F2 Lands $5M Strategic Investment From Golub Capital

Golub Capital will put F2 inside the workflows where its own credit decisions are made, and it is attaching $5M to that choice. The September 28, 2026 strategic investment gives F2 more than capital: it adds a deployment partner managing more than $90B of capital and carrying decades of private-credit judgment into the implementation.

The agreement is designed to help Golub apply F2's AI capabilities across investment processes while using the firm's proprietary data and accumulated institutional knowledge. The parties did not disclose F2's valuation, the ownership stake, board rights, or other transaction terms, so the evidence supports a strategic investment rather than a newly named priced round.

The partnership also establishes the Global Credit AI Council, an invitation-only forum for senior executives from a small number of private-markets institutions. Golub Capital Co-CEO David B. Golub will serve as inaugural Chair, while the founding member list remains pending.

What F2 and Golub Capital Announced

Under the joint announcement, Golub Capital will invest $5M in F2 and deploy the company's technology to augment established investment processes. The stated work includes applying AI across investment workflows and making Golub's proprietary knowledge, historical experience, and investment judgment available inside the systems that support how the firm invests and operates.

That combination separates this event from a conventional software sale. Golub is becoming both an investor and a user, placing the product closer to the people, data, controls, and exceptions that determine whether an enterprise system is useful after the demonstration ends.

F2 describes itself as an agentic operating system for global credit. The platform connects financial models, data rooms, investment memos, credit files, and third-party data inside a shared intelligence layer, then applies agents across screening, diligence, underwriting, portfolio monitoring, and other firm workflows.

Why Institutional Knowledge Is the Product

Private-credit firms do not build judgment from public data alone. Their advantage also sits in declined deals, negotiated structures, past investment-committee debates, covenant decisions, sponsor behavior, portfolio outcomes, and the institutional memory carried by experienced investors.

F2's Institutional Knowledge product is built to turn that history into a structured, queryable record. The company says teams can compare a new opportunity with prior transactions, retrieve the risks and mitigants considered in earlier deals, and build new investment materials with source lineage back to the underlying files.

That architecture matters because foundation models keep changing. A lender that embeds its workflow around one provider can face migration cost, governance work, and fresh validation every time the model layer moves. F2 says its model-agnostic design lets an institution preserve its proprietary data, workflow, and audit trail while adopting different frontier or open-weight models underneath.

Golub's decision makes that thesis concrete. The manager is not only buying access to software; it is trying to encode accumulated experience without allowing that experience to become captive to a single model vendor. Whether the implementation delivers that portability at scale will be an important test of the partnership.

The Investment Comes After F2's $24M Disclosure

F2 launched as a separate company in September 2025 with an oversubscribed $10M round backed by nearly 50 investors from Don Muir's earlier company, Arc. The named backers included NFX, Left Lane Capital, and Y Combinator, and Muir said he invested alongside them.

In June 2026, F2 announced $24M in total equity funding, including a $14M seed round led by HighlandX with participation from Left Lane Capital, NFX, Y Combinator, and Torch Capital. The company said the capital would support customized agentic deployments, product expansion, and growth across private-credit funds and commercial banks.

The new $5M from Golub follows that financing, but the parties did not say whether it legally extends the prior seed round. DevCuration therefore classifies the transaction as a strategic investment and preserves the undisclosed valuation and terms as open fields rather than manufacturing round precision.

F2 Is Moving From Adoption Claims to Deployment Depth

In June, F2 reported more than 100 private-credit funds and commercial banks as customers, more than $400B in assets managed by those clients, and more than 15,000 deals analyzed during the year. It also reported 650% growth in monthly active users and roughly 5x throughput across screening, underwriting, and monitoring.

Those figures are company-reported and do not establish better investment performance. They do show the scale of the operating claim: F2 wants to become infrastructure for work where speed is useful only when the analysis remains traceable, permissioned, and defensible in front of an investment committee.

F2 says its product runs a native Excel engine, tracks formulas and source cells, and maintains an audit trail from claims back to documents. The company also reports SOC 2 Type I and Type II certification, GDPR compliance, zero model-provider data retention, and no training on customer data. Those controls are central in a market where proprietary borrower files and investment judgments cannot be treated like ordinary chatbot context.

The broader enterprise-agent market is confronting the same shift from clever answers to governed action. DevCuration's coverage of Arcade.dev's authorization layer and Ema's cross-system AI employees reflects the pressure to connect agents with real workflows while preserving permissions, context, and accountability.

Why the Private-Credit Market Raises the Stakes

Moody's expects private-credit assets under management to exceed $2T in 2026 and approach $4T by 2030. Its outlook also points to more asset-backed finance, greater competition among lenders, wider interconnection with traditional finance, and rising transparency and liquidity risks.

Growth gives credit teams more opportunities to evaluate, but it also expands the cost of inconsistent process. A faster screening memo has limited value if the system cannot show where a number came from, apply the firm's own precedent, preserve human review, and carry the reasoning forward when the team or model changes.

Golub brings scale and a mature underwriting culture to that problem. The firm's official materials report more than 1,100 employees and more than $90B of capital under management as of July 1, 2026. Its deployment can give F2 a demanding environment for testing whether model independence, institutional memory, and auditability survive the complexity of a large credit platform.

What the Global Credit AI Council Could Add

The Global Credit AI Council is intended to give senior executives a forum for discussing adoption, responsible integration, and emerging AI applications in private markets. If the Council attracts operators who are willing to compare implementation decisions rather than trade polished talking points, it could become a useful source of real operating requirements for F2.

The current announcement does not identify the founding institutions or define the Council's governance, output, or schedule. That leaves its influence unproven, but the structure reveals F2's category ambition: the company wants to help shape how global credit institutions build and govern AI, not merely sell them another point solution.

What to Watch Next

The near-term questions are operational. Golub and F2 have not disclosed the rollout timeline, the first workflows entering production, the number of users involved, or the measures they will use to judge accuracy, speed, control, and adoption.

The larger test will be whether Golub's historical judgment becomes more accessible without being simplified into a generic model response. F2 now has capital, a sophisticated user, and a direct path into the work its product was designed to support. The next evidence should come from what survives inside that workflow: source lineage, permissions, institutional context, and decisions that remain defensible when the underlying models change again.