One sentence in the text shows where this is going: in the AI economy, compute is revenue.
It is written by Jensen Huang himself and it is not a figure of speech. It explains why a chip maker is suddenly talking about plots of land, substations and power lines.
Fact
The facts: on 17 August 2026 NVIDIA announced a partnership with SB Energy to secure land, power and shell at the PORTS-Pike Technology Campus in Portsmouth, Ohio, where NVIDIA compute will be hosted. The tenant will be OpenAI. The company describes those three things as the next critical resource and says it is applying the same discipline it used to secure chips, packaging, memory and networking. The argument: the largest cloud providers and investment-grade enterprises have long secured these things themselves, while everyone else has neither the balance sheets nor the long-term contracts.
What is actually happening
A chip maker becomes a landlord. It secures the sites, puts its own compute there, and rents it to whoever will use it.
If you followed how they fought over manufacturing capacity and memory for the last two years, you will recognise the move. It is the same one, just a step further down the chain.
The constraint is no longer how many chips you can buy. It is where there is power to run them.
And here is the uncomfortable part. When one supplier holds the chips, the networking, the software, and now the sites with the power, the word dependency starts to feel small for the situation.
I am not going to pretend I know how this lands on prices. I do know what every business does once it has gathered the whole chain into one pair of hands. Nobody has ever cut prices at that moment.
For you today: nothing. For the argument about who sets the price of compute in three years: quite a lot.
Official primary sources
NVIDIA: Securing the Infrastructure of Intelligence, Jensen Huang (17.08.2026)The visual is generated code-art, no third-party images.