Eight Microseconds. How the Speed of Data Transmission Affects the Price You Pay for Everything

A $300 Million Cable. For Eight Milliseconds.

2009. The company Spread Networks begins construction.

The task — lay a fiber-optic cable from Chicago to New Jersey. A distance of roughly 1,300 kilometers. The cable has to be as straight as physically possible, given the terrain, rivers, and private property. In some places, crews drill through mountains. In others, they buy right-of-way from landowners. Elsewhere, they negotiate with city governments.

The project’s budget — $300 million.

For what? For 8 milliseconds.

The previous route between Chicago and New Jersey took about 17 milliseconds. The new, straighter cable — about 8.

The company had no trouble finding buyers. Trading firms lined up to lease access to this cable. Because 8 milliseconds is a fortune.

This is the story of how the speed of data transmission became the most expensive commodity on the planet. And why it concerns not just traders. But everyone.

What Dark Fiber Is

In the 1990s, at the height of the internet boom, telecom companies laid cable with enormous surplus capacity.

The logic was simple. The most expensive part of laying cable — up to 90% of the cost — is the physical labor. Chartering ships for undersea cables. Digging trenches through cities. Obtaining permits. Negotiating with landowners.

The cable itself — glass fiber — is relatively cheap. So operators laid 10 to 100 times more fiber into the trenches than the market needed at the time. The logic: if more capacity is ever needed, no one wants to dig again.

When the dot-com bubble burst in 2001, thousands of kilometers of this fiber sat unconnected. No lasers on the ends. No equipment. Just glass, in the ground and underwater.

That’s dark fiber.

For decades it sat as an asset on the books of bankrupt companies. Sold for pennies at auctions. Then someone realized exactly what it was worth.

Who HFT Firms Are, and What They Need

HFT — High Frequency Trading.

This is trading financial instruments using algorithms that make decisions and execute trades faster than a human can blink.

How fast? According to Investopedia, one-hundredth of a microsecond is enough for most decisions and the execution of a typical HFT strategy. A microsecond is one millionth of a second. One-hundredth of that is one hundred-millionth of a second.

For comparison: a nerve signal travels from the eye to the brain in roughly 100–150 milliseconds. In the time it takes you to see a price on a screen, an HFT algorithm has already completed millions of operations.

The global HFT market was valued at $10.4 billion in 2024. That’s not trading volume — that’s the revenue of the HFT firms themselves.

What do they need to operate? Three things.

First — proximity to exchange servers. This is called co-location — placing your own servers in the same data center as the exchange’s servers. The distance from an HFT firm’s server to the exchange’s server is literally meters of cable. That gives a microsecond advantage over anyone farther away.

Second — direct access to market data. Exchanges sell special data feeds — streams of price and order data that arrive several milliseconds ahead of public data. HFT firms pay hundreds of thousands of dollars a year for this.

Third — dark fiber between exchanges. The Chicago Mercantile Exchange (CME) and the New York exchanges — NYSE and NASDAQ — are the two main hubs of the American market. If a futures price changes in Chicago, that information has to reach New York before competitors see it. This is exactly why Spread Networks laid that $300 million cable.

How It Works. A Simple Example.

Your pension fund decides to buy 100,000 shares of Microsoft.

The fund manager enters the order. The order is sent to the exchange. A few milliseconds until execution.

In those few milliseconds, an HFT algorithm notices the signs of a large buyer. It doesn’t know this is specifically your pension fund — it sees a pattern in the order flow. Within microseconds, the algorithm buys up the available Microsoft shares at the current price — say, $42.00. Then it sells them to your pension fund — at $42.05.

Five cents’ difference. On 100,000 shares — $5,000.

This is called latency arbitrage. The algorithm doesn’t know any better than anyone else where the price is headed tomorrow. It simply sees a large buyer’s intent a few microseconds earlier, and manages to insert itself between buyer and seller.

Research from the University of Chicago’s business school estimates that, in typical years, HFT firms earn roughly $5 billion from other market participants through strategies like this.

Who are these other participants? Pension funds. Mutual funds. Institutional investors managing the money of tens of millions of ordinary people.

The Book That Changed Everything. And Didn’t.

2014. Michael Lewis publishes Flash Boys.

Lewis — author of The Big Short about the 2008 crisis, and Moneyball about statistics in baseball. A writer read by more than just finance people.

