AI is eight layers. Most coverage reads only the top one. Strata maps all of them, from the power station at the bottom to the agent at the top, and every number on this site carries a source and a confidence label.
The layers
Band Rules and society: Laws, courts, export controls, and whether your neighbours will let you build.
L6 Applications and agents: The products people actually use. Coding assistants, support bots, agents that run for hours.
L5 Tooling and infrastructure: The plumbing between a model and a product. Serving, retrieval, agent protocols, evaluation.
L4 Models: The labs, their releases, and what a million words of output costs you.
L3 Iron and cloud: Servers, networks, and the companies that rent them to you by the hour.
L2 Silicon: The chips. Processors that do the maths, and the memory that feeds them.
L1 Power and land: Electricity, land, water, permits. Nothing above this exists without it.
Band Money: Who is paying, and with whose balance sheet. Cuts through every layer.
Latest
NextEra and Dominion double the length of promised Virginia bill credits to help win merger approval
NextEra and Dominion filed new paperwork with Virginia's utility regulator on 14 September, offering residential customers four years of 10 dollar monthly bill credits instead of two. The companies say the extension is funded mainly by redirecting credits that would have gone to large data centres, plus about 90 million dollars from shareholders, a figure reported by the Virginia outlet Cardinal News rather than stated in the filing. The package also raises Dominion's low income assistance programme by 100 million dollars through 2038, adds a 100 million dollar workforce fund, a Virginia supplier programme of up to 1 billion dollars a year for five years, and 1,000 direct jobs. Virginia regulators hold hearings in November and are expected to rule in January 2027, with the 67 billion dollar merger still targeted to close in the second half of 2027. If you are a Dominion Virginia customer this means a longer bill credit if it is approved. Anywhere else, it changes nothing for you.
Crusoe opens a second Tulsa factory making the electrical gear AI data centres run on
· L1 Power and land · 400,000 sq ft added, 1M sq ft total · Confirmed
Crusoe, a company that builds AI computing facilities, opened a second factory in Tulsa, Oklahoma that makes electrical equipment such as switchboards, adding 400,000 square feet and bringing its national manufacturing footprint to roughly 1 million square feet. This grows Crusoe's own capacity to build components for data centres; it is not new electricity or computing capacity coming online. The company says it has hired more than 300 local craft workers across its two Tulsa sites, and that its factories have supplied more than 2,500 switchboards to its Stargate Project site in Abilene, Texas. For an ordinary person the concrete change is local: manufacturing jobs in Tulsa, not a change to anyone's electricity bill.
South Korea's Ministry of Trade, Industry and Energy said the country's information and communications technology exports reached 59.98 billion dollars in August, a record and more than double the 22.84 billion exported a year earlier, driven largely by semiconductors. Imports rose to 18.61 billion dollars, leaving a trade surplus of 41.37 billion, the first time Korea's technology trade surplus has topped 40 billion dollars in a single month, according to the ministry's own release. This is a government tally of a whole export sector, not one company's results. Korea's chip exports have been one of the clearest ways the AI boom shows up in a national economy, and a surplus this large can support the country's currency and public finances, though an ordinary person sees no direct change from the figure itself.
Oracle's Larry Ellison cancels his plan to sell company stock
· Band Money · No shares sold under the cancelled plan · Confirmed
Oracle disclosed in a filing that Larry Ellison, the company's executive chair and chief technology officer, has cancelled a pre-arranged plan to sell Oracle stock. The filing says no shares were sold under that plan before it was cancelled, and that Ellison has no other plans in place to sell Oracle stock. These trading plans let executives sell shares on a schedule set in advance, so cancelling one without having sold anything under it says something about an insider's own intentions rather than being a sale in itself. For anyone who owns Oracle stock, or is watching the company's finances as it borrows to build AI computing capacity, this is a small concrete data point about what the company's largest individual shareholder is doing with his own shares, not a change to the business.
Microsoft is reported to be planning to more than triple its data centre capacity to about 38 gigawatts by 2032
· L3 Iron and cloud · about 38 GW by 2032, from about 12 GW · Reported
Bloomberg News reported, citing people it did not name, that Microsoft plans to more than triple its data centre capacity to about 38 gigawatts by 2032, up from about 12 gigawatts today. We read this through a trade outlet's account of Bloomberg's reporting rather than Bloomberg itself, which is why it runs as Reported. Microsoft has not confirmed the target, and the account says the company did not immediately respond to a request for comment. If accurate it would be one of the largest capacity expansions by any single cloud company. Nothing changes today; this is an unconfirmed report about long term planning.
Reuters, citing six people it did not name, reported that the United Arab Emirates is redesigning a planned 5 gigawatt AI data centre project, one of the biggest outside the United States, after last year's Iran war prompted a rethink of how and where critical infrastructure should be built. No UAE government body or company has confirmed this account on the record, so it runs here as reporting rather than a confirmed fact. If accurate, it points to AI data centres increasingly being treated as strategic, security-sensitive infrastructure in the Gulf, which could slow down or reshape how quickly that capacity gets built. Nothing changes today for anything already running.
Together AI, a company that rents out computing power for building and running AI models, lowered the price of fine-tuning, the process of further training an existing model on a company's own data, by 30 to 70 percent across most of the models it supports. For one model, OpenAI's open-weight gpt-oss-20b, the standard training price fell from 1.50 dollars to 0.40 dollars per million tokens processed. Together AI says the reduction comes from its own optimisation of the systems that do this training, not from a change in what the underlying chips cost. For a software team or startup that fine-tunes models to fit its own product, this is a direct reduction in one of the recurring costs of building an AI product, though it applies to training a model rather than running it once it is finished.
Australia's public broadcaster ABC reported that 27 data centre projects already approved but not yet built, with a combined capacity of at least 3.38 gigawatts, will not have to meet new water and power standards the government introduced earlier this year, because the rules only apply to projects approved from now on. Prime Minister Anthony Albanese told ABC Radio Sydney, "You can't retrofit. But as well, many of the centres which are under construction tend to be smaller ones," explaining why existing approvals would not be held to the new requirements. As of March 2026 Australia had 162 data centres running with a combined 1.5 gigawatts of capacity, so the exempted projects alone are more than double what the country already operates. For Australians living near one of these 27 sites, the facility being built nearby will not be held to the newer environmental standards that future projects will have to meet.