The Speed of Light Is Too Slow: Inside Humanity's War Against Latency

Key Takeaway

Light travels at 299,792 km/s in a vacuum - the fastest anything can move in our universe. It is still not fast enough. Fiber optic glass slows light by 31%. Submarine cables add thousands of extra kilometers of routing. And between Earth and Mars, a single Google search takes up to 48 minutes round-trip. Every millisecond of delay shapes what technology can and cannot do - from remote surgery to financial trading to gaming to interplanetary colonization. This is the story of humanity’s oldest engineering battle: the fight against the time it takes information to travel from one place to another.

850,000 Times Faster and Still Not Enough

On August 16, 1858, Queen Victoria sent a 98-word congratulatory message to US President James Buchanan over the first transatlantic telegraph cable. The transmission took 16 hours and 30 minutes. The cable itself failed after six weeks, destroyed by an overzealous engineer who cranked the voltage too high trying to speed things up.

Today, a packet of data crosses the Atlantic in about 28 milliseconds. That is approximately 850,000 times faster than Queen Victoria’s telegraph. And for a growing number of applications - surgery, trading, gaming, military operations - it is still too slow.

The speed of light is not an engineering constraint. It is a law of physics. You cannot innovate around it, optimize past it, or buy your way out of it. Every improvement in communication speed over the past 170 years has been about removing obstacles between the signal and that absolute limit - switching from copper to glass, from glass to vacuum, from indirect routes to straight lines.

The history of telecommunications is the history of getting closer to an unreachable ceiling.

The Refractive Index Trap

Modern fiber optic cables are engineering marvels. A single strand of glass thinner than a human hair carries terabits of data per second using pulses of infrared light. The global internet backbone runs on roughly 550 submarine fiber cables stretching over 1.4 million kilometers across ocean floors.

But there is a problem baked into the glass itself.

Light travels at 299,792 km/s in a vacuum. When it enters glass - specifically, the doped silica glass used in standard telecom fiber (Corning SMF-28, refractive index 1.4677 at 1550 nm) - it slows down. Considerably.

Speed of light in fiber: approximately 204,190 km/s. That is 31.9% slower than vacuum. Roughly one-third of the speed of light is sacrificed the moment a signal enters a fiber optic cable.

This is not a manufacturing defect. It is physics. The refractive index of the glass medium determines how much light slows down when passing through it. No amount of engineering can change it without changing the medium itself.

But the refractive index is only half the problem.

The Routing Penalty

Submarine cables do not follow great circle routes - the shortest path between two points on a sphere. They route around continents, through straits, and along coastlines. They make landfall at multiple points to serve intermediate markets. They avoid earthquake zones, anchoring areas, and geopolitical hotspots (though not always successfully - the Red Sea cables are a persistent vulnerability).

The result: the physical path a signal travels through fiber is often far longer than the straight-line distance between endpoints.

Route	Great Circle Distance	Typical Cable Length	Excess
New York to London	5,555 km	~6,000 km (Hibernia Express)	8%
New York to London	5,555 km	15,428 km (TAT-14, multi-landing)	178%
Singapore to Marseille	10,583 km	20,000 km (SEA-ME-WE 5)	89%
Tokyo to Oregon	10,044 km	11,629 km (FASTER)	16%

Combine the 31% speed penalty from glass with a 16-178% distance penalty from routing, and you begin to understand why real-world internet latency is so far from the theoretical minimum.

Starlink’s Vacuum Shortcut

In 2018, Mark Handley at University College London published a paper titled “Delay is Not an Option: Low Latency Routing in Space.” His finding was counterintuitive and striking: a LEO satellite network with laser inter-satellite links could deliver lower latency than any possible terrestrial fiber network for distances greater than approximately 3,000 km.

The logic is straightforward once you see it.

Starlink satellites communicate with each other using laser links that transmit data through the vacuum of space. In vacuum, light travels at its full speed - 47% faster than in fiber. And satellite paths between distant cities can be more direct than submarine cable routes, which must follow ocean floors and coastlines.

For a New York to London connection:

• Best fiber route (Hibernia Express): 58.95 ms round-trip
• Starlink laser relay through vacuum: potentially 40-45 ms round-trip

For longer routes, the advantage compounds. A connection from London to Singapore via fiber must route through the Mediterranean, Suez Canal, Red Sea, Indian Ocean, and Strait of Malacca - roughly double the straight-line distance. A satellite relay can approximate the great circle path through space.

Handley’s research showed that LEO laser routing could improve latency for 78.76% of all city-pairs globally compared to the best possible fiber path.

By the end of 2023, Starlink had deployed over 9,000 laser inter-satellite links across its constellation, moving 42 petabytes of data per day through the vacuum of space. The theoretical advantage is becoming operational reality.

The $300 Million Straight Line

No industry understands the value of a millisecond better than high-frequency trading.

In 2010, Spread Networks completed a $300 million project: an 827-mile fiber optic cable running in the straightest possible line from the Chicago Mercantile Exchange in Aurora, Illinois to the NASDAQ data center in Carteret, New Jersey. Every existing fiber route between Chicago and New York followed railroad rights-of-way and highway corridors, with all their curves and detours. Spread Networks bored through mountains, trenched under rivers, and cut through private land to shave off every unnecessary kilometer.

