Subscribe
Share
Share
Embed
Living Life in the Digital Age - Living Life with Technology past, present and future. Find me at https://www.lifeat100Mhz.com - Produced by Dave Bunyard - earthdrifter@gmail.com Cover Photo: Charles Sharp Crawford in a Cole car at the Indianapolis speedway in 1910.
In this episode of life at 100 megahertz, what is a quantum computer? How do they put a data center in the ocean? Is a data center in space a good idea or not? Can we put these quantum computers in the ocean or space?
Sam:Okay. So let's unpack this. Because right now, somewhere off the muddy turbulent coast of Shanghai, there's this massive 24 megawatt artificial intelligence brain sitting entirely submerged on the dark ocean floor.
Delilah:Just silently boiling the salt water right around its middle hull.
Sam:Exactly. Silently boiling the water. And at the same time, right, a thousand miles above your head, up in the freezing vacuum of low Earth orbit, there are these fleets of disposable data center satellites. And they're being engineered with the explicit planned intention of, you know, intentionally burning themselves up in the Earth's atmosphere in just a few short years.
Delilah:It's a completely wild reality.
Sam:It really is. Yeah. Because we like to think of the Internet or, you know, the cloud that you are connected to right now listening to this deep dive as something weightless, something invisible.
Delilah:Yeah. It's, it's the ultimate modern illusion. We've spent the last two decades building up this incredibly comforting narrative that our digital lives are ethereal.
Sam:Right. Like magic.
Delilah:Right. Exactly. When you tap a screen to back up your photos or you ask an AI to rewrite an email or you're just streaming a movie, you feel like the data just vanishes into the ether. It's categorized in our minds as existing somewhere else, you know, completely disconnected from the messy heavy resource draining physical world.
Sam:But the reality of the physical infrastructure that's actually required to keep your invisible cloud floating, it's violently crashing into the limits of physics. So today, we're taking this massive sack of sources, research papers, deployment roadmaps, and we're taking you on a journey into the absolute physical extremes humanity's engineering right now just to power the next generation of computing.
Delilah:It's a huge shift.
Sam:We're talking about hardware, fundamentally redesigned to operate at temperatures colder than deep outer space, server farms just abandoned to the crushing pressures of the seabed, and like we said, orbital computing architectures that treat billion dollar spacecraft as temporary disposable commodities.
Delilah:And, you know, if you're wondering why you should care about the physical geography of a server rack, you really need to understand that the invisible cloud you rely on every single day is hitting a brutal physical wall.
Sam:A massive wall.
Delilah:The global computational grid is running out of power. It's running out of space. And the sheer thermal output of modern artificial intelligence is forcing engineers to pull off solutions that are quite literally reshaping the Earth's oceans and our orbital space.
Sam:I mean, just look at the background behind you right now.
Delilah:Yeah. For those of you joining us visually on this deep dive, you can see my usual library backdrop is sort of dissolving into this highly intricate gleaming blueprint schematic of a dilution refrigerator.
Sam:And it's slowly morphing into a sprawling star chart. It looks incredible.
Delilah:Thanks. It really represents the bleeding edge of where our civilization's physical brain is migrating.
Sam:So I wanna start by looking at how the fundamental nature of the machine itself is changing. Because before we can talk about the extreme environments where we're hiding these computers, we have to talk about the fact that we're well, we're moving away from the architecture that has defined computing since the mid twentieth century.
Delilah:Right. The transition from classical computing to quantum computing.
Sam:Exactly. And assuming you already know that classical computers like the phone in your pocket or those massive supercomputers running weather simulations, they operate on silicon chips using billions of microscopic transistors. And they just act like binary light switches. Right? Flipping between zero and one.
Delilah:Yeah. It's the rigid deterministic world of binary logic. A switch is either entirely off or entirely on. There is absolutely no ambiguity.
Sam:But a quantum computer throws that entire deterministic rule book right out the window. It replaces the classical bit with a qubit or quantum bit. And because qubits operate under the wildly counterintuitive rules of subatomic quantum mechanics, they have physical properties that silicon could never replicate.
Delilah:Not in a million years.
Sam:Right. And the first of these core quantum powers is superposition, which I wanna visualize this for you because it is notoriously difficult to grasp.
Delilah:It really defies human logic.
Sam:It does. Think of a standard classical bit as a coin lying completely flat on a table. It is decisively either heads or tails, a zero or a one. But a quibbit operating in superposition is like a coin that is rapidly spinning on that same table.
Delilah:I've always really appreciated that specific visualization because it emphasizes the instability, you know, and the motion that's inherent in the subatomic world.
Sam:Right. While that coin is spinning it's a fluid undefined blur. It's mathematically both heads and tails at the exact same time or like any fractional probabilistic combination of both. It only chooses a definite state. It only falls flat as a zero or a one when the calculation is complete and the machine physically stops to measure it.
Delilah:And what's fascinating here is how that specific fluid property scales when you introduce the second core quantum power, which is entanglement.
Sam:The spooky action at a distance.
Delilah:Exactly. Albert Einstein famously and somewhat dismissively called it spooky action at a distance. It's this cosmic, deeply mysterious link where two or more qubits become fundamentally bound to one another.
Sam:Even across vast distances. Right?
Delilah:Yes. When they're perfectly entangled, the physical state of one qubit instantly dictates the state of its partner regardless of whether they're sitting micrometers apart on a microchip or on opposite sides of the known universe.
Sam:And this changes the math of computational scaling entirely, doesn't it? Because Yeah. Classical scaling is brutally linear.
Delilah:Oh, it is entirely linear. If you want a classical computer to be twice as powerful, you essentially just have to give it twice as many transistors. Two bits, three bits, four bits, it just stacks.
Sam:Right.
