The most consistently profitable traders in the world are not hedge funds or famous individuals but a small group of private firms that buy and sell all day without wanting to own anything at the end of it. They make markets — posting bids and offers in stocks, options, ETFs, futures, bonds and currencies — and they trade on their own capital, so their profits are net trading revenue rather than fees. Almost none of them publishes accounts. The best figures available for 2025 come from documents and anonymous sources reported by Reuters in April 2026, which the firms did not confirm: Jane Street ~$39.6bn (a record; part reportedly from revaluing private stakes, including Anthropic), Hudson River Trading ~$12.3bn and Citadel Securities ~$12.2bn (+25% vs 2024). Citadel Securities then reported a record ~$7.3bn of trading revenue in Q2 2026 (Bloomberg via Hedgeweek, reported). Those numbers, larger than the trading revenue of most global banks, rest on a business that earns fractions of a cent per share and loses on many individual trades. The public evidence for how it works comes from the few listed firms: Virtu Financial's 2014 IPO filing disclosed a single losing day in 1,238 (S-1, 10 Mar 2014, 2009–13), updated to one in 1,485 in the IPO prospectus (16 Apr 2015). The public evidence for how it fails comes from Knight Capital, which lost more than $460m in ~45 minutes on 1 Aug 2012 (SEC order; Knight's early estimate ~$440m) from a faulty software deployment (Shared facts), and from the Flash Crash of 6 May 2010. This section explains what market makers and high-frequency traders do and earn, who the firms are, the latency race and what it costs everyone else, options and ETF market making, payment for order flow, regulation and the big cases. The companion worked example computes the "sniping tax" that a latency race imposes on investors, following Budish, Cramton and Shim (2015).
What the study covers
- 1Overview: renting out immediacy
- 2Units and definitions
- 3What market makers earn and how
- 4The firms
- 5Speed: colocation, microwaves and the latency race
- 6When it breaks: the Flash Crash and Knight Capital
- 7ETF market making
- 8Options market making and gamma-hedging flows
- 9Payment for order flow and regulation
- 10Europe, India and the rest
- 11How to read market-making claims
- 12Key numbers
- 13Landmark firms and cases
- 14Pitfalls
- 15How an expert thinks about it
- 16Glossary
Technical exhibits
Worked examples, checklists and case studies included with this study.
Latency Arbitrage Sniping Tax Worked Example
A stylized single-stock model in the spirit of Budish–Cramton–Shim: the half-spread a liquidity provider must charge to break even against snipers; how it rises with the number of racers, the size and frequency of public price jumps; the annual "sniping prize" and who pays it; and what a frequent batch auction would change (computed in Python, stylized). Not the spread-capture vs adverse-selection example of Appendix A