Co-Location, DMA & Latency in HFT Explained

HFT and algo trading

All HFT is algo trading, but most algo trading is not HFT. An algo is any software that places orders by rules. HFT is the part of it where the edge comes from being faster than other traders.

How HFT firms make money

HFT strategies earn small amounts per trade, many times a day.

• Market making: quoting a bid and an ask at the same time and earning the spread when both sides trade. The risk is holding inventory when the price moves, so market makers update their quotes constantly.
• Arbitrage: trading price differences between related instruments, such as Nifty futures against the stocks in the index, the same stock on NSE and BSE, or option prices that drift away from put-call parity.
• Short-term signals: reacting to changes in the order book, such as a large order appearing or disappearing, before other traders do.

Each trade earns little, so costs, speed and volume decide whether a strategy works.

Co-location

Co-location means renting rack space for trading servers inside the exchange's data centre, next to its matching engine. NSE launched co-location in 2010. A co-located server avoids the long network path an outside order takes, so its orders reach the matching engine in microseconds. It also receives market data before members outside the facility.

By 2024, co-location carried about half of NSE's derivatives trading. Co-located servers accounted for 50.1% of NSE equity derivatives turnover in FY23 and 49.6% in March 2024, up from 8% in FY11, according to exchange data reported by Business Standard.

Racks are offered to exchange trading members, such as brokers and proprietary trading firms. Individual traders cannot rent them.

Direct market access (DMA)

Direct market access lets an institutional client send orders through its broker's systems straight to the exchange, without anyone at the broker handling them. SEBI allowed DMA for institutional investors in 2008. The broker still applies its risk checks to every order.

DMA is slower than co-location but faster than routing orders through a trading app. In March 2024, DMA accounted for 13.6% of NSE derivatives trading, according to exchange data reported by Business Standard.

Latency: where the time goes

Latency is the delay between an event and the response to it, such as a price change and the order that reacts to it. An order from a retail algo passes through several steps:

1. The strategy receives a price update and decides to trade.
2. The order travels over the internet to the platform or broker.
3. The broker's risk system checks margin, quantity and price limits.
4. The order travels to the exchange gateway.
5. The matching engine fills the order or adds it to the book.

For a retail order sent over the internet, these steps usually add up to tens of milliseconds or more. A co-located server sits metres from the matching engine on dedicated network links, so its round trip is measured in microseconds. A millisecond is a thousand microseconds.

Rules for algo trading and HFT in India

The 2012 guidelines still shape how every algo order is handled. Brokers check price and quantity limits before an order leaves their systems, exchanges can stop an algo that malfunctions, and members who send many orders for each trade they complete pay extra charges.

The 2025 retail framework adds a separate route for algos that individuals run through a broker's API. Orders must come from a static IP registered with the broker, after a daily login, and are tagged so the exchange can trace them. An algo can run without registration only if it sends fewer than 10 orders a second. HFT firms do not use this route: they trade as exchange members or through their brokers' co-located systems, under the 2012 rules and the exchange's co-location terms.

In the Jane Street case, SEBI alleged that between January 2023 and March 2025 the firm bought large amounts of bank stocks to lift the Bank Nifty index while holding index option positions that gained from the move. SEBI ordered it to deposit ₹4,843.57 crore of alleged unlawful gains and barred it from trading until it did. Jane Street denies the charges. It deposited the amount, which lifted the bar, and challenged the order before the Securities Appellate Tribunal.

How much Indian trading is algorithmic

Most orders on NSE now come from algos. In FY25, algos accounted for about 70% of NSE equity derivatives turnover and 54% of cash market turnover, according to NSE data. In April 2025, the derivatives share was 69%.

SEBI's September 2024 study of F&O traders found that most of the profits earned by foreign portfolio investors and proprietary traders in FY24 came from algorithmic trading, while 93% of individual traders lost money between FY22 and FY24. SEBI's studies released in August 2026 found that 87.7% of individual traders lost money in FY26. Our algo trading growth report covers the algo share of volumes.

Can retail traders do HFT in India?

Not in practice. Four things stand in the way.

1. Access: co-location racks are for exchange members, not individuals.
2. Order rate: SEBI's retail framework lets your own algo run without registration only if it sends fewer than 10 orders a second, far below HFT rates. See SEBI's algo trading rules.
3. Latency: an order sent over the internet and through a broker's risk checks takes milliseconds, while co-located systems work in microseconds.
4. Cost: rack space, exchange data feeds, network links and specialised hardware cost far more than a retail strategy could earn from speed.

Retail traders can automate strategies where speed is not the edge:

• rule-based intraday strategies such as an opening range breakout or a VWAP strategy,
• option selling with defined stop losses, such as an iron condor,
• pairs trading on daily or hourly data, and
• risk controls such as an MTM stop loss, which act on your own positions.

Speed still matters at the margins. A platform that sends orders within a second of a signal, uses limit orders and keeps slippage low does better than one that lags. Our guide to order types explains how to fill orders well.

Does HFT hurt retail traders?

It depends on the strategy. Market makers that quote both sides narrow spreads, which lowers costs for everyone who trades liquid contracts such as Nifty options. Concerns arise when speed is used to trade ahead of other orders, or when quotes are placed and cancelled with no intention of trading, which exchanges treat as manipulation.

Regulators respond with charges on high order-to-trade ratios, surveillance of algo orders and, since the 2025 framework, Algo IDs that let the exchange trace each algo order to its source.

For a retail trader, the practical effects are:

• spreads in liquid contracts are tight, so marketable limit orders fill close to the quote,
• prices react to news within milliseconds, so trading on headlines is a race retail traders lose, and
• thin contracts can move sharply on small orders, so market orders there are risky.

Related guides

• SEBI rules for option trading: the retail algo framework in full
• Order book dynamics: how bids, asks and depth change
• Slippage meaning: why fills differ from the price you saw
• Order types: filling orders well when speed is not your edge

Frequently asked questions

Sources

• SEBI, circular on safer participation of retail investors in algorithmic trading, 4 February 2025, and glide-path circular, 30 September 2025 (sebi.gov.in)
• SEBI, broad guidelines on algorithmic trading, 30 March 2012, and measures to strengthen algorithmic trading and co-location, April 2018 (sebi.gov.in)
• SEBI, study of profit and loss of individual traders in equity F&O, 23 September 2024, and studies for FY25-FY26, press release 50/2026, 20 August 2026
• NSE data on the algo share of turnover, as reported by Business Standard, "In a first, machines overtake humans in NSE's cash market trades", 15 April 2025 (business-standard.com), and Outlook Business, "Algos have rewritten the script for trading in India", 20 June 2025 (outlookbusiness.com) (checked 20 October 2026)
• Business Standard, "Majority of F&O trading from co-location for second consecutive year", 14 March 2024, and "Share of co-location declines in NSE's trading segments in March", 29 April 2024 (co-location and DMA shares) (business-standard.com)
• SEBI's ₹1,491 crore settlement with NSE and the closing of its Supreme Court appeals, as reported by Outlook Business (outlookbusiness.com) and Free Press Journal (freepressjournal.in), checked 20 October 2026