Delilah:But in a quantum system, because these qubits can be deeply entangled while they are suspended in that fluid spinning state of superposition, the processing power scales exponentially. One qubit represents two states at once. Two entangled quibbins represent four simultaneous states. Three represent eight.
Sam:And this gets out of hand really fast.
Delilah:It really does. If we connect this to the bigger picture, here's the exact data point from our sources that illustrates the sheer terrifying scale of this architecture. 300 perfectly entangled qubits can represent more simultaneous states than there are atoms in the observable universe.
Sam:Okay. Hold on. I really wanna pause and force you, the listener, to just let that statistic sink in for a second because the human brain is not equipped to process numbers like that.
Delilah:No. Not at all.
Sam:We aren't talking about a machine the size of Dyson sphere or a massive planetary supercomputer. We are talking about 300 individual microscopic atomic structures. 300 perfectly entangled qubits hold more simultaneous mathematical states than there are physical atoms in the entire observable universe.
Delilah:It's a staggering concept.
Sam:It completely shatters my intuition of how physical space and information actually relate to one another.
Delilah:Well, it shatters all of our intuitions. Right? Precisely because our brains evolved to survive in a classical physics environment where objects have permanent defined states. A rock is a rock. We're not designed to natively understand exponential quantum scaling.
Sam:Right. We just don't have the mental hardware for it.
Delilah:Exactly. But this incredibly vast landscape of possibilities introduces a massive engineering problem. Because if you have all these simultaneous states I mean, if your computer is holding more possibilities than there are atoms in the universe, how on earth do you actually find the right answer?
Sam:Because if you just pull a random measurement out of a universe of possibilities, finding the correct solution sounds mathematically impossible.
Delilah:It would be. And that brings us to the third core quantum power, the mechanism that actually makes the machine usable which is interference.
Sam:Okay. Let's unpack interference because if you just have a trillion wrong answers and one right answer floating in this quantum blur, that's, I mean, that's infinitely worse than finding a needle in a haystack. Yeah. You need a way to filter the noise.
Delilah:Right. To understand interference, think about the physical mechanics behind high end noise cancelling headphones. How do they actually function? They don't just put thick acoustic foam over your ears to muffle the sound.
Sam:Like the old construction earmuffs.
Delilah:Exactly. Not like that. They have tiny, highly sensitive microphones that actively sample the ambient background noise, say the low droning of an airplane engine. The microchip inside the headphones instantly calculates the exact shape of that sound wave and then the speaker generates a perfect inverted opposite sound wave.
Sam:So they clash?
Delilah:Yes. When the peak of the engine's sound wave physically meets the trough of the headphones artificial sound wave in the air, the two waves literally cancel each other out. That is destructive interference.
Sam:So the quantum computer is essentially doing phase cancellation but with abstract mathematics.
Delilah:That is precisely what a quantum algorithm is designed to do. It acts as a set of mathematical noise cancelling headphones for probability waves. The algorithm utilizes constructive interference to actively amplify the signals of the correct answers, making those specific probability waves larger and stronger.
Sam:And it crushes the bad ones.
Delilah:Simultaneously, yes. It uses destructive interference to cancel out and erase the millions of incorrect answers. The algorithm mathematically flattens the wrong answers down to a zero percent probability leaving only the correct solution standing prominently at the end of the calculation.
Sam:Okay. I wanna try a different analogy here to make sure we're really visualizing the practical operational difference between how a cutting edge classical AI tackles a problem versus how a quantum computer does it. Let's imagine you were trying to find your way out of a massive, incredibly complex, multilayered maze.
Delilah:Okay. I like this.
Sam:A classical computer, even a state of the art supercomputer solves that maze by acting like an incredibly fast, highly methodical mouse.
Delilah:Running down the corridors one by one.
Sam:Exactly. It runs down one path at the speed of light. It hits a dead end. It backs up, writes that dead end into its memory, and tries the next sequential path. Now it might do this millions or billions of times per second, so to our human eyes, it looks completely instantaneous.
Sam:But fundamentally, physically, it is still trying one singular path at a time.
Delilah:Right. It's linear at its core.
Sam:But a quantum computer acts entirely differently. It acts like a flood of water. It enters the maze and flows down every single path simultaneously. There's a need to try paths one at a time because its superposition allows it to explore all possible routes at once, instantly filling the entire volume of the maze and automatically highlighting the singular correct path to the exit.
Delilah:That is an excellent way to visualize the difference in architecture and it actually brings up a profound philosophical point highlighted in our source material. One of the theoretical physicists in the research noted that quantum mechanics is fundamentally the mechanics of possibilities. A classical machine deals in rigid facts, but a quantum computer maps out what physicists call wave functions.
Sam:So the machine isn't looking at hard empirical data points during the calculation at all?
Delilah:No. Not at all. In the classical physics that governs our daily lives, a switch is either on or off. A car is moving down the highway at a singular exact speed. But quantum mechanics dictates that at the subatomic level, nature operates entirely with probabilities and shifting potentials.
Sam:It's all just a blur.
Delilah:A literal blur. A subatomic particle literally exists in a hazy cloud of possible locations until it physically interacts with something that forces it to take a defined position. When you apply that fundamental truth of nature to computer science, the machine you build is quite literally an engine that processes uncollapsed possibilities. Our sources point out that if you have just 50 entangled qubits, that specific computer is holding 1.12 distinct possibilities in mathematical suspension simultaneously.
Sam:Wait. I'm struggling to see how that translates to raw speed though. Let me push back on this for a second. Does the quantum computer actually calculate faster than a regular computer, or does it just erase the wrong answer so effectively that it skips the calculation entirely?
Delilah:That distinction you just made is exactly the shift in perspective required to understand quantum computing. No. It does not calculate faster in a traditional linear sense.
Sam:It's not a faster mouse.
