Pushback on the sub‑nanosecond arms race

A fight over less than a millionth of a second is spilling into public view. In a widely shared post on X, a chief executive in the exchange and market-technology world argued that exchanges chasing sub-nanosecond speed are mostly helping the very fastest high-frequency trading firms, while pushing everyone else into a more expensive race. (x.com) A sub-nanosecond is less than one billionth of a second. At that scale, the contest is no longer about whether a human trader can react faster, or even whether normal software can react faster. It is about whether one firm’s hardware, cables, and physical placement inside a data center can beat another firm’s by tiny slices of time. (quantvps.com) That race has been building for years. High-frequency trading firms buy co-location, which means placing servers next to an exchange’s matching engine, and they spend heavily on custom chips, network cards, and feed handlers because a few microseconds can change who gets to a quote first. (quantvps.com) The usual defense of this speed race is that faster markets are better markets. If more firms can update quotes faster, the argument goes, prices should reflect new information more quickly and spreads should stay tight. That is the engineering story. (bis.org) The criticism is that the engineering story leaves out the bill. When exchanges and trading firms compete to shave latency from microseconds into nanoseconds, the cost lands on everyone who needs to connect, measure, and respond at that new standard, even if only a small group can actually win at that game. (quantvps.com) There is also a market-structure complaint, and it is more specific than “fast trading is bad.” Researchers at the Bank for International Settlements and the Quarterly Journal of Economics estimated that latency arbitrage accounts for about one-third of effective spread and price impact, and that market designs that eliminate it could reduce the cost of liquidity for investors by 17 percent. (bis.org, academic.oup.com) Latency arbitrage is the basic problem underneath the argument. One trader sees that prices are about to move, races to hit stale quotes before slower traders can cancel them, and earns money from being first rather than from taking long-term market risk. (academic.oup.com) That can make displayed liquidity look deeper than it really is. If market makers know they are likely to be picked off whenever prices jump, they often respond by quoting smaller size, widening spreads, or canceling faster, which can leave the market looking liquid until the instant it matters. (onlinelibrary.wiley.com, sciencedirect.com) That is why the X post’s proposed remedy was not “build even faster.” It was to use deliberate delays of about 100 microseconds, often called speed bumps, so exchanges compete on market design instead of forcing every participant to fund the next round of hardware escalation. (x.com, sec.gov) A speed bump sounds dramatic, but in market terms it is tiny. Investors Exchange, better known as IEX, built its equities venue around a 350-microsecond delay created with 38 miles of coiled fiber, and it describes that mechanism as a way to make the exchange a fair referee for trading. (iexexchange.io, sec.gov) Regulators are still arguing over how far that idea should go. In 2025, IEX proposed an options exchange with a 350-microsecond delay on incoming orders and quotes, and the filing drew support from some market participants who said it could protect market makers from stale-price risk and criticism from others who said it could create inaccessible or misleading quotes. (sec.gov, sec.gov, sec.gov) Academic evidence on speed bumps is not one-sided, but it is serious enough that the idea has stayed alive. A recent study on New York Stock Exchange American found that a speed bump can lower adverse-selection costs by reducing informed trading, while other papers warn that the exact design of a delay can shift liquidity in unintended ways. (onlinelibrary.wiley.com, sciencedirect.com) That is what makes this latest pushback notable. The complaint is not that engineers should stop optimizing systems. The complaint is that if exchanges reward only the last few hundred nanoseconds, they are choosing a market structure that concentrates advantage in a narrow slice of firms and pushes capital toward speed infrastructure instead of broader liquidity provision. (x.com, bis.org) The audience that reacted strongly to the post was not a general political crowd. It was the trading-technology community itself, which is why the reaction matters: the debate is moving from private complaints in colocation rooms and exchange meetings into a public argument about what markets should optimize for. (x.com) The open question is whether exchanges and regulators will treat this as a technical tuning issue or as a rulebook issue. If they treat it as a rulebook issue, then tools like speed bumps, quote protections, and fee design become ways to redirect competition away from sub-nanosecond races and toward the quality and durability of liquidity. (sciencedirect.com, sec.gov)