Flash Boys tells the story of Canadian banker Brad Katsuyama, who worked at RBC and discovered a strange pattern. Every time he tried to buy shares, the price on his screen was one thing, but the actual execution price came in higher. As if someone saw his intent and managed to get ahead of him.

Katsuyama figured out the mechanism. Drew out the scheme on paper. The scheme described latency arbitrage exactly as in the example above.

The book came out — and created a scandal. In interviews, Lewis declared that “the market is rigged so that only the fastest win.” That this was a form of front-running — using knowledge of others’ intentions for personal gain. That ordinary investors pay a hidden tax every time their pension fund executes a trade.

The HFT industry fired back. Part of the argument had real weight.

Before HFT existed, bid-ask spreads were significantly wider. Market makers — the people who provided liquidity — charged more for their services. HFT made markets cheaper for small retail investors in terms of spreads and commissions.

That’s true. A retail investor buying 100 shares through a brokerage app today pays less than they would have in 2000.

But that doesn’t contradict the fact that a pension fund buying 100,000 shares pays more than it would without HFT. Because the scale of the trade makes it visible to the algorithms.

Two truths that exist simultaneously.

Microwaves. When Fiber Optics Became Too Slow.

The story doesn’t end there.

After Spread Networks laid its $300 million cable — and after dozens of HFT firms paid for access to it — a problem emerged.

Light travels slower in fiber-optic glass than in a vacuum. Glass has a refractive index of about 1.5. That means light travels through fiber at roughly 200,000 kilometers per second instead of 300,000 in a vacuum.

And a straight line between Chicago and New Jersey is shorter than any cable route. Because a cable can’t fly through the air.

The solution: microwave towers.

Companies Jump Trading and McKay Brothers built networks of microwave relay towers — towers that transmit signals through the air. Radio waves in the atmosphere travel at a speed close to the speed of light in a vacuum. A straight-line route is shorter than any cable route.

The result: signal transmission time from Chicago to New Jersey over microwave networks — about 8.5 milliseconds. Over the best fiber cable — about 13. A difference of several more milliseconds.

Firms pay tens of millions of dollars a year for this competitive edge.

Then another solution appeared. Hollow-core fiber. Fiber whose core isn’t glass but air or a vacuum. Light travels through it significantly faster than through standard fiber — close to the speed of light in a vacuum. EU Networks in the UK deployed such a connection for HFT clients.

The race for microseconds continues. Every time one technology hits its physical limits, the next one is found.

The Flash Crash. When Algorithms Devoured a Trillion Dollars in 36 Minutes.

May 6, 2010. An ordinary trading day, against the backdrop of the Greek debt crisis. Markets are nervous.

2:32 PM Eastern time. A large mutual fund begins selling E-Mini S&P 500 futures. A large volume — 75,000 contracts worth $4.1 billion. The selling algorithm is set for speed, not price — sell fast, regardless of what happens to the price.

Selling begins. HFT algorithms spot the large seller. The algorithm’s logic: if someone big is selling, sell too. Algorithms start selling. This pushes the price down further. Other algorithms see the price falling — they sell. A chain reaction.

Over 36 minutes, the Dow Jones fell nearly 1,000 points. A trillion dollars in market capitalization evaporated.

Procter & Gamble lost 36% of its value in a matter of minutes. Accenture shares traded at one cent. Literally.

Then the algorithms stopped. Prices recovered. Within two hours, everything was almost back to where it started.

What happened to the HFT firms during the crash? A study by the CFTC and academic researchers found: HFT firms didn’t lose money that day. On the contrary — they earned more than usual. While pension funds and traditional market participants took losses, the algorithms extracted profit from the chaos they themselves had amplified.

After the 2010 flash crash, smaller-scale flash crashes started appearing — in individual stocks. Google, Symantec, Anadarko. Shares lost tens of percent within seconds — and recovered. Researchers have documented that events like this occur regularly.

A market that operates faster than a human can perceive is a market that can move in ways humans can’t keep up with controlling.

The Connection to the Data Collected About You

Here’s where a question arises that ties HFT to the subjects of the previous articles.

HFT algorithms run on data. Their advantage isn’t just signal transmission speed. It’s also the quality of the information they work with.

Exchanges sell so-called proprietary feeds — streams of order and trade data richer and more detailed than the public data. In these feeds, patterns are visible — who’s buying, in what increments, at what intervals.

But that’s just the beginning.

HFT firms buy alternative data — anything that can provide an edge over the market by a few seconds or minutes.

