Your bank's most important technology partner probably isn't a bank. It's a startup that has never turned a profit. And that, according to Moody's, is becoming a systemic risk the financial world can't afford to ignore.
On August 9, the ratings agency published a stark warning: the banking sector's race to embed AI into every layer of operations is creating what it calls a "systemic dependency" on a tiny cluster of Silicon Valley companies. OpenAI and Anthropic were named. Both are loss-making. Both face mounting pressure from investors to turn a profit. And both now hold pricing power over an industry that can't function without them.
The Numbers That Should Worry You
More than three quarters of UK financial services firms are already using AI, with insurers and international banks leading the charge. Lloyds Banking Group alone is pouring £13 billion into a technology overhaul that includes £2 billion in cost cuts — and a frank admission that AI will reshape its workforce. Moody's attached a stark probability to the broader displacement risk: a one-in-five chance that AI will be able to do the work of a capable mid-level employee by 2030.
But the cost savings Moody's envisions come with a catch. "Substantial investments" will be required up front, and the agency fears that because so many rivals are racing toward the same goal, much of the benefit will simply be "competed away." In other words: banks will spend billions, only to end up roughly where they started — except now they'll be dependent on a handful of external providers they can't easily replace.
The Vendor Concentration Problem
This is not a theoretical concern. The structural dependency Moody's describes has a clear mechanism.
A single outage at a foundation model provider — say, OpenAI's API goes down — could cascade through the entire financial sector in minutes. Because banks, insurers, and payment processors all lean on the same APIs, a failure at one node becomes a failure everywhere. That's the definition of systemic risk.
Then there's pricing. OpenAI and Anthropic are burning cash. At some point, investors will demand returns. When that day comes, the banks that have woven these APIs into their core operations will have no leverage. They'll pay what they're told to pay, or they'll face the impossible task of ripping out and replacing their entire AI stack on short notice.
Moody's does note that banks aren't entirely powerless. They control proprietary data — the one asset AI providers can't replicate. Large institutions have decades of experience grinding down technology contracts. And open-source models, alongside selective partnerships, offer alternatives worth exploring.
Regulators Are Already Circling
The warning doesn't come in a vacuum. The Bank of England disclosed in April that it is actively stress-testing AI as a systemic financial stability risk — not just as a compliance or conduct issue. HM Treasury proposed an assurance framework for third-party AI and scrutiny of critical suppliers in the same week Moody's published. When the rating agency and the regulator arrive at the same diagnosis independently, the signal is hard to ignore.
The UAE Angle
For banks and financial institutions in the UAE — where AI adoption is accelerating under national strategies — the Moody's warning carries added weight. The region's banks are modernizing rapidly, often partnering with the same global AI and cloud providers that dominate Western markets. The concentration risk is not someone else's problem. It's local.
The Deposit Flight Risk Nobody's Talking About
One of Moody's more unsettling projections sits at the intersection of AI and consumer behavior. As AI-powered tools make switching between bank accounts trivially easy, deposits could start moving at short notice. The old friction that kept customers loyal — paperwork, hassle, inertia — evaporates when an AI assistant can compare rates, fill forms, and execute a full switch in under a minute. Moody's flagged this as a distinct risk to funding stability and depositor trust. In a world where capital can flee at the speed of an API call, the banks that built the AI gateway are also building the escape hatch.
The Open-Source Escape Hatch
Not every path leads to vendor lock-in. Open-weight models — from Meta's Llama family to the increasingly capable open-source alternatives — offer banks a way to run AI on their own infrastructure, with their own data, under their own terms. It's not cost-free: running models locally requires hardware, expertise, and ongoing maintenance. But it eliminates the per-call pricing model that will eventually make proprietary APIs unsustainable for high-volume use cases.
For banks building their AI strategy in 2026, the smart money splits the difference: proprietary models for low-sensitivity, high-experimentation use cases; open-weight models running on local servers for everything that touches customer data, risk, and compliance. That's the architecture Moody's is implicitly recommending — even if they're too diplomatic to say it outright.
What This Means for Your Organization
The takeaway isn't "don't use AI." It's "don't build your entire AI strategy on a single supplier you can't walk away from." Three practical moves:
- Multi-vendor resilience. Don't bet the farm on one AI provider. Build your stack so that models and APIs can be swapped without rewriting everything.
- Own what you can. For sensitive workloads — customer data, risk modeling, compliance — consider running open-weight models on local infrastructure. Control your data and your costs.
- Prepare for the regulator's questions now. When supervisors come asking about vendor concentration, you want to have an answer that isn't "we hadn't thought about it."
A Dependency We Built Ourselves
The Moody's report isn't a prediction of doom. It's a diagnosis of a dependency the industry chose to build. Banks raced toward AI hoping to cut costs and gain an edge. Instead, they may find themselves locked into supplier relationships that look more like toll booths than partnerships.
The solution isn't less AI — it's smarter AI adoption. Own your stack. Diversify your suppliers. And never let a single vendor become your single point of failure.