Delilah:It doesn't run the maze faster than the classical mouse. No. It completely bypasses the need for speed by choreographing possibilities. The entire job of the quantum algorithm is to carefully manage and manipulate the waves of probability during the calculation. Remember, while the machine is running, it exists in this fluid, unobserved, deeply isolated state where millions of answers are interacting as waves.
Delilah:The actual output, the final answer only happens at the very end when the machine is forced to take a measurement.
Sam:When the spinning coin is finally slapped down onto the table.
Delilah:Yes. In quantum physics, this moment is called collapsing the wave function. The exact millisecond the measurement occurs, that vast fluid cloud of a quadrillion possibilities instantly snaps into a single concrete classical reality. And because the algorithm utilized interference to perfectly choreograph the collapse beforehand, the single reality left standing is overwhelmingly likely to be the exact answer you were looking for.
Sam:That's wild.
Delilah:It doesn't find the answer by searching. It makes the correct answer an absolute inevitability.
Sam:Okay. So if a quantum computer is this magical flood of water that can instantly solve mazes and hold quadrillions of possibilities at once, the obvious question you must be asking yourself is, why aren't we running advanced video games on them right now? Why don't I have a quantum processor in my smartphone?
Delilah:Because of the thermodynamic nightmare.
Sam:Exactly. This brings us directly into the thermodynamic nightmare of quantum hardware. Because if you were to walk into a top tier quantum computing lab today, you wouldn't see a sleek minimalist desktop tower with glowing LED lights.
Delilah:No. Definitely not.
Sam:You would see what looks like a massive highly polished copper and gold industrial chandelier hanging from the ceiling. It looks like a piece of alien technology. It's almost steampunk.
Delilah:It really is beautiful. But the irony is that the massive gleaming chandelier structure is not actually the computer itself. That structure is called a dilution refrigerator.
Sam:Right. The actual quantum chip, the brain holding those quibbets, is incredibly small, and it sits at the very bottom tip of this massive chandelier. And to keep those incredibly fragile quibbets stable, to prevent them from accidentally bumping into each other or getting hit by stray ambient energy, which causes a catastrophic failure event called decoherence.
Delilah:Which ruins the whole calculation.
Sam:Right. To prevent that, that microscopic chip must be cooled to point zero one five Kelvin.
Delilah:Let's translate that scale for a moment because it's hard to grasp. Point zero one five Kelvin is roughly negative 459 degrees Fahrenheit. That is significantly colder than the ambient temperature of deep outer space.
Sam:That is just mind boggling. Colder than the vacuum of space, how do they even achieve that on Earth?
Delilah:By using really complex mixtures of liquid helium three and helium four isotopes pumped continuously through the layers of the chandelier. The entire rest of that massive beautiful copper structure is simply the thermal infrastructure required to pump those isotopes and painstakingly shield that tiny chip from any microscopic wave of heat or any stray electromagnetic interference or even the slightest physical vibration from the building.
Sam:Just vibration.
Delilah:Oh, absolutely. If a truck drives by outside the laboratory and the chandelier isn't properly isolated, the vibration alone can collapse the wave function and destroy the calculation.
Sam:Which completely vaporizes the myth of the home quantum computer. Because I constantly see people theorizing online, oh, computers used to be the size of an entire room in the nineteen sixties, and now they are in our pockets. Give it twenty years, engineers will miniaturize it, and I'll have a quantum laptop.
Delilah:Yeah. That is fundamentally misunderstanding the physics involved. A personal superconducting quantum computer is thermodynamically impossible for a consumer to maintain. And frankly, it would be practically useless to you even if you had a dilution refrigerator in your living room.
Sam:Useless? How can a machine holding a quadrillion possibilities be useless for a consumer?
Delilah:Because quantum computers are not simply faster versions of classical computers. They are entirely different computational tools designed for entirely different problems. For standard linear tasks like streaming a four k video, typing a text document, browsing the web, or rendering the graphics of a video game, quantum computers are actually terribly slow and incredibly cumbersome.
Sam:So they're terrible at daily stuff.
Delilah:Exactly. They are exclusively built for extreme optimization and massive multi variable mathematics. We are talking about molecular modeling, natively simulating the complex atomic chemistry of new life saving pharmaceuticals. We're talking about global logistics, instantly calculating the absolute most efficient shipping routes for thousands of cargo ships navigating unpredictable weather. Or cryptography factoring massive prime numbers to break advanced encryption protocols.
Sam:Let's look at the actual physics of that claim though because our sources do mention companies like SpinQ that are literally selling desktop sized quantum computers right now for educational Are those just fake? I've seen videos of them.
Delilah:No. They aren't fake, but it's vital to clarify what those machines actually are. Those are two to three qubit educational models. They do not use superconducting qubits that require deep space temperatures.
Sam:Okay. So how do they work?
Delilah:Instead, they operate at room temperature using a completely different, far less powerful physical mechanism known as nuclear magnetic resonance. They essentially use magnetic fields to manipulate the spin of molecules in a small vial of liquid. They are fantastic tools for teaching university students the basic mathematical concepts of quantum gates and superposition, but they cannot run real software or solve any utility scale computational problems.
Sam:So they're just demonstrators?
Delilah:Precisely. You will never simulate a new protein on a room temperature NMR desktop toy.
Sam:So if we aren't getting them in our homes, what is the actual timeline for this technology to impact your daily life? Because our sources lay out a very clear three phase road map for the deployment of utility scale quantum computing.
Delilah:Yes. And the crucial takeaway from this roadmap is that your interaction with quantum hardware will be entirely invisible and seamless. We are currently living in phase one, which the industry calls the cloud era. Anyone with a laptop and a basic understanding of Python can log in today to cloud portals run by IBM, Google, or Amazon Web Services and write experimental code that actually executes on a real physical quantum processor sitting in a freezing dilution refrigerator in a laboratory somewhere. You're simply renting fractional time on the machine.