Traffic data from mobile apps. If people are flocking to a particular retail chain’s stores en masse, that’s a sales signal before the company even reports earnings. Location data from phones — aggregated and anonymized — gets sold as financial data.

Transaction data. Aggregated credit-card purchase data reveals consumer trends faster than official statistics.

Text-sentiment data. Algorithms analyze millions of news articles, social-media posts, and company earnings-call transcripts — and trade on them seconds before an ordinary investor has even read the headline.

Satellite-imagery data. The number of cars in shopping-mall or oil-rig parking lots is a signal purchased before official statistics come out.

The connection is direct. The data you hand over to apps — your location, purchase history, online behavior — gets aggregated and sold as financial data. Those who trade fastest are, in part, trading on your behavioral patterns.

You didn’t know you were part of this system. But you are.

What Happened Next. IEX and the Attempt to Slow the Race Down.

Brad Katsuyama of Flash Boys didn’t stop at identifying the problem.

He founded IEX — the Investors Exchange. An exchange with a deliberate delay.

The idea is simple and elegant. IEX adds 350 microseconds of delay to all incoming orders. Physically, this is a coil of fiber-optic cable roughly 60 kilometers long, wound up in a cabinet. The signal passes through it and loses time.

350 microseconds is enough to strip HFT algorithms of their main advantage. They see the order and try to react — but the delay prevents them from getting ahead of the large buyer in time.

IEX received exchange status in 2016. Goldman Sachs and other major institutional players came on board. Because for pension funds trading in large volumes, execution on IEX turned out to be more favorable.

This didn’t solve the problem globally. Other exchanges continue operating without a delay. HFT firms adapted. The race continues.

But IEX proved one important thing. The market doesn’t have to be structured this way. It’s an architectural choice. It can be changed.

Italy, Which Taxed Microseconds

In 2013, Italy became the first country in the world to introduce a special tax on high-frequency trading.

The rate — 0.02% on stock transactions lasting less than 0.5 seconds.

Not a large sum. But it’s a principle. The state acknowledged that there’s a separate category of trading activity — too fast to be ordinary trading — and that it can be a subject of regulation.

Europe’s regulator, ESMA, introduced requirements for algorithmic traders — algorithm testing, limits on the number of order cancellations, system-resilience requirements.

France introduced a similar tax on HFT. The UK debated it. Germany introduced registration requirements for HFT firms.

This is movement. Slow. With no guarantees. But movement nonetheless.

What Dark Fiber Means for You

Let’s return to the beginning. A $300 million cable, for 8 milliseconds.

You’ll never see that cable. It’s in the ground. You never interact with an HFT firm directly. They trade on exchanges — not with you personally.

But you’re part of this system. Every time your pension fund executes a trade. Every time the mutual fund holding your savings buys or sells shares. Researchers estimate other market participants’ cumulative losses to HFT strategies at roughly $5 billion a year.

$5 billion that shifts from pension funds and institutional investors — to algorithms.

This isn’t a catastrophe. Markets function. Pensions get paid. But it’s a hidden tax built into the structure of the financial system — invisible, unwritten, officially nonexistent.

And it became possible because one hundred-thousandth of a second acquired monetary value. Because dark fiber buried in the ground was sold to whoever found a use for that value. Because exchanges decided to sell data first to whoever pays the most.

All of this is architectural choice. Not natural law.

One Last Thing. About the Invisible Layer Beneath the Visible One.

We’re used to thinking of financial markets as a place where buyers and sellers meet. Where prices form from real supply and demand.

That’s true — on one level.

On another level — beneath it — there’s a race for microseconds. Algorithms trading faster than human perception. Dark fiber worth a fortune. Microwave towers transmitting signals through the air because light in glass is too slow. Data about your behavior, aggregated and sold as financial signals.

This layer is invisible to the ordinary market participant. But it determines the prices at which trades execute. And prices in financial markets determine the cost of capital. The cost of capital determines the cost of credit. The cost of credit determines the cost of a mortgage. The cost of a mortgage determines the price of an apartment.

Eight milliseconds in a cable underground. The price of your apartment.

This isn’t a direct dependency. It’s a complex system in which the speed of data transmission has become one of the determining factors at every level.

Understanding how this layer works isn’t a duty for every citizen. But it’s knowledge that changes how you hear the phrase “the market decided.”

The market doesn’t decide. The algorithm decides. In eight microseconds. Over fiber lying beneath your feet.