Wall Street Picked Its AI Partners. Look at What They Actually Bought.
The biggest firms in wealth and asset management just signed billion-dollar partnerships with the frontier AI labs. The most important detail is the one nobody is talking about: the intelligence itself was never what they were paying for.
On June 9, Rockefeller Capital Management, a 140-year-old firm responsible for $212 billion in client assets, announced it is building an AI-enabled wealth platform directly with Anthropic. The same day, CNBC reported that JPMorgan, which already gives 250,000 employees access to OpenAI and Anthropic models, plans to deploy AI agents that work for hours without human input. Five weeks earlier, Anthropic and OpenAI launched rival enterprise services ventures within the same 24 hours: one a new venture with Goldman Sachs, Blackstone, and Hellman & Friedman valued at $1.5 billion, the other raising more than $4 billion at a $10 billion valuation from nineteen investors led by TPG. The day after that, Dario Amodei and Jamie Dimon shared a stage in New York.
If you run a firm that is not named Rockefeller or JPMorgan, the natural reading is that the giants just locked up the best AI. They did not, because the intelligence was never for sale. Everyone already has it. What the giants bought is integration, and that distinction points directly at what every other firm in this industry should do next. That is the whole argument of this essay, in three steps.
Everyone Is Renting the Same Brain
The megadeals are not exclusive licenses to a smarter model. The Claude that Rockefeller is building on is the same Claude available through an API to any firm with a credit card and an engineer. Bridgewater, Citadel, and Carlyle already use it. LPL is wiring it to more than 30,000 advisors, and Orion, iCapital, and CAIS are embedding it in the platforms mid-sized firms already run on. Frontier intelligence has become a utility: enormously powerful, broadly available, and priced like software rather than like an army of analysts. When everyone runs on the same models, the model is not the moat.
What the Billions Actually Buy
So what are the giants paying for? Follow the money and the people in these deals, and three line items repeat.
Engineering presence. FIS did not buy a chatbot. It got Anthropic engineers embedded in its teams to co-design a financial crimes agent that compresses anti-money-laundering investigations from hours to minutes. Goldman has Anthropic engineers inside its trade accounting and client onboarding functions. The scarce resource is not intelligence. It is the labor of wiring intelligence into regulated workflows.
Data plumbing. Anthropic's financial services launch came with Moody's, S&P Capital IQ, Morningstar, FactSet, and PitchBook connected natively. A model is only as useful as the data it is permissioned to see, and the giants are paying to make that data flow cleanly and compliantly.
Unglamorous workflows first. Rockefeller's first phase is client meeting intelligence, operational workflows, and internal support. Morgan Stanley's flagship deployment, which OpenAI reports is used by over 98 percent of the firm's advisor teams, started with knowledge retrieval and meeting notes. The most sophisticated buyers in finance, spending the most money, all started in the same place: the repetitive internal work where hours visibly leak.
Here is the simplest way to hold it: the model is the engine, and everyone is buying the same engine. The partnerships purchase the drivetrain, the connection between raw horsepower and wheels that actually touch the road.
The Gap Is Readiness, Not Access
The integrated deployments are producing real numbers. Norges Bank Investment Management reports roughly 213,000 hours saved annually, about a 20 percent productivity gain. Now set that against where the average firm stands: Deloitte finds 73 percent of advisory firms use AI in some form, but only 6 percent use agentic tools and only 5 percent have AI integrated across their systems. And when Objectway and FT Longitude surveyed 300 senior wealth executives this year, the biggest barrier to scaling AI was not technology or talent. It was integration. In this industry, integration is never just technical: it has to be explainable, auditable, and permissioned before it touches a client.
Most firms do not have an AI problem. They have a data problem, a workflow problem, or a governance problem. The intelligence is available to everyone. The integration is not, and that is exactly the gap the giants are spending billions to close.
What This Means If You Are Not a Megafirm
Two things, both encouraging. First, the capability stack is being productized downward fast. Anthropic's new services venture explicitly targets mid-sized enterprises that the large consultancies do not prioritize, and its wealth plugins exist precisely so RIAs, broker-dealers, and TAMPs can build private versions powered by their own data and controlled by their own compliance teams. Capability that required a nine-figure partnership in May will be a product with a settings page within a couple of years.
Second, mid-sized firms carry far less integration debt, the kind that accumulates naturally across decades of systems, mergers, and acquisitions. That is no knock on the giants; coordination at their scale takes time and money, which is exactly why they hired the labs' engineers. It does mean a firm where decisions travel one floor can run the same validated playbook faster than its authors did: pick one internal workflow where hours leak. Wire the data into it properly. Keep a human owning every client-facing decision. Build the audit trail as you go, not after.
The Takeaway
Access to frontier intelligence commoditized faster than anyone expected, and the largest firms responded by moving their spending up the stack: to engineering, data, and workflow integration. The advantage in this industry no longer belongs to whoever has the model. Everyone has the model. The race is no longer a technology race. It is an implementation race, and it is won by the firms whose data can feed the model, whose workflows can absorb it, and whose governance can defend it in front of a regulator.
The most expensive due diligence in financial services history has already been done, and the conclusions are published in press releases. Treat the partnership wave as a readiness checklist rather than a spectator sport, start the unglamorous work now, and you will hold a structural advantage over everyone still waiting for clarity. For firms deciding where to begin, that is the gap our AI Foundation engagement was built to close: a fixed-scope diagnosis of where AI creates real leverage, and a build plan that puts it into production inside a quarter.