off that did not exist when people were applying.¹ On 10 January 2008 it issued 121 LoIs, and priority went not to whoever applied first but to whoever could show up with a demand draft first. Some applicants completed their compliance by mid-afternoon that same day, drafts already made out; whether they had prior word of the date is exactly what the prosecution later alleged, and could not prove.² 120 licences were issued in a single day, two more followed in July. That’s your 122.¹

And the regulator? TRAI’s August 2007 recommendations are the most misquoted document in this whole affair. It did recommend no cap on the number of operators.³ On pricing, it said, in the same document, that the 2001 entry fee “is also not the realistic price for obtaining a license. Perhaps, it needs to be reassessed through a market mechanism”… and then declined to recommend revising it for new 2G entrants, on level-playing-field grounds, while pushing auctions for every band except 2G.³ So both camps get to quote TRAI, and both are quoting half a paragraph.

Institution one: the auditor and the number that ate the story

Here’s the thing almost nobody who cites ₹1.76 lakh crore knows: the CAG never gave one number. Report No. 19 of 2010-11 gives a table of four estimates, each against a different benchmark:¹

Benchmark the CAG used	Presumptive loss
Swan Telecom’s equity sale	₹57,666 crore
S Tel’s offer to the government	₹67,364 crore
Unitech’s equity sale	₹69,626 crore
2010 3G auction prices	₹1,76,645 crore

The word doing all the work is presumptive, and it’s the CAG’s own word, stated up front: “we have only attempted to arrive at a presumptive value in this Report.”¹ The ₹1.76 lakh crore figure is what you get if you assume 2G spectrum in 2008 would have fetched what 3G spectrum fetched in 2010. You can think that’s a reasonable proxy or a wild one (former RBI Governor D. Subbarao called it “contestable”⁴), but the report itself hands you a range where the top number is three times the bottom one.

The headline kept the top number and dropped the range. It also dropped this: the CBI, the agency actually prosecuting the case, put the loss in its chargesheet at ₹30,984.55 crore.⁵ The auditor and the prosecutor never agreed on the size of the harm, and no court has ever fixed a number. Eighteen years in, the size of the loss is still nobody’s finding.

Where did the benchmarks come from? From the market itself, which is what makes them sting. Swan paid ₹1,537 crore for licences in 13 circles and, per the CBI’s FIR, sold 45% of itself to Etisalat for ₹4,200 crore within months. Unitech paid ₹1,658 crore for all 22 circles and sold 60% to Telenor for about ₹6,100 crore.⁵ Neither had rolled out a single service. Whatever the licences were worth, it was not what DoT charged for them.

Institution two: the Supreme Court, on process

On 2 February 2012, in Centre for Public Interest Litigation v. Union of India, the Supreme Court quashed all 122 licences.⁶

Read the judgment and two things jump out. First, what it decided: that the exercise was “wholly arbitrary, capricious and contrary to public interest apart from being violative of the doctrine of equality.” A first-come-first-served queue where the queue order itself got redefined mid-game is not a lawful way to hand out a scarce public resource. That’s a constitutional finding about process.

Second, what it pointedly did not decide. The Court explicitly declined to rely on the CAG’s findings (Parliament’s committees were still examining them). It imposed costs on the companies that had flipped equity, but it convicted no one, and its final direction says in terms that the criminal case must proceed “uninfluenced by this judgment.”⁶ Also, a Constitution Bench clarified later that same year that auctions are not a constitutional mandate for all natural resources; the auction direction was specific to this spectrum.⁷ The popular version of 2012 (“Supreme Court says everything must be auctioned”) is not what happened.

Institution three: the criminal court, on proof

On 21 December 2017, Special Judge O.P. Saini pronounced three judgments, not one. The CBI main case: 17 accused, all acquitted. The ED’s money-laundering case: 19 accused, all acquitted. And the case everyone forgets, against the Essar and Loop promoters: 8 accused, all acquitted.⁵⁸⁹

The sentence everyone quotes is real, and having read the 1,553-page judgment I can give it to you exactly: “I have absolutely no hesitation in holding that the prosecution has miserably failed to prove any charge against any of the accused, made in its well choreographed charge sheet.”⁵ Paragraph 1818, second-to-last page. The judge didn’t stop at “not proved beyond reasonable doubt”; he said the chargesheet itself recorded facts that were factually incorrect, and named examples.

