AYA CORE
Who Pays for the Watts?
The technology industry is racing to build bigger furnaces. Crypto already ran that experiment — and the results are in. Why the future of computing, and of money, belongs to whoever stops burning.
There is a race on, and you are paying for it whether you entered or not.
Microsoft, Google, Amazon, Meta and Oracle will spend somewhere between $600 and $700 billion on data centers and AI infrastructure this year alone. The International Energy Agency expects global data-center electricity consumption to roughly double by 2030 — from about 415 TWh in 2024 to around 945 TWh, on its way to 1,200 TWh by 2035. If data centers were a country, they would soon be the world's fifth-largest electricity consumer, wedged between Japan and Russia.
This is not an abstraction. Data centers already consume 26% of Virginia's electricity. Residential bills in Ohio and western Maryland are rising by $16–18 a month specifically because of data-center load. Jensen Huang estimates a single one-gigawatt "AI factory" costs $40 billion to build. Dozens are planned.
The bet behind all this spending is simple: whoever burns the most, wins. More chips, more megawatts, more cooling towers, more river water evaporating into the sky.
We think this bet is wrong. Not morally wrong — arithmetically wrong. And we say this as a technology project, not as spectators: our industry, crypto, ran this exact experiment first. The results are in.
The experiment crypto already ran
Bitcoin currently consumes somewhere between 138 TWh (Cambridge's estimate) and 204 TWh (Digiconomist's) of electricity per year — about as much as Poland or Thailand, roughly half a percent of all electricity generated on Earth. Add up Cambridge's annual figures for the past decade and you get on the order of a thousand terawatt-hours: approximately one full year of Japan, spent on keeping one ledger honest.
Then there is the water. In a study published in Cell Reports Sustainability, financial economist Alex de Vries calculated that Bitcoin mining consumed over 1,600 gigaliters of water in 2021, heading toward 2,300 gigaliters — water used to cool the machines and the power plants behind them, much of it evaporated and gone. His per-transaction framing (about 16,000 liters, a backyard swimming pool per payment) is disputed, and fairly so — per-transaction accounting is a blunt tool. The network-level figure is not disputed. In the United States alone, Bitcoin's water footprint matches the household consumption of roughly 300,000 American families.
What did the electricity cost in money? At the cheap industrial rates miners hunt for, a decade of Bitcoin's power bill runs into tens of billions of dollars. For scale: the UN World Food Programme has estimated that about $3 billion a year would feed every one of the world's hungry schoolchildren — at 25 cents a meal. One year of Bitcoin's electricity bill is two to three years of school lunches for every hungry child on the planet. The comparison is unfair in the way all such comparisons are unfair, and it is still worth sitting with.
Here is the part of the story that gets told less often. In September 2022, Ethereum — then the second-largest furnace in crypto — switched off its mining and moved to proof-of-stake. Its electricity consumption fell from tens of terawatt-hours to 7.87 GWh a year: a 99.98% reduction, achieved by a change in software architecture. The entire global Ethereum network now uses less electricity than half the British Museum. The average node draws 105 watts — a bright light bulb.
Nothing about the hardware changed. Nothing about physics changed. The architecture changed, and 99.98% of the energy turned out to be unnecessary.
Hold that thought.
The eighty-year-old habit
Almost every computer on Earth — your phone, the miners, the AI clusters — is built on an architecture John von Neumann sketched in 1945: memory over here, processor over there, and a bus shuttling data between them. It was a brilliant design for 1945. It has a flaw that only became fatal at modern scale: the shuttling itself is now the main cost.
IBM's researchers put a number on it: moving 64 bits of data from memory to the processor consumes 10,000 to 2,000,000 times more energy than actually performing a multiplication with those bits. Read that again. The computation is nearly free. The commute is the bill. Modern AI is a warehouse worker with a brilliant brain and a wheelbarrow: the thinking is instant; the fetching is forever. The data-center race, at its core, is a race to buy more wheelbarrows.
