Think of an AI data centre as a factory. Electricity goes in, and tokens come out. A token is a small piece of text, roughly a word.
So what limits the factory? Not the building, and not even the chips. It is the power. A site is connected to the grid for a fixed amount of electricity.
The Measure That Matters
If power is fixed, the question becomes simple: how many tokens can you produce from each unit of power? That measure is called tokens per watt.
It is a ratio of what the factory makes to what it uses. Because the power is the part that cannot grow, the ratio is where improvement has to come from.
A Worked Example
The video uses round numbers. A site has 100 MW. With one generation of chips it produces 1 billion tokens a day. A newer generation, on exactly the same power, produces 3 billion.
Same building, same electricity, three times the output. Since tokens are what the factory sells, that is three times the revenue.
Three times the output from the same power. Round numbers for illustration.
How It Changes the Way Chips Are Judged
When power is the limit, the cheapest chip is not the best one. The best one is the chip that turns each watt into the most tokens, even at a higher price.
It also explains why older chips are replaced while they still work. They are not broken. They are occupying power that a newer chip could turn into more output.
Tokens per Watt builds on GPUs. Tokens per watt judges chips by how many tokens each unit of power yields, so it needs the GPU lesson’s picture of AI chips.
Tokens per Watt builds on Tokens. Tokens per watt measures a data centre’s output in tokens, the unit the tokens lesson defines.
The Business Engineer framework maps where each part of the AI stack sits and who holds the leverage. These pills are short lessons from the same library.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
