Moody's has published a warning that deserves more attention than it is getting: banks that rush into AI are quietly signing up for a new kind of dependency, and they may not fully understand the terms.
The rating agency's argument is simple. Banks and insurers are moving core work, like processing insurance claims and checking creditworthiness, onto AI systems built by a small number of companies. Those systems mostly run on cloud computing services owned by an even smaller group of firms.
When that many banks lean on the same handful of providers, one bad day for a tech company becomes a bad day for the entire financial system. This is not theoretical.
On October 20, 2025, a fault at Amazon Web Services took down banking services at Lloyds, Halifax and Bank of Scotland, disrupting access for as much as a quarter of UK banking customers in a single day. That is exactly the kind of chain reaction Moody's is describing, and it already happened before the ink on this report dried.
The second risk is about money, not uptime. OpenAI and Anthropic are spending far more than they bring in, with reported losses running into the billions each year, and both face growing pressure from investors to turn that around ahead of possible public listings.
Moody's point is straightforward: once a bank has spent years building its systems around one company's AI, walking away becomes costly and slow. That gives the AI provider room to raise prices once the losses need to stop.
Banks are not defenseless here. They have large contracts, decades of experience squeezing better terms out of tech vendors, and their own valuable asset that AI companies want: customer data. Some are also mixing in open-source AI models, which they can run themselves, as a way to avoid being fully boxed into one provider's pricing.
The part of the report that got less attention is about people. Moody's puts the odds that AI can fully replace a solid mid-level employee by 2030 at around one in five, which is a real number but far lower than the panic headlines suggest.
Lloyds' own plan, a £13 billion investment tied to £2 billion in cost cuts, backs this up. The bank is not planning to erase jobs wholesale, but it is openly telling staff that reskilling is coming whether they like it or not.
The lesson stretches well beyond banking. Any company plugging its operations into a handful of AI providers, whether through a cloud contract, a chatbot subscription, or an AI tool built into everyday software, is taking on the same kind of dependency risk banks now face.
Banks can at least negotiate. Most smaller businesses cannot, which makes this a risk worth planning for now, before the bill arrives.