The ED case collapsed for a cleaner reason: money-laundering needs “proceeds of crime,” the alleged crime was the one the CBI case just failed to prove, so the foundation was, in the court’s word, knocked out.⁸

An acquittal is a finding that the prosecution failed to discharge its burden. It is not a finding that the allocation was lawful. That’s not me editorialising to save the story… that’s the next judgment.

The reconciliation, from the Court itself

This is the paragraph that settles the “so was it all nothing?” question, and it’s the reason I think this post can exist under a verifiably-correct constraint at all.

In 2022, Loop Telecom came to the Supreme Court asking for its ₹1,454.94 crore entry fee back. Reasonable ask, on its face: the licences were quashed for the government’s own arbitrariness, and Loop’s promoters had since been acquitted. The Court said no, and said why:

“…the acquittal of the promoters of the appellant of these criminal charges does not efface or obliterate the findings which are contained in the final judgment of this Court in CPIL… as a beneficiary and confederate of fraud, the appellant cannot be lent the assistance of this Court for obtaining the refund of the Entry Fee.”¹⁰

Held in pari delicto: equally at fault, and the loss lies where it falls. So the reconciliation isn’t a blogger’s construction, it’s a holding. The criminal court answered “did the prosecution prove a conspiracy beyond reasonable doubt?” No. The constitutional court answered “was the allocation lawful?” Also no. Three institutions, three different questions. The answers only look contradictory if you assume they were answering the same one.

The referee with no stake in Indian politics

However… you could still object that every institution above is Indian, and every reading of this case in India is political. Fine. There was one adjudication from entirely outside.

Loop’s Mauritian shareholder took India to investor-state arbitration under the India-Mauritius investment treaty (Khaitan Holdings (Mauritius) v. India, PCA Case No. 2018-50). A tribunal seated at The Hague, chaired by a New Zealand professor, with no horse in the Indian race. In September 2025 it rejected the claims on the merits: as reported, no legitimate expectations or expropriable rights could arise from 2G licences that were issued in violation of Indian law.¹¹ The award itself isn’t public, so I’m citing the specialist arbitration press on its reasoning, and flagging that.¹²

Sistema and Telenor, the Russian and Norwegian investors burned by the quashing, both threatened treaty arbitration too (Telenor floated a $14 billion figure) and neither ever filed; no such case exists in any arbitration registry.¹³ Both quietly bought their way back in at the subsequent auctions instead. Of every foreign investor in this story, not one recovered a rupee through arbitration, and the only one who pushed to an award lost on reasoning that tracks the Indian courts’. Make of that what you will. I find it the single most under-reported fact in the whole affair.

What the market said

The counterfactual got tested, sort of. The November 2012 re-auction, the first after the quashing, committed just ₹9,407.64 crore; the entire 800 MHz band got zero bids, and Delhi, Mumbai, Karnataka and Rajasthan drew no bids at all in 1800 MHz.¹⁴ Vindication for the “the loss was imaginary” camp? Slow down: February 2014 committed ₹61,162.22 crore¹⁵ and March 2015 committed ₹1,09,874 crore.¹⁶ (All committed bid values, not cash in hand; the money comes in instalments.)

So the auctions neither confirmed ₹1.76 lakh crore nor embarrassed it. They showed what auctions show: prices move with timing, reserve prices, and who needs spectrum that year. I spend my working days around price discovery, so maybe I’m biased toward the auction as an instrument… but an instrument is all it is. It discovers the price of what you sell, when you sell it. It cannot tell you what 2008 spectrum “really” was worth. Nothing now can. That is precisely why the CAG called its own number presumptive, and why nobody should quote it as a fact.

Where this stands (as of 16 August 2026)

This section is the living part of this post. I checked the Delhi High Court’s own case-status records two days ago, and I’ll update this table as things move.¹⁷

Matter	Case number	Status	Next hearing
CBI’s appeal against the main acquittals	CRL.A. 277/2024	Pending, leave granted March 20242	28 October 2026
ED’s plea to appeal its case	CRL.L.P. 184/2018	Still awaiting leave, eight years on	28 October 2026
CBI’s plea to appeal the Essar/Loop case	CRL.L.P. 257/2018	Still awaiting leave	13 October 2026

Upcoming, concretely: two hearing dates this October. If the High Court decides the CBI appeal, the ending of this story changes, in either direction. Worth saying: no news outlet has reported on any of these matters since February 2025. The docket shows all three alive and listed. The press silence is real; the dormancy is not.

