30 Apr 2019. A whole time member of SEBI signed an order in the matter of NSE colocation. Its directions to NSE were to pay Rs 624.89 crore with 12% interest from 1 Apr 2014, and to stay out of the securities market for six months. The appellate tribunal, SAT, ruled on the appeals on 23 Jan 2023, and on 13 Sep 2024 SEBI passed a second order that disposed of the remanded proceedings against NSE and seven individuals without any direction. Between those dates sits a question every fast trader should be able to answer: what is fair access to an exchange's data, and what does a millisecond buy? This lesson reads the orders and circulars as they are dated and prices the millisecond. It records what they say, takes no side and is not legal advice.
3By the end you can
- Say what colocation is and what SEBI's 2015 and 2018 circulars ask of exchanges.
- Read the 2019 SEBI order, the 2023 SAT order and the 2024 SEBI order for what each holds, with its date.
- Size a millisecond of Nifty movement in rupees, and see why a retail algo cannot race.
The idea, plainly
What is colocation, and is it legal?
Colocation means a broker's or data vendor's servers sit inside or close to the exchange's premises and reach the trading platform over a direct private network. SEBI's circular of 13 May 2015 describes it so. The facility is legal and regulated. The circular tells exchanges to offer it in a fair, transparent and equitable manner, to give every participant fair and equal access to facilities and data feeds, to see that all users get similar latency from the exchange's own infrastructure, to publish quarterly reports on latencies and to be able to identify orders from colocated servers. The case that follows was never about whether colocation may exist. It was about whether access was equal.
How did the data feed reach members, as the orders describe it?
The tick by tick feed, TBT, reports every addition, change and cancellation in the order book. In the period the orders examine, 2010 to April 2014, NSE sent it over TCP/IP, which delivers to one connection after another and does not broadcast. As the 2024 order summarises it, the feed passed from the exchange's data centre to POP servers, three primary and one secondary kept for an emergency, each with three ports. Members were to log in to the primary servers. On a port, members were served in the order in which they had logged in.
Where a member sat in the queue depended on server, port and login order
What did SEBI order on 30 Apr 2019?
The order followed complaints dated 8 Jan 2015, 10 Aug 2015 and 3 Oct 2015. As the 2024 order records the findings, SEBI found that a member who logged in early on a port gained an advantage, that with no load balancer the IPs were allotted unevenly across ports and servers, and that the secondary server was the least crowded, with no documented policy or monitoring of members who used it. It found violations of regulation 41(2) and regulation 42(2) of the SECC Regulations by NSE and no fraud under the PFUTP Regulations without evidence of collusion. Its directions on NSE were to disgorge Rs 624.89 crore with 12% interest from 1 Apr 2014 within 45 days to the Investor Protection and Education Fund, to stay out of the securities market for six months, and to run system audits, reconstitute its Standing Committee on Technology and frame a whistle blower policy.
The maths: how the Rs 624.89 crore was reached
Disgorgement =
revenue x net margin
The order took NSE's revenue from the colocation facility for FY2010-11 to FY2013-14, leaving out rack charges because they were not tied to the TBT feed. NSE supplied the figures: Rs 811.54 crore. It then applied NSE's consolidated net profit margin over the same four years, 77.00%, which is profit after tax over revenue from operations. Rs 811.54 crore times 77.00% is Rs 624.89 crore. Interest at 12% a year from 1 Apr 2014 ran on top.
What did SAT hold on 23 Jan 2023?
SAT's order in appeals against the 2019 orders agreed with some findings and set others aside. As the 2024 order quotes it, SAT held that a member logging in first on a port gets a probabilistic advantage, not a certain one, and that no extra randomizer was required. It held that the absence of a load balancer had favoured some members and that NSE had not ensured fair, transparent and equal access in allotting IPs. It found no violation of regulation 41(2) or regulation 42(2), held the 2015 circular did not apply to the period of 2010 to April 2014, found a violation of the 2012 circular, and called the failures human failure and a lapse in monitoring the secondary server. On NSE it held the direction to disgorge unwarranted, since it found no unethical act by NSE and no unjust enrichment, and said the lapse could invite a penalty or another direction instead. The 2024 order does not quote the direction SAT gave in its place. NSE's red herring prospectus of 10 Sep 2026, filed with SEBI, records it: a deposit of Rs 100 crore in the Investor Protection and Education Fund for lack of due diligence. It affirmed the violations found against one trading member, set aside that member's disgorgement and sent four issues back to SEBI.
The term this section turns on: Disgorgement
What did SEBI conclude on 13 Sep 2024?
The order was passed in compliance with SAT's orders. Of the remanded issues, only the charge that NSE and its former officials colluded or connived with that member was part of this proceeding. SEBI wrote that NSE had no detailed defined policy for use of the colocation facility and did not monitor use of the secondary server, and that this did not on its own show collusion. It noted that 93 trading members in the F&O segment logged in to the secondary server between 2010 and 2016, and that the experts' reports showed no direct or indirect evidence of collusion. It held the test of preponderance of probability was not met and disposed of the proceedings against the noticees without any direction.
What happened on appeal, as NSE reports it?
SEBI appealed SAT's order to the Supreme Court. For what followed, the record I could find is NSE's own prospectus, so read it as NSE's account and not as a SEBI order. It reports a Supreme Court interim order of 20 Mar 2023 directing SEBI to refund Rs 300 crore to NSE, settlement applications for the colocation and dark fibre matters filed on 20 Jun 2025, SEBI's approval in principle by email on 30 Jul 2026 of revised terms of Rs 1,491.21 crore for both matters together, and a payment by NSE of Rs 714.74 crore on 31 Jul 2026. As the prospectus states it, the applications to dispose of SEBI's colocation appeals were pending in the Supreme Court on 10 Sep 2026.