Sam:And phase two is the hybrid consumer era, which the sources slate for the late twenty twenties through about 2035. This is when the apps on your smartphone will start invisibly offloading highly specific optimization tasks to a quantum back end.
Delilah:Right. Without you ever knowing.
Sam:Exactly. For example, if you are using an advanced financial app to dynamically rebalance a wildly complex multivariable investment portfolio based on real time global market fluctuations, your phone's classical silicon chip doesn't do the heavy math. It shoots the variables to a quantum cloud, the machine runs the wave function, it collapses into the optimal portfolio strategy, and the answer flashes back to your screen in seconds.
Delilah:Precisely. The classical chip in your phone acts as the manager, delegating the impossible math to the quantum coprocessor. And then we reach phase three projected for roughly 2035 to 2040 and beyond. This is the era of fault tolerance scale. This is the holy grail when quantum computers achieve millions of stable error corrected qubits.
Sam:This is the disease curing era.
Delilah:Exactly. This is the era where they fundamentally revolutionized material science, perhaps inventing next generation battery architectures or room temperature superconductors. But even in 2040, those massive physical machines will still be entirely locked away in highly secure freezing corporate or government data centers.
Sam:Now here's where it gets really interesting. You mentioned the quantum computers, despite their incredible power, draw a relatively tiny amount of electrical power. Our sources indicate a typical quantum setup draws roughly 25 kilowatts of power and almost all of that is purely to run the massive deep space refrigerator, not the chip itself.
Delilah:Which is practically nothing.
Sam:Right. Contrast that 25 kilowatts with classical AI data centers, which are currently devouring hundreds of megawatts of power.
Delilah:It is a stunning dichotomy. If quantum architecture represents the elegant, hyper efficient, mathematically precise future of computing, classical AI data centers represent the brutal, brute force, power hungry reality of the present.
Sam:There has been a viral fear sweeping through financial and tech circles lately. You've likely seen the headlines asking whether the massive build out of AI data centers will eventually bankrupt the global economy because the incredibly expensive hardware expires and becomes completely useless every three to four years.
Delilah:Yeah. This touches on the economic obsolescence of silicon, which is a very real, very terrifying dynamic for cloud providers. The concern isn't based on science fiction. The data center industry is currently experiencing a level of financial churn that is historically unprecedented. The economic useful life of a cutting edge AI chip like NVIDIA's flagship GPUs is brutally punishingly short.
Sam:But why is it so short? I'm struggling to see how a piece of solid state electronics degrades that fast. Do they just physically break down and melt from the intense heat they generate?
Delilah:No. They don't physically break or melt. The hardware is remarkably resilient. The issue is purely economic obsolescence driven by the insatiable demands of large language models.
Sam:So they just age out?
Delilah:They age out incredibly fast. The architectural efficiency of these GPUs advances at such a blistering exponential pace that a cutting edge cluster of chips you buy today for hundreds of millions of dollars will be completely uncompetitive and drastically inefficient to operate compared to the generation of chips released just thirty six months from now. The rental rates that cloud providers can charge developers for older GPUs drop off a cliff. Eventually, if you are operating a massive server farm, running the older chips actually costs you more electrical bills and cooling costs than they generate in computational value.
Sam:Hence, the fear of bankrupting the economy. I mean, tech giants have to replace tens of billions of dollars worth of infrastructure every three years just to stay relevant, That sounds like an unsustainable financial meat grinder.
Delilah:It is absolutely a meat grinder of capital, but we need to provide some historical and economic context on why this specific cycle likely won't trigger a massive collapse.
Sam:Okay. Let's hear it.
Delilah:During the dot com crash of two thousand, the telecommunications companies that were burning cash on fiber optic cables and servers were essentially pre revenue startups heavily leveraged on speculative high interest debt. Today, the company's driving this massive AI infrastructure build out Meta, Microsoft, Google, Amazon are sitting on the most lucrative business models in human history.
Sam:They just have the cash.
Delilah:They are generating astronomical amounts of free cash flow. They are funding this rapid hardware churn with cash on hand, not speculative debt.
Sam:Plus, to be fair to the engineering side, they aren't demolishing and rebuilding the entire physical building every three years.
Delilah:Exactly. The physical shell of the data center, the poured concrete, the massive high voltage grid hookups, the advanced liquid cooling manifolds, those infrastructural elements have a lifespan of fifteen to twenty years. When a specific generation of silicon chips becomes economically obsolete, technicians simply pull the old server blades out of the standardized racks and slide the new vastly more efficient blades in. The surrounding infrastructure remains intact.
Sam:But even with the physical shell remaining, the raw power consumption is exploding. These classical AI data centers are pulling hundreds of megawatts continuously. We are seeing reports of local municipal power grids straining under the load and tech companies delaying the retirement of coal plants just to feed the GPUs.
Delilah:It's a huge problem.
Sam:And this is where the synergy between the AI crisis and the quantum road map comes into play. Yeah. It is a fascinating two way street of technological salvation. Let's first look at how classical AI is actually saving quantum technology today.
Delilah:It is one of the most beautiful ironies in modern physics. We discussed earlier that building a functional quantum computer is incredibly difficult because qubits are fundamentally fragile. The slightest micro fluctuation in temperature or ambient vibration causes decoherence, destroying the wave function.
Sam:Right.
Delilah:Right now, classical machine learning, the very AI models devouring gigawatts of power is actively watching over these delicate qubits. The AI models are trained to spot the microscopic signals of a qubit beginning to drift or flip incorrectly. The AI calculates the error in real time and dynamically adjusts the system to mathematically fix the error before decoherence cascades and ruins the calculation.
Sam:That's wild.
Delilah:And the AI is also continuously calibrating the thousands of precise microwave pulses needed to control the qubits. Without the brute force of classical AI managing the noise, current quantum hardware would be practically unusable.
