Latency and Market Access Infrastructure
Latency is the delay between something happening in the market and your response reaching the exchange. In competitive electronic markets it is measured in microseconds, and occasionally nanoseconds.
Why it is worth money
The obvious reason is offensive: reacting first to a price change lets you trade at a stale price before it updates.
The less obvious reason matters more to a market maker, and it is defensive. Your quotes are resting in the book. When the market moves, those quotes are briefly wrong, and everyone faster than you can trade against them at prices you no longer want. Being slow does not merely mean missing opportunities; it means being systematically picked off.
This reframes the arms race. Firms invest in speed not primarily to be predatory but to avoid being prey. A market maker who cannot cancel quickly must compensate by quoting wider, which costs them flow.
Speed is mostly defensive. The main return on latency investment for a market maker is fewer fills on quotes that are already wrong.
Where the latency budget goes
A round trip decomposes into several parts, and firms optimise each:
Market data processing. Decoding the exchange feed. Often moved onto FPGAs, hardware programmed to parse packets directly rather than running software.
Decision logic. Your strategy computing a response. Simple logic can also live in hardware; complex logic must run in software and costs more time.
Order transmission. Getting the message to the matching engine, which is where co-location matters.
Exchange processing. The engine's own time, which you cannot control but can measure and account for.
The physical layer
Co-location places your servers in the same data centre as the exchange's matching engine. Exchanges sell this, along with the fairness guarantee that every cabinet has an equal-length cable to the engine.
Microwave and millimetre-wave links carry data between financial centres faster than fibre, because light travels roughly 30% slower in glass than in air, and because the microwave route is a straighter line. The Chicago to New Jersey path is the famous example. The tradeoff is bandwidth and weather sensitivity: microwave links degrade in heavy rain, so firms maintain fibre as backup.
Kernel bypass lets an application talk to the network card directly, skipping operating system overhead worth several microseconds.
The economics
This is expensive and the advantage is temporary. Every improvement is matched by competitors, so the winner is whoever spent most recently rather than whoever is fastest in absolute terms. Some describe it as a race with no finish line, and there is a reasonable argument that the social value of the last few microseconds is small.
The practical consequence is that speed alone is a hard business to sustain. Most successful firms combine adequate speed with something more durable: better pricing models, better risk management, or access to flow.
"Is speed the only edge?" is a common interview question. A strong answer notes that speed is necessary to avoid adverse selection but rarely sufficient, and that it depreciates faster than pricing or risk skill.
Where it does not matter
Not every strategy needs microseconds. A relative-value trade held for days is unaffected by a millisecond, and an options market maker's edge lies mostly in pricing volatility correctly rather than in reacting first. Match your infrastructure spend to your holding period, not to the industry's headline numbers.