The Short Answer
Block applied on 8 September 2026 for Builders Bank & Trust, an uninsured national trust bank that cannot take deposits or lend, and the OCC signed Revolut's preliminary conditional approval on 2 September 2026. No accounting or ERP vendor appears on American Banker's 2026 charter tracker. The lock-in is the data layer, not the charter. With ibl.ai you own all the code and the data.
A banking licence changes who holds the money. It does not change who holds the record, and the record is the part that is expensive to move.
Which software and payments companies have actually applied for a bank charter?
Four filings carry the story, and each is a different instrument.
Block announced on 8 September 2026 that it had applied to the Office of the Comptroller of the Currency to establish Builders Bank & Trust, N.A., a national trust bank offering custody and related fiduciary services, including for bitcoin and stablecoins.
It would be uninsured, and would not take deposits or make loans. Block's Square division already holds a bank: it obtained a Utah industrial loan charter in 2020 and opened Square Financial Services in 2021.
PayPal filed on 15 December 2025 with the Utah Department of Financial Institutions and the FDIC to establish PayPal Bank, a Utah-chartered industrial loan company.
The release states PayPal has distributed more than $30 billion in loans and working capital to over 420,000 business accounts worldwide since 2013 — lending already underwritten on platform data, through partners.
Stripe's stablecoin subsidiary Bridge applied in October 2025 and was conditionally approved on 12 February 2026 for a national trust charter.
Revolut filed a de novo application on 10 March 2026, and the OCC's preliminary conditional approval letter for Revolut Bank US, N.A. is dated 2 September 2026.
The volume is the real signal. The OCC received 40 de novo applications in 18 months, against 48 across the fourteen years from 2011 through 2024.
Does a bank charter mean your accounting software is about to become your bank?
Not on the evidence, and the distinction matters more than the headline does.
American Banker's tracker of 2026 charter activity lists roughly thirty applicants. They are neobanks, payments companies and digital-asset firms. No accounting, ERP or general business-software vendor appears on it.
Charter type is the second correction. A national trust bank is not a bank in the sense a CFO means: Block's proposed entity would be uninsured and is barred from deposits and lending.
Approval status is the third. Revolut's is preliminary conditional — the OCC letter records that its FDIC deposit-insurance application is still under review and that final authorization to open has not been granted.
PayPal's ILC application is pending and contested. The Bank Policy Institute and ICBA filed comments on 30 January 2026 urging the FDIC to withhold deposit insurance from ILC applicants.
What software companies are actually doing is quieter. Revolut told the OCC it already serves U.S. customers "through FDIC-insured bank partners," and wants the charter to do it "at a lower cost and with greater efficiency than through its existing partner bank model."
That is the pattern: banking arrives inside software through a partner bank long before anyone files for a charter. The filing is the moment the partner is removed, not the moment the product appears.
Why is the data layer, not the banking licence, what actually locks a finance team in?
Because the vendor had the data first, and the licence only changes what it can do with it.
By the time a platform applies for a charter it already holds the transaction history, the counterparty list, the payment timing and the billing relationships. PayPal's $30 billion of lending was underwritten on exactly that, years before the ILC filing.
Switching a deposit account is a week of work. Switching the system that holds five years of invoices, categorisations, approvals and reconciliations is a project with a budget.
Add AI and the asymmetry widens. The vendor's agents read the vendor's copy of your data, and the prompts and outputs are retained under the vendor's policy.
The trail of what the model saw is theirs to produce — the absence-of-a-trail problem that decides whether an AI-assisted number survives an audit.
None of that appears in a charter application, because none of it is a regulated activity. It is a product decision the customer never gets to vote on.
What changes for a CFO when the vendor holding the ledger also holds the money?
Concentration, in one direction, with the exit priced accordingly.
The honest version first: a vendor bank is often better for the customer. Removing a partner bank removes a margin layer and a reconciliation step, which is precisely the efficiency Revolut cited to the OCC.
The risk is not that the vendor becomes a bank. It is that software, records, AI and now credit and payments sit behind one commercial relationship, and each additional layer raises the cost of leaving any of them.
Credit is the layer that converts inconvenience into dependence. If the working-capital line is underwritten on platform data and priced off platform behaviour, changing systems is also a refinancing.
Concentration can also arrive without a filing. Capital One completed its acquisition of Brex on 7 April 2026, so a finance-software platform became part of a bank by M&A, on a timetable its customers did not set.
The defensible position is not vendor purity. It is keeping one thing — the data layer — on infrastructure the vendor does not control, so every other decision stays reversible.
Which questions should a finance team ask before a vendor becomes its bank?
Four, and each has a document-level answer rather than an opinion.
- What exactly can you export, and how? Full history or a rolling window; machine-readable with schema and relationships intact, or a PDF pack; on your schedule via API, or on request. Ask for a test export now, not at renewal.
- Where does the AI over your data run, and who keeps the record? Whose infrastructure, what is retained, for how long, and whether you can inspect the code that handles it rather than accept a description of it.
- Is credit separable from software? If the vendor extends credit off your platform data, write down what an exit costs in refinancing terms before you need it.
- What survives a change of control? Assignment clauses, data rights and pricing protections are cheap to negotiate at signature and unobtainable the week an acquisition is announced.
None of this requires predicting which vendor charters a bank. It requires knowing, in writing, what leaving costs today.
How does ibl.ai keep the data layer on infrastructure you own?
By running the platform, and its source, inside your own perimeter, and reading your systems where they already are.
With ibl.ai you own all the code and the data.
Codebase transfer is a one-time fee plus a perpetual licence: full source code, self-hosted, so the components that touch your ledger are ones your team reads and changes rather than vendor behaviour you observe.
The ontology runs inside your network and exposes the systems you already operate — systems of record, warehouses, line-of-business SaaS, document stores — to agents as one role-scoped MCP server, queried in place with no data extraction.
Every tool execution is written to a tamper-resistant audit log.
It is model-agnostic across any LLM, billing is usage-based with no per-seat pricing, and you can deploy anywhere: your own cloud, on-premise, GovCloud, or a fully air-gapped network.
The practical consequence is that a vendor's roadmap — including a decision to become a bank — stops being a question about your data.
1.6M+ users across 400+ organizations run the platform this way, including NVIDIA, MIT, and Syracuse University.
ibl.ai is family-owned and operated from New York, NY.
Related reading: finance AI with no audit trail is evidence, not speed — the three controls that make an AI-assisted number defensible; who owns your data when you use ChatGPT or Copilot — the same question at the model layer; and financial AI agents ship as SKUs, integration doesn't — why the integration, not the agent, is the moat.
Sources: Block's application and Square Financial Services' history from Payments Dive; PayPal's filings and lending figures from its 15 December 2025 release; Bridge's conditional approval from Banking Dive; Revolut's application date, approval status and partner-bank rationale from OCC Corporate Decision #1390; the applicant list from American Banker; the 40-in-18-months figure from PYMNTS; the ILC objections from BPI and ICBA; the Brex close from Capital One.