Sam:That's incredible. The power hungry brute is babysitting the fragile genius.
Delilah:That's a great way to put it.
Sam:But the street runs the other way for the future. How will quantum save AI tomorrow? Because we desperately need saving from this physical footprint explosion, we cannot continue building 100 megawatt data centers indefinitely.
Delilah:That salvation lies in the promise of quantum machine learning or QML. If we strip classical machine learning down to its pure mathematical essence, it is simply the process of taking a massive chaotic mess of unorganized data and searching for hidden patterns within a highly complex multi dimensional mathematical space.
Sam:Which is why it requires thousands of classful GPUs processing those matrices step by step, row by row for weeks on end, drawing massive amounts of power to move all that data around.
Delilah:Precisely. It is incredibly inefficient, but a quantum computer by its very nature computes using wave functions across thousands of dimensions simultaneously. So over the next ten to fifteen years as those fault tolerant quantum co processors are integrated into the hybrid cloud we discussed, they will be able to map this chaotic training data into multidimensional quantum states natively.
Sam:So it just skips all that linear processing?
Delilah:Exactly. They will navigate that mathematical space and find those hidden patterns using a tiny fraction of the computational steps and an unbelievably small fraction of the physical energy. A 25 kilowatt quantum processor and a dilution refrigerator could theoretically execute the mathematical heavy lifting that currently requires a 100 megawatt city size warehouse expansion of classical AI data centers.
Sam:Okay. That sounds like a perfect technological utopia for twenty forty, but we can't wait fifteen years for quantum coprocessors to mature and save the local power grids. Tech companies are training massive trillion parameter classical AI models right now. We need a radical way to cool thousands of massive heat generating AI chips today without draining local water tables or crashing municipal power grids.
Delilah:And the solution engineers have landed on is honestly wild.
Sam:Just throwing the computers entirely into the ocean.
Delilah:The Deep Blue fix?
Sam:Yes. If you want an ultimate cheat code for thermal management, look at the ocean. And this isn't just a whiteboard theory anymore. Our sources detail massive real world deployments. Microsoft famously proved the concept a few years ago with Project Natick, sinking a sealed pod of servers off the turbulent coast of Scotland to see if they survived.
Sam:And they did.
Delilah:They absolutely did.
Sam:Fast forward to June 2026, and China is officially switched on a massive commercial scale, scale. 24 megawatt underwater data center just off the coast of Shanghai.
Delilah:What is genuinely fascinating about these aquatic deployments is how elegantly they solve three massive escalating terrestrial problems simultaneously.
Sam:Okay. Let's go through them.
Delilah:First, and perhaps most importantly for the climate, is zero freshwater usage. Standard land based data centers gulp down trillions of liters of clean, potable drinking water every year through massive evaporative cooling towers. They literally evaporate drinking water to pull heat away from the silicon.
Sam:Which is crazy when you think about it.
Delilah:It is. Underwater data centers, however, use a completely closed loop system of internal copper piping that transfers the heat directly to the exterior metal hull, which is exposed continuously to the cold flowing seawater. They chill the servers twenty four seven without consuming a single drop of drinkable water.
Sam:That alone is a geopolitical game changer for coastal cities dealing with severe multiyear droughts.
Delilah:Absolutely. Second, underwater placement allows for direct hookups to offshore green energy. You mentioned the Shanghai facility. That entire 24 megawatt operation is wind powered. By dropping the server pods right next to the massive pylons of offshore wind farms, you hook the competing power directly into the clean energy source at the point of generation.
Sam:Oh, so you don't even have to send the power back to land.
Delilah:Exactly. You completely bypass the need to build incredibly expensive high voltage undersea cabling to transport that raw power to the mainland grid, which relieves massive pressure on aging civilian infrastructure.
Sam:That's brilliant.
Delilah:And third, proximity to populations. Over 40% of the global human population lives within a 100 kilometers of a coast. Placing AI data centers offshore provides ultra low latency for user requests. The data doesn't have to travel through hundreds of miles of fiber optic cables from a massive rural server farm in the desert. It bounces straight from the coast to your device.
Sam:So zero freshwater waste, direct wind power integration, and incredibly fast response times. It sounds like the perfect engineering solution. But when I was reading through these deployment specs, the very first logical question that popped into my head was maintenance.
Delilah:Yes. The big question.
Sam:How on earth do you manage hardware failures? Who fixes a broken sparking hard drive underwater? Are they seriously training scuba divers to swim down with specialized waterproof screwdrivers to swap out motherboards?
Delilah:This is the operational reality that shocks most people when they first study aquatic data centers. Nobody fixes them.
Sam:Oh my, nobody. You just leave millions of dollars of hardware to rot.
Delilah:Nobody touches them. The entire software and hardware architecture is built around a philosophy known as let it die.
Sam:Let it die. That is a brutal design philosophy. Walk me through how that actually works in practice without degrading the network.
Delilah:The software architecture running inside an underwater pod is built with aggressive massive redundancy. If a specific server blade experiences a critical failure, say a memory module burns out, the overarching management software instantly flags the hardware, cuts power to that specific blade to prevent a thermal cascade, and uses virtualization to seamlessly instantly shift the active workload to a healthy dormant standby server located in the exact same pod.
Sam:So you really don't send anyone down?
Delilah:You do not send a diver down. You do not deploy or repair a submarine. You just let the broken silicon die in the dark and you logically route the data around it.
Sam:But surely that is a losing battle. You would eventually run out of backup servers. Right? If things are breaking constantly, the pod becomes deadweight.
Delilah:You would naturally assume so given how often terrestrial servers fail. But here is an incredible counterintuitive detail from the engineering data. Because these pods are perfectly sealed on land before deployment, operators pump out all the corrosive oxygen and ambient humidity, and they fill the pods with a 100% dry nitrogen gas.
