Nooks Took $500 Million Without Diluting — AI-App Growth Capital No Longer Comes From Equity Markets

On Nov. 4, 2026, Nooks announced that General Catalyst's Customer Value Fund had committed $500 million to fund the sales and marketing push behind its revenue-agent platform, and used the same day to launch Node, its first AI personal assistant for sales. The capital arrived without a share, a board seat or a valuation: CEO Dan Lee confirmed only that General Catalyst takes a share of the returns the money helps create, and declined to disclose terms. What Nooks did publish were operating numbers that explain why the fund showed up at all — 3x year-over-year growth, 135% net revenue retention, more than 2,000 customers, and annual recurring revenue on track to nearly triple by year-end.

The closest published description of this instrument is General Catalyst's $1 billion facility with IM8, filed with the SEC in July 2026: up to 70% of monthly marketing spend is financed, and repayment comes only from the collections of the customer cohorts the money bought, capped at a fixed multiple of capital deployed to each cohort. There is no equity, warrant or convertible instrument, no maturity date, no fixed repayment obligation and no financial covenant. General Catalyst's own framing calls it an asset-liability match — the fund is paid only if the company gets paid, so it owns the downside. The proof point is Fivetran, where customer-acquisition returns moved from above 75% into the high three figures once the fund financed the upfront spend, and cash grew by more than $60 million instead of burning more than $100 million.

General Catalyst has told reporters the Customer Value Fund has deployed more than $5 billion across over 65 companies since inception, so the six publicly announced facilities are a large sample of a much larger book. The mix has shifted decisively: the four facilities announced between mid-July and early November 2026 total $2.27 billion, against $1.2 billion across the two 2025 facilities, and three of those four went to companies whose product is software that sells itself. The structural point is not size. It is that the marginal dollar of growth capital for a proven AI application can now arrive without a valuation attached to it.

Downstream is where the repayment gets verified, and that is where the asset class migrates next. The fund's return is a claim on collected revenue, and collection happens on the rails: Mastercard and Visa now run agent credentialing and machine-speed settlement programmes, while FIS ships a transaction platform that gives issuing banks know-your-agent data to authorise and monitor agent-initiated purchases. Whoever can prove which agent bought what, under whose permission, and whether it was paid holds the evidence needed to underwrite cohort paper — which is why the natural next owners of this paper are networks and banks rather than software investors.

FIS launched what it calls an industry-first AI transaction platform in January 2026, built with Mastercard and Visa, letting issuers identify and authorise agent-initiated purchases inside existing authorisation, authentication and dispute frameworks, with know-your-agent data and rollout to all its issuing bank clients by the end of the first quarter. Mastercard's June 2026 launch of Agent Pay for Machines goes further: it credentialed agents with verifiable intent, enforced programmatic spending limits and spending permissions, and settled machine payments across cards, accounts and stablecoins, with Adyen, Cloudflare, Coinbase, Global Payments, OKX, Stripe and others in the first participant cohort. JPMorgan Chase, which reported $76.2 billion of revenue in its September quarter and $64.5 billion of trailing net income, argues the same point from the issuing side: universal protocols that cross over between banks and networks are what make machine commerce bankable.

The spread is the market's answer to the question this deal sharpens. Applications that must buy their growth trade at roughly 2.5x to 6.2x revenue, while the card networks that verify and settle the resulting transactions trade at 16x to 17x, and Twilio, whose minutes every automated call consumes, sits in between at 5.4x. HubSpot is the cleanest illustration: its market value fell from $24.6 billion to $10.1 billion in the twelve months to September 2026 while trailing revenue grew 20% to $3.59 billion. That de-rating is inference, not a stated causal chain — but it is consistent with a world in which growth capital for proven AI applications no longer has to be raised from equity markets, so the equity's claim on the marginal dollar of that growth is worth less.