The exits from this trap already exist in the lab. In-memory computing writes a neural network's weights directly into the physical state of the chip — frozen into phase-change glass, written once — and then computes in place, where the data lives, with nothing to fetch. IBM's digital cousin of the idea, the NorthPole chip, ran a language model 73 times more energy-efficiently than the best GPU it was tested against. Neuromorphic designs go further and borrow the brain's rule: components that have nothing to do draw almost nothing — the chip works when there is work, and sleeps when there isn't. Google's TPUs, to give credit, already took a half-step this way years ago, streaming data through the chip instead of round-tripping it to memory. The human brain, for reference, runs the best-known intelligence on about 20 watts. The furnace is not the destiny of computing. It is a habit.
The honest conclusion from both stories — Ethereum's and the chip labs' — is the same: when energy use falls by 99% or more, it is never because someone built a bigger power plant. It is because someone stopped doing unnecessary work.
Who pays when the bet fails
Now, about the race. Not every one of these hundred-billion-dollar bets can win; that is what makes it a race. Some of this capacity will sit dark. Some of these companies — and the utilities borrowing $1.4 trillion to serve them — will be wrong.
We have seen this movie, and we know who buys the tickets. When banks bet well, the winnings fund yachts — and a banker's yacht is never at risk; it is moored on the far side of the balance sheet. The side that sinks is traditionally reserved for depositors and taxpayers. In 2008 the losses were "systemic," which is the technical term for yours. The mechanism is already warming up in miniature: the grid build-out is on your utility bill today, years before a single promised miracle arrives. If the miracles come, splendid. If they don't, the write-downs will be socialized with the usual speech about how nobody could have known.
And here is a detail our banker friends prefer not to advertise: the traditional banking system — its data centers, branches and ATMs — consumes an estimated 260 TWh a year, nearly twice the Cambridge estimate for Bitcoin. The old money burns more than the new money, with better marble and calmer press coverage.
The path we chose
AYA CORE was designed on the other side of this argument, from the first line of code.
There is no mining in our network. No lottery, no race, no warehouse of machines guessing numbers and discarding quintillions of wrong answers every ten minutes. Consensus is reached by a small quorum of licensed validators using post-quantum signatures — ordinary servers doing ordinary work.
More importantly, the network is event-driven. Epochs advance when there are transactions to finalize. When there is nothing to do, the validators do almost nothing — a heartbeat, and silence. No work is manufactured to justify a reward, because there is no block reward, no emission, no yield. Bitcoin burns the same gigawatts on a quiet Sunday as on its busiest day; our quiet Sunday costs approximately a quiet Sunday.
Today the entire AYA network — every validator on Earth — draws well under a couple hundred watts. That is not a rounding error of Bitcoin's consumption; it is a rounding error of a rounding error, roughly eight orders of magnitude less. And the property that matters is structural, not circumstantial: there is no component in AYA whose income grows when electricity burns. A proof-of-work chain rewards whoever adds the next megawatt, so its appetite grows with its price, forever, by design. Ours cannot. Growth in users adds transactions, not furnaces; even a future network of thousands of validators, at Ethereum's measured 105 watts per node, would fit inside the power budget of one small office building.
Are we "the greenest"? Post-Merge Ethereum deserves genuine credit, and plenty of lean BFT networks exist. We will happily lose a greenness beauty contest to anyone. Our claim is narrower and harder: efficiency is not our policy, offset, or pledge — it is our construction. You cannot un-build it, the way you cannot mint our 635,836th symbol.
The industry will keep racing for a while — bigger campuses, taller cooling towers, louder announcements. Physics and arithmetic will keep whispering the same thing the Merge and the chip labs already proved out loud: the cheapest, cleanest, most honest watt is the one you never draw.
Energy saved is the only subsidy that never needs a bailout.
The AYA wallet runs in the browser at ayacoin.online and is available for iPhone on the App Store. Documentation: portal.ayacoin.online/docs.
This article is educational and does not constitute financial or investment advice.