Sam:So there is zero oxidation, no rust, no corrosion on the microchips?
Delilah:None whatsoever. Furthermore, because there are no human technicians walking the aisles, there is no one accidentally bumping into racks, vibrating the hard drives, or tracking in microscopic dust particles that clog the cooling fans.
Sam:Humans are the problem.
Delilah:We really are. The internal environment is an absolutely pristine chemically inert vacuum of dry nitrogen. When Microsoft analyzed the data from their Project Natick deployment, they discovered that the underwater servers were actually eight times more reliable than identical servers operating in a human accessible data center on land.
Sam:Eight times more reliable. That is a staggering improvement in hardware longevity just by removing the human element. So they just leave them down there indefinitely until the entire pod eventually fails.
Delilah:Not indefinitely. No. Because we have to remember the brutal three to five year economic obsolescence cycle of AI chips we discussed earlier. The hardware doesn't physically fail, but it becomes financially useless. So operators perform an incredibly heavy duty maneuver called a pull and refresh.
Sam:What does that look like?
Delilah:Every three to five years, a specialized maritime crane vessel sails out to the coordinates, hooks onto the pod, and hauls the entire multi ton pressure vessel up from the seabed onto a barge. They power wash off all the accumulated barnacles, kelp, and algae, transport it to a coastal facility, pop the seal, pull out all the perfectly functioning but economically outdated silicon, slide in the newest generation of AI chips, reseal the hole with fresh nitrogen, and sink it right back down to the bottom.
Sam:Okay. I have to challenge this from an environmental standpoint. Yeah. Because we are talking about structures the size of shipping containers, eventually scaling up to football field sized deployments, generating 24 megawatts of pure thermal energy. Does a 24 megawatt data center essentially boil the ocean?
Sam:Are we inadvertently cooking the local marine ecosystem just to process AI requests?
Delilah:It is a vital nonnegotiable question that marine biologists have scrutinized heavily. To understand the thermal impact, we have to distinctly separate the macro global perspective from the micro local perspective. From a macro perspective, the global ocean temperature is entirely mathematically unaffected.
Sam:Okay. So it's a drop in the bucket.
Delilah:The natural surface of the world's oceans constantly vents roughly 36 petawatts of heat energy into the atmosphere through natural evaporation. A 24 megawatt facility doesn't even register as a rounding error on a global odynamic scale.
Sam:But locally, right next to the metal hull where the heat is transferring directly into the water.
Delilah:Locally, it is a very different highly concentrated story. The passive heat exchangers on the hull create what oceanographers call a thermal plume. The water immediately surrounding the pod experiences a measurable localized temperature increase of roughly 0.1 to one point zero degrees Celsius depending on the current.
Sam:So not literally boiling, but definitely creating a significantly warmer microclimate.
Delilah:Yes. And this creates a fascinating unintended ecological consequence. Artificial reefs. Marine life doesn't instinctively flee from these warm metal structures. They're actually intensely attracted to them.
Sam:Wait. Really?
Delilah:Oh, yeah. Microscopic algae and barnacles rapidly colonize the wormhole. Small fish arrive to graze on the algae, and larger pelagic predators follow the small fish. In the winter months, that thermal plume acts like a mild localized hot tub, drawing in massive amounts of marine life seeking shelter from the colder surrounding waters.
Sam:Wait. On the surface, that sounds like a massive environmental win. We are accidentally creating thriving fish habitats.
Delilah:It can be a net positive for local biomass, but marine biologists warn of serious insidious risks that aren't immediately visible. The real danger isn't the heat, it is the noise.
Sam:Oh, from the servers.
Delilah:Yes. The constant unrelenting mechanical of thousands of high speed cooling fans spinning inside the metal hull reverberates into the water creating acoustic shielding. This localized wall of noise heavily masks the natural underwater acoustic signals that fish, dolphins, and whales rely on to communicate, navigate, and avoid predators.
Sam:That's terrible.
Delilah:Furthermore, localized warm spots can confusingly alter natural migratory and breeding patterns for temperature sensitive species. The engineering solution to mitigate this is entirely location based. You cannot drop these pods in calm, shallow, biologically dense coastal coves.
Sam:You need moving water.
Delilah:Engineers specifically place these massive pods in areas with high velocity turbulent tidal currents. The rushing heavy volume of water instantly shears apart the thermal plume and rapidly dilutes the heat and the sound into the open ocean, minimizing the impact on local ecosystems.
Sam:Before we leave the ocean and look to the sky, I have to ask about the crossover here. Can we put our freezing cold, highly sensitive quantum computers underwater? Is that a viable road map?
Delilah:Yes. And architecturally, it makes a profound amount of sense. The deep seabed provides absolute isolating silence. A mile underwater naturally blocks out ambient terrestrial Wi Fi signals, stray cellular radiation, atmospheric cosmic radiation, and completely dampens the seismic surface vibrations from human activity.
Sam:All the exact things ruin the quibbets.
Delilah:All the exact stray energies that cause a delicate quibbet to experience decoherence. Using the sheer mass of the ocean as a natural silent cryogenic shield for future quantum hardware is a major highly anticipated long term goal.
Sam:Okay. Let's look at the brutal economics of this for a second. Sinking computers into the ocean costs roughly $226,000,000 for a massive 24 megawatt facility. It uses zero fresh water, It has an eight x y ability boost, and it provides a great shield against interference. It seems like the perfect ultimate solution.
Sam:So, why on earth is Elon Musk currently building a sprawling 11,000,000 square foot Gigasat factory in Texas to mass produce AI data centers and shoot them into space for billions of dollars. This brings us to the ultimate hedge against terrestrial limits, orbital computing.
Delilah:Space is undeniably the final frontier for digital infrastructure, but thermodynamically, it is the most counterintuitive and hostile environment imaginable for processing data.