From the first complaint to a reported settlement took over eleven years
- First complaint
Complaints to SEBI on the tick by tick feed run 8 Jan, 10 Aug and 3 Oct 2015
- SEBI colocation guidelines
Fair, transparent and equitable access, similar latency, quarterly latency reports
- SEBI circular on algo and colocation
Free tick feed, latency statistics, managed colocation, an identifier for every algo
- SEBI order on NSE colocation
Disgorgement, six months out of the market, audits and a whistle blower policy
- SAT order
Some findings agreed, some set aside, four issues sent back to SEBI
- Supreme Court interim order
On SEBI's appeal, a part refund to NSE, as NSE reports it
- SEBI order on the remand
Proceedings against NSE and seven individuals disposed of without direction
- Settlement in principle
SEBI accepts NSE's revised terms in principle, as NSE reports it
What did SEBI change for everyone?
In its discussion paper of 5 Aug 2016 SEBI described HFT as a subset of algo trading made of latency sensitive strategies. It floated five ways to level access: a minimum resting time for orders, frequent batch auctions, random speed bumps, randomising orders that arrive in a short window and a minimum transaction to order ratio. The circular of 9 Apr 2018 took a different route. Exchanges were to give every member the tick feed free of charge, to publish latency statistics including the 50th and 99th percentile, to offer a managed colocation service for smaller members, and to tag every order with its algo's identifier. The circular of 24 Jun 2020 set order to trade ratio slabs. The 2018 circular did not include the five.
Worked example: the crore and the millisecond
Start with the money in the case. NSE's colocation revenue, excluding rack charges, was Rs 811.54 crore across four years, and the chart splits it by year as the order tabulates it.
The revenue the 2019 order worked from grew every year
A millisecond of Nifty movement is about two rupees on one lot
The maths: a millisecond of volatility
Move = S x vol
x sqrt(t / year)
x lot size
A year has 252 sessions of 22,500,000 milliseconds each, and the square root of their product is about 75,299. The Nifty close of 23,118.60 times the volatility 11.5%, divided by that root and multiplied by 65 units, gives the figure for one millisecond. It is a typical size of movement. An edge is something else: a gain you can capture after fees, over many trades.
Read 3 more paragraphs, 226 words
Now the other side of the case: what a unit of speed is worth. At the average volatility of 11.5% over the 300 sessions to 15 Sep 2026, one standard deviation of the Nifty move in a millisecond is about 2.30 rupees on one lot of 65 units, and in a second about 72.76.
Read the two charts together. With every member on the same feed at the same moment, a millisecond is a few rupees on a lot and the differences left are the wire and the code. When the order of service differs by even a few milliseconds, as the orders say it did on some ports, the difference is worth having only for a firm trading in size. That is why the access rules in the 2015 circular and the free feed of 2018 focus on equal timing and published latency, not on banning speed.
For a retail algo on a broker's API the point is different. A retail algo sends its orders through the broker's API and risk checks, is held to ten orders a second per exchange unless registered, and has no rack of its own at the exchange, so the colocation race is not one it can enter. The speed bands are in HFT, MFT and LFT compared and the retail framework in Building a compliant retail algo.
Read the orders yourself
Open the order of 13 Sep 2024 linked under the sources and read paragraph 62, its directions, then paragraphs 59, 60 and 61, its conclusion, and paragraph 12.9, its summary of what SAT held on the disgorgement. Open the order of 30 Apr 2019 and read paragraph 11.1. For each order write three lines: its date, what it found and what it directed. Then take the rupee figure for a millisecond above and work out how many lots a firm would have to trade to earn one lakh rupees from a movement of that size. The answer shows why only firms trading in size care about a few milliseconds.
The common mistake
The belief: The colocation case proved that fast traders rigged NSE.
The reverse claim: It ended with no direction, so nothing was wrong.
Neither matches the orders. The order of 30 Apr 2019 directed disgorgement and a six month bar. The tribunal of 23 Jan 2023 agreed with some findings, set aside others, held the disgorgement on NSE unwarranted, directed a deposit in its place by NSE's account, and sent four issues back. The order of 13 Sep 2024 found the material insufficient to show collusion between NSE and one trading member, recorded that NSE had no detailed colocation policy and did not monitor the secondary server, and disposed of the proceedings without direction. SEBI then appealed, and NSE reports a settlement in 2026. Each date has a different holding. Quote the order and its date, and do not quote a headline.
Two terms to hold on to from 19.4
Key takeaways
- Colocation is legal and regulated. The case was about equal access.
- The 30 Apr 2019 order directed disgorgement and a six month bar on NSE.
- SAT set aside NSE's disgorgement; NSE's prospectus records a deposit direction instead.
- The 13 Sep 2024 order disposed of the proceedings without any direction.
- A millisecond of Nifty movement is about two rupees on one lot.
- Quote an order and its date, never a headline.
Where this section sits in module 19.4
Before you trade. Derivatives trading carries a high risk of loss. SEBI's study "Profitability of Individual Traders in the Equity Derivatives Segment (FY25 to FY26)" (August 2026) found that 87.7% of individual traders in equity F&O made net losses in FY26, with aggregate net losses of Rs 91,685 crore. Read the SEBI study