So, the distinct picture, in four sentences. An indefensibly rigged process sold public spectrum at a 2001 price in 2008, and the Supreme Court struck it down for exactly that. The criminal conspiracy alleged behind that process was tried for six years and not proven, three times over. The one number everyone remembers was the top of a four-value presumptive range that its own author refused to call a fact. And the question the public actually argues about, “was there a scam?”, is the one question no institution has answered, because “scam” is not a legal or an accounting category… it’s a story word.

The record is genuinely unfinished. Come the last week of October, we’ll know if it moves again. I’ll keep the table above honest either way.

References

Footnotes

1. CAG of India, Report No. 19 of 2010-11: Performance Audit of Issue of Licences and Allocation of 2G Spectrum (PDF). Entry fee and 2001 pricing: Ch. 5, p. 51. Benchmark table and presumptive framing: para 5.5, p. 56, and the preface. Cut-off and application-spurt findings: paras 4.1.1–4.1.2. The 121 LoIs / 120 + 2 licences reconciliation: footnote 1, p. 3. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

2. CBI v. A. Raja & Ors, Delhi High Court, CRL.L.P. 185/2018, order of 22 March 2024 (2024:DHC:2350). Reserved 14 March, pronounced 22 March; prima facie satisfaction at para 270; LoI compliance-time evidence recited at para 258. ↩ ↩²

3. TRAI, Recommendations on Review of license terms and conditions and capping of number of access providers, 28 August 2007 (PDF). No-cap recommendation: para 2.37. Entry-fee passages: paras 2.73 and 2.78. ↩ ↩²

4. Business Today, “Ex-RBI Governor D Subbarao on 2G scam: ‘CAG’s Rs 1.76 lakh crore presumptive loss contestable’”, 1 May 2024. ↩

5. CBI v. A. Raja & Ors, Special Judge O.P. Saini, 21 December 2017. Accused list and chargesheet dates: pp. 1–5. Chargesheet loss figure of ₹30,984.55 crore and the ₹200-crore allegation: chargesheet recital. Swan/Unitech stake sales: FIR recital, p. 5. The verbatim holding: para 1818, p. 1552. ↩ ↩² ↩³ ↩⁴

6. Centre for Public Interest Litigation & Ors v. Union of India & Ors, Supreme Court of India, 2 February 2012. Arbitrariness holding: para 77. Quashing and directions, including the instruction that the criminal case proceed uninfluenced: para 81. ↩ ↩²

7. In re Special Reference No. 1 of 2012, Supreme Court of India (Constitution Bench), 27 September 2012. Auction not a constitutional mandate: paras 146–149. ↩

8. ED v. A. Raja & Ors (PMLA), Special Judge O.P. Saini, 21 December 2017. Proceeds-of-crime reasoning: paras 166–168. ↩ ↩²

9. CBI v. Ravi Kant Ruia & Ors, Special Judge O.P. Saini, 21 December 2017. Acquittal: paras 466–467. ↩

10. Loop Telecom and Trading Ltd v. Union of India, Supreme Court of India, 3 March 2022 (PDF via LiveLaw). Refund claim: paras 1–2. In pari delicto conclusion and the acquittal-does-not-efface passage: para 60. ↩

11. IAReporter, “Khaitan v. India tribunal finds no breach on the merits, ruling that no legitimate expectations or expropriable rights could arise from 2G licenses issued in violation of Indian law” (2025). The award text is not public; the merits reasoning is cited as reported here. ↩

12. Khaitan Holdings (Mauritius) Limited v. Republic of India, PCA Case No. 2018-50: case page at italaw. ↩

13. Knowledge at Wharton, “Revoked Indian Telecom Licenses Spur Legal Action”, March 2012, on the Sistema and Telenor treaty notices. “Never filed” rests on the absence of any Sistema or Telenor case against India in the italaw, PCA and UNCTAD ISDS registries as of August 2026. ↩

14. Department of Telecommunications, Annual Report 2012-13 (PDF, archived), pp. 35–36: November 2012 auction result and unsold spectrum. ↩

15. Department of Telecommunications, Annual Report 2013-14 (PDF, archived), pp. 4 and 30: February 2014 auction, ₹61,162.22 crore total. ↩

16. Press Information Bureau, “The Auction of Spectrum in 2015 ends on Day 19 after 115 Rounds of Bidding”, 25 March 2015. ↩

17. Delhi High Court case-status portal, records for CRL.A. 277/2024, CRL.L.P. 184/2018 and CRL.L.P. 257/2018, accessed 15 August 2026. The portal is captcha-gated and carries an information-only disclaimer; statuses in the table reflect that date. ↩