Sam:I always fundamentally assumed space was a fantastic place to put a computer because it's essentially an infinite free freezer. Isn't space incredibly cold?
Delilah:This is perhaps the greatest, most pervasive misconception about aerospace engineering. Yes, the ambient temperature of deep space is technically near absolute zero, but far more importantly, space is a vacuum. Right. On Earth, whether you are using cold air in a warehouse Iceland or freezing sea water in the ocean, you cool a computer through the physical processes of convection and conduction. The heat transfers directly from the hot metal into the physical fluid or gas and the fluid naturally carries the heat away.
Sam:But in space
Delilah:In the vacuum of orbit, there is no air. There is no water. There are no physical molecules to carry the heat away. The only physical mechanism a satellite has to shed the massive thermal load of an AI processor is through radiation literally emitting slow, incredibly inefficient infrared light into the void.
Sam:So putting a massively hot 24 megawatt AI computer in a vacuum isn't putting it in a freezer. It's putting it in a perfect inescapable heat trap.
Delilah:A brutal, potentially catastrophic heat trap. To keep an orbital AI data center from physically melting its own circuitry into slag, it requires gargantuan, ultra bulky liquid radiator wings. These deployable radiator panels are fundamentally massive, and they are often physically larger and heavier than the computing hardware itself.
Sam:And that sheer mass absolutely destroys the cost efficiency of the launch. Right? Let's look at the brutal cost comparison from our sources. We established a 24 megawatt facility in the ocean costs about $226,000,000. To achieve that exact same 24 megawatts of computational power in orbit, you would need to launch a constellation of 160 to 200 massive radiator equipped data center satellites.
Sam:And the estimated cost for that orbital constellation.
Delilah:Between 4,500,000,000 and $6,000,000,000. And that staggering figure is with the massive cost reductions provided by reusable rocket architecture.
Sam:So I have to ask the obvious question. Why do it? Who underwrites a $6,000,000,000 orbital heat trap when they could deploy the exact same computational power in a $226,000,000 ocean pod. Seems like financial madness.
Delilah:Tech giants and nation states pursue this extreme path because of the strategic absolute advantages that no terrestrial or oceanic location can ever match. First and foremost, absolute geopolitical sovereignty.
Sam:Okay. What does that mean in this context?
Delilah:An underwater data center, no matter how deep, is still legally and physically tethered to earth. It heavily relies on a subject to local maritime laws and data sovereignty regulations. Most importantly, it can be catastrophically affected by terrestrial energy crises, sudden political shifts, or the intentional hostile physical sabotage of its undersea fiber optic cables by a foreign submarine. An orbital data center is an independent, untouchable island. It doesn't rely on local power or local politics.
Sam:Which perfectly brings up the second major driver, bypassing the nightmare of terrestrial red tape. Building a massive data center on land or even off a coast takes years of agonizing environmental impact reviews, zoning board approvals, and waiting endlessly in line for utility grid hookups. In space, once you have your launch clearance from the FAA or equivalent body, you can scale the constellation exactly as fast as your factories can build rockets. There are no beachfront homeowners suing you over the visual impact of your orbital satellite.
Delilah:Exactly. The regulatory friction is practically zero. And the third, perhaps most critical long term driver, is radical unfiltered solar efficiency. While the ocean gives you free infinite cooling, space gives you free infinite uninterrupted power.
Sam:As long as you position it right.
Delilah:Right. If orbital engineers place these AI satellites in specific trajectories called dawn, dusk, sun synchronous orbits, the satellites ride the terminator line between day and night. They are essentially never physically shattered by the earth. Because there is no thick atmosphere, no clouds, and no weather to scatter or absorb the incoming light, space grade solar arrays provide five to 13 times more energy annually than the exact same solar panels bolted to the ground on It is a completely unfiltered, uninterrupted fire hose of pure stellar energy.
Sam:And there's also the massive logistical benefit of true edge computing. Right? Our sources noted that there are currently thousands of commercial and military satellites capturing petabytes of high resolution earth observation data every single day. If those imaging satellites have to beam all that raw uncompressed data down to terrestrial servers just to be processed, it creates a massive sluggish bandwidth bottleneck.
Delilah:Yes. The downlink bandwidth is a massive choke point. By putting the AI data centers directly into orbit right next to the sensors, you create the ultimate edge network. A climate monitoring satellite can simply use high speed lasers to shoot its raw terabyte sized imagery imagery directly over to the neighboring AI data center satellite.
Sam:Who processes it up there?
Delilah:The AI processes the imagery instantly in orbit, identifies the crucial information, and only beams down the compressed highly specific answer like a wildfire has been detected starting at these exact GPS coordinates. It saves massive amounts of time, terrestrial bandwidth, and actionable latency.
Sam:But this brings us right back to the brutal three to five year AI chip replacement cycle. Yeah. The economic obsolescence of silicon doesn't stop just because the chip is in zero gravity. If I can't send a scuba diver down to the ocean floor to fix a server, I certainly cannot justify sending a human astronaut up in a capsule to manually swap out obsolete GPUs every three years like they did for the Hubble Space Telescope. The math just doesn't work.
Delilah:No. Human maintenance is entirely out of the question. The maintenance cycle in space is an aggressive, entirely automated financial meat grinder. There are two distinct architectural paths for managing this depending entirely on the specific orbit of the satellite. Let's look at low earth orbit or LEO first.
Delilah:The strategy here is ruthlessly simple. Launch, depreciate, and burn.
Sam:Wait. Burn as in, like, intentionally plunging them into the atmosphere to literally burn us.
Delilah:Literally vaporize into plasma. After three to five years when the hardware becomes economically obsolete or the delicate silicon is critically degraded by the constant bombardment of solar radiation, the satellite is instructed to use its last remaining reserve of thruster fuel to autonomously de orbit.
