The Speed Race: HFT and the $300 Million Cable That Saved 1.4 Milliseconds | Inside the Machine EP.5

In 2010, a company called Spread Networks spent $300 million to dig a fibre-optic cable in a near-straight line from Chicago to New Jersey. Previous cables followed road corridors. Roads curve. Curves add distance. Distance adds milliseconds. Spread Networks cut through mountains and rivers to save those milliseconds. The saving: 1.4 milliseconds. High-frequency trading firms paid $14 million a month to use that cable.

This is Episode 5 of Inside the Machine: How Markets Really Work.

What Is High-Frequency Trading?

High-frequency trading encompasses several distinct strategies with very different implications for other market participants. Electronic market making — providing continuous bid and ask quotes in exchange for the spread — is the most common and most beneficial. Statistical arbitrage — identifying temporary price discrepancies between related instruments — improves price efficiency across markets. Both make markets better for most participants.

Latency arbitrage is more predatory: being faster than other participants to the same information and trading against their predictable reactions before they can act. Momentum ignition — placing and cancelling orders rapidly to create artificial price signals — is considered market manipulation and has been prosecuted. The policy debates around HFT collapse all four strategies into one category, producing confused conclusions. The question is not whether HFT is good or bad overall, but which specific strategies create genuine market value and which extract it.

Co-location: The Speed Hierarchy

Co-location is the physical infrastructure behind the speed hierarchy. Exchanges openly sell a service where trading firms place their servers inside the exchange’s own data centres — the same buildings housing the matching engines. A co-located server connects to the matching engine with feet of cable. A retail platform connects through hundreds of miles of network infrastructure.

A co-located server receives market data in under a microsecond. A retail platform displays the same price change 50 milliseconds later. In those 50 milliseconds, a co-located algorithm has executed hundreds of trades, hedged positions, and updated quotes across dozens of instruments. Co-location is publicly available and openly priced — but the economic logic only applies to strategies operating at microsecond speeds, which retail traders do not and should not attempt.

After Spread Networks’ fibre cable, the race moved to microwave. Signals travel through air at the speed of light — approximately 3% faster than through glass fibre. Jump Trading and several other firms built microwave tower networks on the Chicago-to-New-Jersey route, cutting latency from roughly 14 milliseconds (fibre) to 8.5 milliseconds. The cable Spread Networks spent $300 million building was commercially obsolete within a few years.

The Flash Crash: When HFT Liquidity Withdrew

On May 6, 2010, HFT market-making firms collectively withdrew their quotes as volatility exceeded their operational parameters. Each firm’s risk system independently concluded that conditions were too uncertain to continue quoting. Because they all used similar risk frameworks responding to the same signals, they stopped at approximately the same time.

Order books that had appeared deep became nearly empty in milliseconds. Not because sellers overwhelmed them — because the buyers turned off. The Flash Crash was not caused by malicious HFT activity. It was caused by the concentration of liquidity provision in systems that all respond to stress signals in similar ways at similar speeds — a structural vulnerability that electronic market design has not fully resolved.

What This Means for Your Trading

Do not compete on speed. Any strategy requiring action on a price signal in under a second is permanently structurally disadvantaged against co-located algorithms. This is not a skills gap — it is a physics gap. The appropriate response is a different timeframe, not faster hardware.

Trade at timeframes where HFT has no edge. HFT models are not built to operate at the four-hour or daily timeframe. A swing trade held for hours or days competes against human analysis and fundamental judgment — where your research, pattern recognition, and method discipline are genuine advantages.

Recognise HFT crowding patterns. When multiple momentum-following algorithms detect the same signal simultaneously, they pile in together, creating abnormally fast moves that often reverse sharply. The fingerprint: extreme velocity relative to typical candles, compressed time for a large move, clean reversal. Distinguishing machine-driven momentum from genuine directional conviction is a method skill that starts with understanding the mechanism.

Frequently Asked Questions

Is high-frequency trading legal?

Yes, most HFT strategies are legal. Electronic market making, statistical arbitrage, and co-location-based speed advantages are all legal and publicly disclosed. Momentum ignition — creating artificial price signals to manipulate other algorithms — has been charged as market manipulation in several cases. The distinction is between strategies providing genuine market function and those distorting prices to extract value without contributing liquidity.

Does co-location give HFT firms an unfair advantage?

Co-location creates a speed hierarchy that advantages firms with resources to pay for proximity to matching engines. It is sold openly by exchanges to any firm willing to pay. At microsecond trading speeds, the advantage is decisive. At multi-hour or daily trading timeframes, co-location provides zero advantage — retail traders and HFT firms are operating in entirely different competitive environments.

How did HFT contribute to the Flash Crash of 2010?

HFT contributed through collective withdrawal of liquidity. As a large selling cascade intensified, HFT market-making firms’ risk models simultaneously identified conditions exceeding their operational parameters. Each firm stopped quoting — rational individually, catastrophic collectively. Order books emptied, causing prices to gap lower on cascading sell orders with no bids available to catch them. The pause that followed restored conditions and prices recovered.

What replaced the Spread Networks cable?

Microwave transmission networks replaced fibre-optic cables as the primary low-latency route between Chicago and New Jersey. Microwave signals travel through air approximately 3% faster than light through glass fibre. Jump Trading and other firms built relay tower networks reducing latency from approximately 14 milliseconds (fibre) to 8.5 milliseconds (microwave). More recently, laser transmission networks have been developed for conditions where microwave is impractical.

Can retail traders benefit from understanding HFT?

Yes, in two practical ways. First, by choosing timeframes where HFT has no competitive advantage — multi-hour and daily analysis is outside the operational range of any HFT strategy. Second, by recognising HFT crowding patterns: fast, high-velocity moves generated by simultaneous algorithmic entries that often reverse sharply when the crowd unwinds. Understanding these patterns helps distinguish machine-generated momentum from fundamental trend continuation.