Models, chips, power
AI & Computing
Every model runs on silicon, every chip runs on power, and every watt shows up in a quarterly report somewhere. We follow the stack from the wafer to the invoice.
Long reads
Every read in AI & Computing
AI & Computing
The chip race
Everything runs on somebody else’s wafer
The AI boom is a story about three companies, one island and a memory shortage. Understanding where the bottleneck sits explains most of what happens to prices, products and geopolitics.
AI & Computing
Power
AI is an electricity business now
A frontier training cluster draws as much power as a small city, and the grid connection queue is now a longer wait than the chips. This is the constraint that will shape the next five years.
AI & Computing
Agents
What an AI agent actually does all day
An agent is a language model in a loop with tools and permission to act. That sentence contains both the reason it works and every reason it fails.
AI & Computing
Reasoning
Models that think before they answer
Adding compute at answer time, rather than at training time, turned out to be a second scaling axis. It made models better at hard problems and much more expensive per question.
AI & Computing
Edge AI
The model in your pocket
Running a model on the device instead of in a data center solves latency, privacy and cost at once. The price is that the model has to be small, and small is a real constraint.
AI & Computing
Regulation
The AI Act, for people who have to comply with it
Europe wrote the first broad law for artificial intelligence. It is a product safety regulation, not a technology regulation, and that distinction decides almost everything about what it means for you.
AI & Computing
Quantum
Quantum computing, without the hype and without the sneer
Error correction started working. That is the milestone that matters, and it is why the field went from a permanent ten years away to a schedule with dates in it.
AI & Computing
AI in science
The instrument that reads everything
Protein structure prediction won a Nobel prize. Weather forecasting got faster and better. Materials discovery is next. The pattern in each case is the same, and it is not general intelligence.
Tools
Tools that belong to AI & Computing
Tokens per month to money
AI Inference Cost
What an AI feature costs per month at scale: requests, tokens in and out, model tier, cache hit rate, and the comparison with renting a GPU for a self-hosted model..
OpenGPUs to megawatts
Data Center Power
How much power a cluster of accelerators draws, what it costs a year, how much heat it rejects and what a hyperscale site needs from the grid, from GPU count, chip power, overhead and electricity price..
OpenBasics
Explainers for AI & Computing
How a blockchain works
A ledger copied across thousands of computers, extended by a rule everyone can check, where changing the past would mean redoing everything after it.
02Wallets, keys and custody
A wallet does not hold coins. It holds the key that lets you move them, and losing that key is the same as losing the coins.
03How large language models work
A very large function that predicts the next chunk of text. Everything these models can do is a consequence of doing that extremely well.
04Why AI needs GPUs
A neural network is millions of multiplications that can all happen at once. A graphics chip was already built to do exactly that.
05What is a data center?
A building full of computers, power and cooling. For AI the interesting unit is not square metres but megawatts, and that is the constraint on the whole industry.