Sam:That is wild.
Delilah:It drops into the thick layers of the atmosphere at hypersonic speeds and burns up 100%. It is a complete intentional capital loss. No space debris is left behind.
Sam:I am struggling to accept that. That sounds like absolute financial madness. Yeah. You were telling me they intentionally burn up a multimillion dollar satellite just because the internal silicon chip is 36 months old.
Delilah:It sounds like madness until you look at the long economics. It is only financially viable because next generation heavy lift reusable rockets like Starship are radically driving the cost to launch bulk cargo down to an expected unprecedented $200 per kilogram.
Sam:Oh, I see.
Delilah:At that historically low price point, it actually becomes cheaper to mass produce cheap disposable AI satellites on an assembly line, launch them in bulk, burn them up after three years and launch brand new ones rather than paying for a decade of expensive terrestrial electricity, land taxes, and cooling infrastructure. The software layer treats the entire orbital constellation as a single fluid self healing cloud. If one specific satellite begins its fiery death plunge, the network autonomously and seamlessly redistributes its computing tasks to the healthy satellites flying nearby. The user on Earth never notices a drop in service.
Sam:Wow. Okay. I see the brutal logic in LEO. But what about the much higher, more expensive orbits? Geostationary orbit or GEO, where the satellites stay in a fixed position 22,000 miles above one specific spot on Earth.
Sam:You can't just easily or cheaply drop those massive platforms back into the atmosphere to burn up. It requires too much fuel.
Delilah:For GEO, we enter the fascinating, highly complex era of robot repair fleets. Space agencies and private aerospace firms are currently deploying autonomous space servicing vehicles or SSVs. These are essentially highly advanced robotic mechanics permanently stationed in orbit.
Sam:So zero human hands involved, just autonomous robots repairing other autonomous robots 22,000 miles in the sky.
Delilah:Exactly. To make this work, future GEO satellites are being designed with modular satellite architecture. Instead of the traditional method where every wire is custom soldered and sealed shut behind metal panels, these new data centers are built structurally like a standard desktop PC. They feature exterior standardized plug and play component base.
Sam:So the robot can just grab it?
Delilah:An SSV robot can fly up, autonomously dock with a massive data center satellite, use precision robotic arms to pull out a fried power module or a financially outdated GPU blade, and seamlessly slide a brand new freshly launched upgrade module right into place. The satellite is upgraded without ever leaving its orbit.
Sam:So we've talked extensively about classical AI migrating to space, but what about the end game? Can we put our ultimate computational engines or quantum computers in orbit?
Delilah:For the superconducting quantum computers, the massive beautiful chandeliers that require dilution refrigerators cool to near absolute zero no. They are a profound thermodynamic nightmare for space architecture. The cryogenic liquid helium fridges are far too heavy and power hungry to launch efficiently and the incredibly fragile qubits would be instantly, constantly destroyed by the intense cosmic radiation of deep space and the unavoidable micro vibrations of the satellite's attitude control thrusters.
Sam:That makes sense.
Delilah:However, as is always the case in engineering, there is a massive game changing exception.
Sam:I knew there had to be a workaround. What is the exception?
Delilah:Trapped ion and photonic or light based quantum notes. These specific alternative quantum architectures do not require extreme cryogenic cooling. They can operate at or near standard room temperature. Global governments and military intelligence agencies are highly intensely interested in mass launching these specific nodes into space.
Sam:What for if not for data processing?
Delilah:They aren't launching them to act as massive data processing centers, but rather to serve as the backbone for a secure quantum Internet. By placing light based quantum nodes on orbital satellites, we can use targeting lasers to shoot entangled photons across the vacuum of space without the degrading interference of the thick atmosphere or the distance limitations of terrestrial fiber optic cables. This creates an unhackable instantaneous global quantum cryptographic communications network.
Sam:Okay. Let's take a deep breath and synthesize all of this extreme engineering. What does this all actually mean for the trajectory of human technology? We started by looking at how the very atomic nature of computing is fundamentally evolving, moving away from rigid linear binary switches to choreographing vast fluid possibilities at temperatures colder than the vacuum of space. And to fuel this quantum evolution and to temporarily feed the massive ravenous power demands of classical AI in the meantime, we are literally stretching the known laws of physics and logistics to their absolute breaking point.
Delilah:We really are.
Sam:We are intentionally boiling microplumes in the ocean to cool servers without wasting fresh water, and we are mass producing fleets of disposable satellites explicitly designed to burn up in the atmosphere just so we can harvest unfiltered solar energy. We're doing all of this, spending hundreds of billions of dollars just to keep our digital world running seamlessly.
Delilah:And if we connect this extreme physical reality to the bigger philosophical picture, I wanna leave you with a final slightly lingering thought to mull over. As we aggressively push the physical brains of our civilization into environments that are entirely lethally hostile to human life, the crushing dark depths of the seabed, the freezing irradiated vacuum of orbit, and confined cryogenic chambers colder than deep outer space, what fundamentally happens to our relationship with our own technology?
Sam:It's a great question.
Delilah:When the physical hardware that runs our financial systems, our communications and our scientific discoveries becomes entirely permanently inaccessible to the human hand. When a machine is operating a thousand miles above your head silently upgrading and repairing itself with autonomous robotic arms in the dark, who really holds the off switch?
Sam:That is a chilling, incredible reality to consider. Because as we said at the very beginning of this journey, we deeply love to imagine our digital lives exist in a weightless, invisible, harmless cloud. But the physical reality is a massive, incredibly heavy, violently hot, and increasingly untouchable machine. Thank you so much for taking the time to join us on this deep dive.
Dave Bunyard:Is it worth it to put a data center in space or in the ocean? You decide. I hope you have enjoyed this episode. Find more episodes at life at 100.