Latency Arms Race: Venue Design To Physics Limits
Sources: 1 • Confidence: Medium • Updated: 2026-04-11 19:06
Key takeaways
- High-frequency trading has pushed trading speed faster and faster over time as a dominant trend in electronic markets.
- Market-making HFT provides liquidity by posting non-immediately-executable bids and offers, while liquidity-taking HFT trades against existing orders when opportunities arise.
- The episode raises a normative question about whether society benefits from micro-latency investments such as being meters closer to servers.
- MacKenzie is starting a research project on AI scaling focused on how far actors proceed along diminishing-returns scaling curves given infrastructure spending and carbon costs.
- Many market participants do not understand end-to-end what happens between placing an order and its execution in modern markets.
Sections
Latency Arms Race: Venue Design To Physics Limits
- High-frequency trading has pushed trading speed faster and faster over time as a dominant trend in electronic markets.
- Latency competition in HFT has included physical proximity and infrastructure arms races, including co-location and the 'wire wars'.
- Island's distinctive feature was a very fast electronic order-book matching engine that automated execution by matching bids and offers without direct human negotiation.
- Automated trading and fast electronic venues reinforced each other by attracting liquidity to fast venues, which pressured incumbent exchanges to reorganize technologically.
- Island reduced execution time from seconds to roughly two milliseconds in the late 1990s.
- By roughly 2018–2020, nanosecond-scale latency became practically relevant in high-frequency trading.
Cross-Market Stale-Quote Race As The Core Micro-Mechanism
- Market-making HFT provides liquidity by posting non-immediately-executable bids and offers, while liquidity-taking HFT trades against existing orders when opportunities arise.
- A core HFT speed arms race arises when index-futures price moves in Chicago quickly render equity quotes in New Jersey stale, prompting makers to cancel while takers race to execute against stale orders.
- The maker-cancel versus taker-execute race can occur on nanosecond timescales.
- Eric Sufert and colleagues have estimated that profits from exploiting structural cross-market relationships such as the futures–equities linkage are meaningful but on the order of single-digit billions of dollars.
- Much liquidity taking in practice occurs between HFT firms because many quotes are placed by HFT market makers and many executions against them are by HFT liquidity takers.
Disputes, Welfare Questions, And Unresolved Systemic-Risk Effects
- The episode raises a normative question about whether society benefits from micro-latency investments such as being meters closer to servers.
- It is unresolved in the episode whether electronic and high-frequency trading create feedback loops that increase market volatility.
- Accusations between HFT liquidity takers and makers are described as mutual blame, with each side pointing to the other as responsible for problematic behavior.
- High-frequency trading is described as being discussed less frequently today than during the post-2008 period when it was a prominent target of blame.
Ai Scaling As An Analogous Arms Race With Diminishing Returns Uncertainty
- MacKenzie is starting a research project on AI scaling focused on how far actors proceed along diminishing-returns scaling curves given infrastructure spending and carbon costs.
- It is unresolved whether AI industry actors will stop on a diminishing-returns curve or persist due to beliefs that qualitative breakthroughs like AGI emerge past a threshold.
- The episode claims both HFT and AI exhibit arms-race dynamics in which firms continue investing despite worsening marginal economics because falling behind can be existential at the firm level.
- Sam Altman has described intelligence as roughly logarithmic in the resources devoted to training and running AI systems, implying diminishing returns to additional compute and spending.
Market Microstructure Opacity And Regulatory-Driven Complexity
- Many market participants do not understand end-to-end what happens between placing an order and its execution in modern markets.
- In modern equity markets, a broker routes an order into an exchange order book, where it either matches immediately and executes or rests until canceled or matched later.
- Equities trading workflows have become more complicated over time partly due to regulatory and market-structure changes such as Regulation NMS.
Watchlist
- It is unresolved in the episode whether electronic and high-frequency trading create feedback loops that increase market volatility.
- MacKenzie is starting a research project on AI scaling focused on how far actors proceed along diminishing-returns scaling curves given infrastructure spending and carbon costs.
Unknowns
- Do electronic and high-frequency trading strategies create net positive feedback loops that materially increase volatility, and under what conditions?
- How large is the total addressable profit pool for cross-market latency arbitrage today, and how is it changing over time?
- What fraction of the observed latency arms race spending produces measurable improvements in end-user outcomes (e.g., spreads, depth, execution quality) versus primarily shifting rents among intermediaries?
- To what extent did Regulation NMS specifically increase fragmentation/routing complexity versus other contemporaneous drivers?
- What is the empirical basis and scope of the 'logarithmic intelligence in resources' claim for AI systems, and how stable is it across model generations and evaluation sets?