NSE co-location creates latency asymmetry

The National Stock Exchange of India’s co-location controversy has often been described as a market-abuse case. The technical record shows it was also an infrastructure case. SEBI’s 2019 order said NSE’s dissemination architecture up to March 2014 was based on time priority, meaning trading members who logged in first on a dissemination server received market data before those who logged in later. That detail matters because co-location is supposed to narrow physical distance, not create new advantages inside the exchange’s own network. SEBI’s 2015 circular said exchanges must ensure “fair and equal access” to co-location facilities and that users experience “similar latency” with respect to exchange-provided infrastructure. The result is that the NSE dispute is best understood as a question about how fairness is engineered at the packet level, not only how misconduct is punished after the fact. (sebi.gov.in) That framing is consistent with the recent technical analysis circulating on X, but it is also grounded in SEBI’s own findings and NSE’s published colocation architecture documents. ### How could co-location still produce unequal speed inside the same venue? (sebi.gov.in) SEBI’s 2019 order said NSE’s system disseminated tick-by-tick data through TCP/IP-based POP servers and that members who logged in first on a server were disseminated data earlier than those who logged in later on the same server. The regulator said some TBT servers were faster than others and that OPG Securities consistently logged in first to two TCP/IP-based POP servers and their backup servers. (sebi.gov.in) That means the edge did not have to come from being closer to the exchange building. It could arise from where a participant landed in the exchange’s server sequence, which server handled the connection, and whether the participant repeatedly secured the earliest session on a faster path. The advantage described by SEBI was measured in the order of message arrival, which in high-frequency trading can translate into microsecond-to-millisecond differences. That last point is an inference from the architecture and trading context, not a quoted SEBI finding. (sebi.gov.in) ### Where do server allocation and login sequence enter the story? NSE’s own colocation guidelines say the exchange provides TAP IP addresses, user IDs and rack-based infrastructure inside the colocation data center. The guidelines also say the exchange provides 40 IPs per full rack and 20 IPs per half rack. SEBI’s 2019 order said the number of IPs used by OPG Securities was significantly larger than the majority of other trading members availing tick-by-tick data during the period when OPG logged in first. (sebi.gov.in) That finding matters because repeated connection attempts, broader IP use or more aggressive session management can affect which gateway or dissemination point a participant reaches first. (inspection.nseindia.com) A recent NSE protocol document for non-NEAT frontends says users initially connect to a gateway router, which then assigns a gateway. That shows the exchange path includes assignment logic before messages reach the participant application, adding another layer where timing and allocation can matter. ### Why does TCP/IP dissemination keep appearing in the case record? (sebi.gov.in) SEBI’s findings focused on TCP/IP-based dissemination because TCP sessions are stateful and sequenced by connection order. In the 2019 order, the regulator said the architecture up to March 2014 effectively rewarded the first login on a server. That is why recent commentary has stressed packet timestamps, login order and order-entry timing rather than broad claims about “faster traders.” If fairness depends on who connected first, which server they were assigned and how quickly they could act on the feed, then the evidence has to come from the full chain of connection, dissemination and order submission. (nsearchives.nseindia.com) That is an inference drawn from SEBI’s findings and NSE’s network documents. (sebi.gov.in) ### What would prove or disprove asymmetry more cleanly? SEBI’s 2015 circular already requires exchanges to publish quarterly latency reports and to identify orders emanating from co-located servers and the resulting trades. A fuller audit trail would go further. The relevant measurements would include gateway assignment logs, login timestamps, hardware-synchronized packet timestamps, server-level dissemination times and the elapsed time from feed receipt to order entry. (sebi.gov.in) NSE’s current colocation guidelines say the facility offers both NTP and PTP time-synchronization services, which indicates the exchange already operates timing infrastructure that could support tighter observability. (sebi.gov.in) ### Where does the case stand now? On April 21, 2026, Moneycontrol reported that SEBI’s High-Powered Advisory Committee had recommended NSE’s settlement application and that the next step was review by a panel of two whole-time members. The report said NSE had offered 1,387.39 crore rupees in June 2025 and that the final recommended amount could be about 1,880 crore rupees for the co-location and dark-fibre cases. (inspection.nseindia.com) If SEBI’s whole-time members approve the proposal, Moneycontrol reported that SEBI and NSE are expected to move to withdraw the case from the Supreme Court. NSE declined comment to the publication, and SEBI did not respond by the time of that report, Moneycontrol said. (moneycontrol.com)