The result: round-trip latency dropped from 17 milliseconds to 13.1 milliseconds. Later refinements brought it down to 12.98 ms.

Four milliseconds. That was worth $300 million to financial traders, because in high-frequency trading, the firm whose order arrives first captures the profit. Four milliseconds was the difference between making money and losing it on thousands of trades per day.

Then someone realized that microwaves travel through air at 99% the speed of light.

Within two years, proprietary microwave relay networks - chains of radio towers with line-of-sight connections across the landscape - had beaten Spread Networks’ fiber. Microwave latency on the Chicago-to-New-Jersey corridor dropped to roughly 8.0-8.5 milliseconds round-trip. The one-way advantage: 2.75 milliseconds over the fastest fiber.

Today, 15 to 20 private microwave networks operate between New York and Chicago, each costing approximately $8 million. Jump Trading purchased a decommissioned Belgian military tower once owned by NATO. Citadel Securities processes $652 billion in daily trades with execution times measured in microseconds. The global HFT market reached $10.36 billion in 2024.

Spread Networks was sold to Zayo Group in 2017 for $127 million - less than half its construction cost. The vacuum of air had beaten the glass of fiber.

When Milliseconds Save Lives

On September 7, 2001, surgeon Jacques Marescaux stood in a New York operating room and removed a gallbladder from a patient lying on a table in Strasbourg, France - 6,230 kilometers away. The procedure, known as Operation Lindbergh, was the first transatlantic telesurgery performed on a human patient.

The connection used redundant fiber optic lines provided by France Telecom, carrying a round-trip signal across approximately 14,000 km. The measured latency: 155 milliseconds. The surgery took 45 minutes. The patient recovered fully.

That 155 ms was carefully chosen. Research has established clear thresholds for telesurgery latency:

Below 200 ms, surgeons can operate with full precision - the delay is imperceptible to human motor control. Between 200 and 300 ms, surgeons begin consciously compensating for the lag. Above 700 ms, the delay becomes dangerous.

This creates a hard geographic boundary for telesurgery. At the speed of light in fiber, 200 ms of round-trip latency corresponds to a maximum distance of roughly 20,000 km - enough to cover most of the planet. But real-world routing, processing overhead, and network congestion eat into that budget. In practice, reliable telesurgery requires routes where the latency margin is comfortable, not razor-thin.

In July 2025, surgeons in Strasbourg performed the first transcontinental bariatric surgery on a patient in Indore, India - 8,500 km away - “without any perceptible lag.” A 2024 multicenter study across 37 patients at distances up to 5,000 km reported a 100% success rate. The technology is maturing fast.

But there is one place telesurgery will never work in real-time: Mars. At a minimum 3-minute one-way delay, a surgeon on Earth would see their patient’s vital signs change and be unable to respond for at least 6 minutes. Mars colonists will need their own surgeons - or surgical robots capable of autonomous operation.

The Gaming Ceiling

Competitive gaming has its own latency hierarchy, with a different set of thresholds:

Latency	Experience
0-15 ms	Professional esports level
15-50 ms	Competitive play
50-100 ms	Casual play
100-150 ms	Noticeable lag
150-300 ms	Unplayable for FPS

Research from Worcester Polytechnic Institute found that reducing network latency by 100 ms improves player accuracy by 2% and increases scoring by approximately 2 points per minute - equivalent to roughly one additional elimination per minute of gameplay. In a professional match where the prize pool can reach millions of dollars, that difference is enormous.

The esports industry generated $2.1 to $4.1 billion in revenue in 2024, projected to reach $7.5 billion by 2030. This entire industry exists within the constraints of latency. Professional tournaments enforce maximum ping limits. Game server locations are chosen based on distance to player populations. The netcode in every competitive multiplayer game is essentially a collection of clever tricks to hide or compensate for the delay that physics guarantees.

LEO satellite internet has already crossed the threshold for casual gaming. Starlink’s typical 25-35 ms latency is competitive with many terrestrial broadband connections. But for professional esports, where the difference between 5 ms and 15 ms matters, only local fiber connections provide the consistency required.

The Mars Problem

Everything discussed so far - the refractive index of glass, the routing of submarine cables, the microwave towers of Wall Street - takes place within a narrow band of latencies between 1 and 200 milliseconds. These are problems of optimization: getting closer to the speed-of-light limit over relatively short distances.

Mars is a different category of problem entirely.

At its closest approach, Mars is about 55 million km from Earth. Light takes 3.1 minutes to cross that distance. At its farthest, with the Sun between the two planets, the one-way delay stretches to 22.3 minutes. A round trip - sending a request and receiving a response - takes between 6.2 and 44.6 minutes.

