Schwab's Claude exclusive is a standards play with an expiry date

The most telling name on Anthropic's charter list for Claude for Financial Advisors belongs to Vanguard, the firm that says it will not use the product for the thing every other name on the list is using it for. Vanguard appears among the launch cohort Anthropic announced on Sept. 14, and it will not deploy the system in RIA custody or with its in-house advisors, directing questions about the arrangement back to Anthropic. What Anthropic's own release says about Vanguard runs a single sentence: the firm is feeding information on its model portfolios and advisor investment solutions into Claude so that advisors can draw on its research, portfolio construction and asset allocation work when making decisions for clients.

That is a research feed with a charter badge on it, and it is worth holding next to Schwab Advisor Services' claim that it is first, and exclusive among RIA custodians, to use Claude for Financial Advisors. Anthropic announced the release on Sept. 14 with a list of charter members that included Schwab, and Schwab described itself as the first and only RIA custodian included in the early release. Both statements can be true simultaneously: the exclusivity is scoped to a category, and the categories appear to have been drawn by how the launch's partner list was segmented rather than by anything the custody market decided.

Anthropic's head of asset and wealth management, Peter Nolan, framed the Schwab piece as reach: "Collaborating with Schwab and the 16,000-plus independent RIAs it serves brings Claude to the firms doing that work every day, starting with a clear view of what has changed for each client." Joel Bruckenstein, the producer of the T3 conferences, read the same arrangement with more conviction: "Schwab is in the driver's seat now. Schwab is the gorilla, and they're going to get first crack at taking it to 16,000 advisors." The same coverage carries the opposing reading, that the offering is generic and late-coming, and Fidelity, asked where its RIA custodian stands on Claude or on artificial intelligence generally, declined to answer.

The 16,000 travels in two units through that coverage, RIAs in one quote and advisors in the other, and those are not the same population. Nobody in the material treats the distinction as load-bearing, which is its own small measure of how loosely the launch has been framed.

When the Schwab arrangement surfaced, the case was that the custodian had taken a zero-revenue seat at the front of Anthropic's RIA queue, and that the seat would be worth whatever the plumbing underneath it turned out to be worth. Nothing since has attached a dollar figure or a stated term to the arrangement. That silence is consistent with what a category exclusive on a horizontal product is: a real asset, but a narrow one. Going first to a vendor's roadmap buys design influence, an escalation path into support, and a reference logo; it does not buy scarcity, because Anthropic's commercial interest is selling Claude to every firm with a client meeting to prepare, and the release went out to a charter list the vendor assembled and defines. If a competing custodian wants Claude in the next cohort, the binding constraint is Anthropic's schedule, not Schwab's consent.

The strongest version of the bull case is that Schwab's workflow shapes the product while competitors are still waiting, so that by the time Fidelity or a smaller custodian buys in, the tuning assumes Schwab's process map. That argument is real, and it has a ceiling. Workflow tuning is portable software; vendor roadmaps get repriced at every release, and Anthropic has shown with this launch that it will seat an asset manager and a custodian in the same cohort and let each describe its participation in whatever terms it likes.

Which is not the same as saying Schwab bought nothing. The valuable half of arriving first is almost never the model; it is the right to write defaults: which accounts a system may read, which fields it may summarize, what an advisor has to sign before a recommendation leaves the desk, and what the file shows when a regulator asks how the recommendation got made. Publish a permissioning scheme to 16,000 firms and competitors inherit a choice between copying your rules and explaining to their advisors why their version differs. That is a standard. A standard is worth more than a head start, and it closes on a clock the standard-setter does not control.

Vanguard's posture is the most informative of the three, because Vanguard is doing the expensive thing and declining the cheap one. It is acquiring Altruist, the RIA custodian, in a $4.6 billion deal announced in early September, pairing a firm with $11.1 trillion in assets with a custodian of 348 employees. Renting Claude for RIA custody would have cost far less and produced a faster demo. Feeding the model a research-content arrangement instead points to a firm that wants the advisor-facing layer to sit inside the custodian it owns rather than inside a model it rents. That is an inference rather than anything the firm has said; the coverage does not spell out Vanguard's AI plans for Altruist.

The pattern underneath it has tracked since Vanguard moved on Altruist: buyers in wealth are paying for distribution rails rather than client books, and a custodian is the rail in its plainest form. A firm that owns the workflow can swap vendors without asking permission; a firm that rents the model and rents the connector negotiates from a seated position, and the leverage only tightens as the model becomes more central to the daily work. Fidelity's position is unknown from the material at hand—the coverage records its non-answer and nothing more—so the comparison for now runs between the two custodians who have shown their cards.

The AI premium in wealth management has moved from model quality to the governed client record, and whoever owns the permissioned data owns the meeting. The fight has also moved into the advisor-client conversation itself, where the constraint is no longer what a model can generate but what the firm can lawfully put in front of it. That framing makes the charter list read as a set of data-access negotiations rather than model deployments, which is precisely what Vanguard's one sentence describes.

The scarce input here is the position file, not intelligence. Cost basis, corporate actions, held-away accounts, the statement, and the meeting note that has to be written, stored and defended are the raw material worth arguing over, and any custodian can staple a chat box to an advisor desktop inside a quarter. The firm that can hand a model a governed slice of client data with an audit trail, and let advisors use it in the meeting, holds something rivals would have to rebuild rather than license. A layer up, the same logic shows up in tax, where the platform wrapping other managers' models captures the value of the models it wraps; the model commoditizes, the connector is where the rent sits.

Schwab's exclusive is a standard-setting play being marketed as a technology lead, and those two claims expire on different schedules. A technology lead compounds; a standard-setting window closes. Anthropic will name another cohort of financial firms, probably within quarters, and the first test is whether a second RIA custodian appears on it. The second test is whether Schwab's data and permissioning rules reach its 16,000 firms before a competitor's rules reach theirs, or after. If it is after, the exclusive will have bought a quarter of headlines, a talking point for the conference circuit, and a first-mover story that its rivals answered by buying the rail underneath the conversation instead of the model on top of it. Vanguard's $4.6 billion answer is the one that will still be on the balance sheet when the next charter list lands.