ChatGPT Is Now an Ad Platform, and It Got There Faster Than Anyone Expected
OpenAI's advertising pilot has hit $100 million in annualised revenue in under two months, working with more than 600 advertisers and with projections of $30 billion in ad revenue by 2030. The company that set out to build safe, beneficial AI has quietly become one of the fastest-growing ad businesses in tech history, and the tension between that revenue model and its stated mission is only going to sharpen.
Read full story →Meta Is Building a Gigawatt Data Centre in the Desert, and It Just Got Six Times Bigger
Meta has expanded its El Paso, Texas data centre investment from $1.5 billion to $10 billion, targeting 1 gigawatt of capacity by 2028. The site is part of a broader pattern: hyperscalers are now committing to infrastructure at a scale that rivals national electricity grids, raising real questions about where the power comes from and who bears the cost.
Read full story →The Week AI Agents Got a Control Panel
Microsoft's Agent 365 platform goes live in May, giving enterprises a unified control plane for governing autonomous AI agents. Meanwhile Anthropic quietly launched "auto mode" for Claude, letting the model decide on its own which actions are safe to take without asking permission. Two very different approaches to the same problem: how do you manage AI that acts without being told to?
Read full story →Two Washingtons, One AI Future
This week, Silicon Valley executives and Trump officials celebrated AI in one DC venue while labour leaders and senators strategised resistance in another. Bloomberg's account of the split reveals that the politics of AI are no longer simply pro-innovation versus cautious -- they are now shaped by two genuinely incompatible stories about who benefits, and the bipartisan consensus may be cracking.
Read full story →Harvey's $11 Billion Bet: Legal AI Moves from Experiment to Infrastructure
Harvey has raised $200 million at an $11 billion valuation, signalling that legal AI has graduated from productivity tool to core law firm infrastructure. The timing is pointed: it arrives the same week a senator noted that major law firms have stopped hiring first-year associates because AI can now do that work.
Read full story →AI Adoption Is Spreading, But Not Evenly
Anthropic's new Economic Index report shows AI use converging across US states, but diverging globally, with the countries already using AI most per capita pulling further ahead. The rate of augmentation is rising and task diversity is broadening, but the data raises an uncomfortable question: spreading usage and equitable access to AI's gains are not the same thing.
Read full story →The Traders Who Bet on the Forecast About the Forecast
European energy traders are using AI not just to read weather forecasts, but to predict when the dominant ECMWF forecast is about to shift before it does. It is a perfect case study in how competitive markets consume information edges: once everyone has the same model, the advantage moves up one level of abstraction.
Read full story →The Ad Model Is the Original Sin
OpenAI hitting $100 million in ad revenue in two months is not a business milestone. It is a structural commitment. Once you build a product around attention and engagement, the model's job is no longer to be useful; it is to keep you clicking. Peter Harrison on why the advertising model is incompatible with the safety mission, and why we've seen this movie before.
Read opinion →