This means:

• A web search takes 6 to 45 minutes to return results
• A video call is impossible - you would say something and wait up to 45 minutes for a response
• TCP/IP (the protocol that runs the internet) breaks completely, because it requires acknowledgment packets that would take minutes to arrive
• Every two years, when Earth and Mars are on opposite sides of the Sun (solar conjunction), all communication is blocked for roughly two weeks

No engineering can fix this. The speed of light is the speed of light. Mars is far away. These are not problems to be solved; they are constraints to be designed around.

NASA and internet co-inventor Vinton Cerf recognized this in the late 1990s and developed Delay-Tolerant Networking (DTN) - a “store and forward” protocol where data is packaged into self-contained bundles, stored at each node, and forwarded whenever a communication link becomes available. DTN has been operational on the International Space Station since 2016.

A Mars colony’s internet would look nothing like Earth’s. It would be a local network - satellites in Mars orbit, surface base stations, cached copies of Wikipedia, medical databases, and engineering references stored on local servers. New content from Earth would arrive in batches, hours old at best. Email would work. Streaming would not. Social media would operate on a delay measured in minutes, not milliseconds.

Mars internet would feel less like broadband and more like the postal service - reliable, eventually, but never real-time.

The Quantum False Promise

Whenever the speed-of-light problem comes up, someone mentions quantum entanglement. The reasoning goes: entangled particles affect each other instantly across any distance, so we should be able to use entanglement for faster-than-light communication.

This is wrong. And the reason it is wrong reveals something deep about the nature of information itself.

Quantum entanglement is real. When two particles are entangled, measuring one instantly determines the state of the other, regardless of distance. Einstein called it “spooky action at a distance” and was uncomfortable with it his entire life.

But here is the catch: when you measure an entangled particle, you get a random result. The other party also gets a random result. The results are perfectly correlated - but neither party can tell whether the other has measured yet, or what the other’s result was, without communicating through a classical (light-speed-limited) channel.

This is not a technical limitation waiting to be overcome. It is a mathematical theorem. The No-Communication Theorem, proven within the framework of quantum mechanics itself, demonstrates that it is impossible to transmit classical information by manipulating entangled quantum states. As Caltech’s quantum physics group puts it: “It may be tempting to think that the particles are somehow communicating with each other across these great distances, but that is not the case.”

What quantum networks actually do is Quantum Key Distribution (QKD) - using entangled particles to create encryption keys that are provably secure against eavesdropping. China has built a 2,000 km quantum communication backbone from Beijing to Shanghai for exactly this purpose. QKD is genuinely useful. But it operates at the speed of light or slower. It does not, and cannot, break the latency barrier.

The speed of light is not an engineering problem waiting for a quantum solution. It is a fundamental property of spacetime.

The Absolute Floor

Let us put a final number on the constraint.

Earth’s equatorial circumference is 40,075 km. At the speed of light in a vacuum, a signal takes 133.7 milliseconds to circumnavigate the planet. One-way to the antipodal point (the farthest place from you on Earth): 66.8 milliseconds.

In fiber optic glass, those numbers become 196.3 ms and 98.1 ms respectively.

No technology that transmits information as electromagnetic radiation - radio waves, microwaves, laser beams, light in fiber - can beat these numbers. They are the floor. Everything else is overhead.

Route	Distance	Minimum Latency (Vacuum)	Minimum Latency (Fiber)	Typical Real-World
Same city	50 km	0.17 ms	0.24 ms	1-5 ms
NY to Chicago	1,150 km	3.8 ms	5.6 ms	15-20 ms
NY to London	5,555 km	18.5 ms	27.2 ms	70-80 ms
NY to Sydney	16,000 km	53.4 ms	78.4 ms	200-250 ms
Antipodal (max)	20,038 km	66.8 ms	98.1 ms	300+ ms

Human perception research shows that interactions feel sluggish beyond 100-200 ms and become noticeably delayed beyond 300 ms. This means that for users on opposite sides of the planet, physics itself guarantees a perceptible delay. No future technology operating within known physics will change this.

The inescapable numbers: 66.8 ms one-way to the antipodal point in vacuum, 98.1 ms in fiber. Human perception turns sluggish beyond 100-200 ms. We have improved 850,000x since 1858 - and we are still bumping against the ceiling.

The War Goes On

We have improved transatlantic communication speed by a factor of 850,000 since Queen Victoria’s telegraph. We have gone from 16 hours to 28 milliseconds. And we are still fighting for every remaining fraction of a millisecond.

Wall Street builds microwave towers to save 5 ms. SpaceX launches thousands of satellites with laser links to beat fiber by 15-20 ms on long routes. Surgeons choose hospital locations partly based on network latency. Gamers choose internet providers based on single-digit ping differences. And NASA designs entirely new networking protocols because the 3-minute delay to Mars makes the existing internet fundamentally unusable.

The pattern is always the same: get closer to the speed of light, get a straighter path, remove every obstacle between signal and destination. The tools change - copper wire, glass fiber, microwave relay, satellite laser - but the enemy never does.

The speed of light is 299,792 km/s. It is the fastest anything can move in our universe. For an engineer trying to connect a surgeon’s hands to a patient 8,500 km away, or a trader’s algorithm to an exchange 1,300 km away, or a Mars colonist to a database on Earth - it is too slow.

And it always will be.