Stock exchanges compete on speed. Faster execution attracts order flow, and for years the assumption was that faster is always better. But speed creates its own problems, and a small but growing number of exchanges have started deliberately slowing things down. Adding delay seems counterintuitive in a market measured in microseconds. The question isn't whether speed is good. It's who benefits from it, and who pays for it.
Modern equity trading is fragmented across dozens of venues. When a large order enters the market, it typically gets routed to multiple exchanges simultaneously. But "simultaneously" is a fiction. Information travels at different speeds to different places, and the gaps between arrival times, even gaps measured in microseconds, create opportunities.
Latency arbitrage exploits these gaps. A firm with faster connections can see an order arrive at one exchange and race ahead to others, buying up available shares before the rest of the order lands. The original buyer pays a slightly higher price. The fast firm pockets the difference. On any single trade, the cost is negligible. Fractions of a penny. Across billions of shares daily, it becomes significant. The money flows from slower participants (pension funds, mutual funds, retail brokers) to faster ones (proprietary trading firms with optimized infrastructure).
This isn't illegal. It's not even controversial in the way insider trading is controversial. But it raises a structural question about whether market design should reward speed itself or reward better decisions.
A speed bump is an intentional delay built into an exchange's matching engine. Every incoming order passes through the delay before it can execute.
IEX, the exchange that popularized the concept, uses a 350-microsecond delay, roughly the time it takes light to travel 65 miles. They implemented it physically, coiling 38 miles of fiber optic cable into a box that every order must traverse. Symmetry is the key: the delay applies to everyone equally. But its effect is asymmetric. For a long-term investor submitting a buy order, 350 microseconds is imperceptible. For a latency arbitrage strategy trying to race ahead of that order, 350 microseconds is an eternity.
What makes this effective is timing. The exchange can update its quotes before processing incoming orders. When prices move on other venues, IEX's displayed quotes can adjust before a fast trader arrives to pick them off. The speed advantage that made latency arbitrage profitable disappears.
IEX built more than a speed bump. The exchange structured its entire model around removing advantages that come from infrastructure rather than insight.
Most exchanges offer co-location, server space adjacent to the matching engine where proximity means speed. IEX doesn't. All participants connect through the same access point. Complex order types are another source of edge, and sophisticated traders use them to manage execution in ways casual participants don't understand. IEX offers a simplified set. Pricing structures also differ: many exchanges use maker-taker models, paying rebates to some participants and charging fees to others in ways that can misalign broker incentives. IEX uses flat fees.
None of these choices is radical on its own. Together, they reflect a philosophy: competition should happen on the quality of trading decisions, not on who spent more on infrastructure.
IEX wasn't the last exchange to experiment with intentional delay.
NYSE American adopted a 350-microsecond speed bump in 2017, then removed it two years later. The delay didn't fit their market's characteristics. Canada's NEO Exchange took a different approach: an asymmetric delay that applies only to orders that remove liquidity, targeting the specific behavior latency arbitrage depends on. Nasdaq avoided speed bumps entirely but introduced an order type called M-ELO that holds orders for half a second before they become eligible to execute. That's a delay measured in milliseconds rather than microseconds, aimed at a different problem.
These variations reflect genuine uncertainty about what fairness means in market structure. Symmetric delays treat everyone equally but may not address the specific harm. Asymmetric delays target the problem more precisely but introduce complexity and judgment calls about who deserves protection.
Studies examining trading activity on IEX have found narrower spreads for stocks that trade there frequently. The SEC published research on this in 2018. IEX's market share has grown steadily since it became a national exchange in 2016. By late 2024, its displayed market share had tripled from earlier that year. In 2025, the exchange received industry recognition for innovation.
That growth suggests the model appeals to at least some segment of the market. Whether speed bumps should become standard or remain a niche alternative is still an open question.
Speed bumps don't eliminate high-frequency trading. They don't prevent all forms of latency arbitrage. They address one specific structural advantage and leave others intact.
For most retail investors, the choice of exchange happens invisibly. Your broker's smart order router decides where your order goes. But the design of those exchanges affects the price you receive. When latency arbitrage runs freely, you pay a small tax on every trade. When exchanges build mechanisms to limit it, some of that tax disappears.
Market structure involves choices. Faster isn't automatically better. In some cases, slowing things down changes who benefits from the market's design.