$110M and a Valuation They Won't Tell You: AI's Fintech Moment Is Getting Serious

The Number That's Missing Is the Point

4,208. That's how many new SIC 64.20 companies - holding companies and investment vehicles, the structural backbone of fintech capital formation - were registered in the UK in Q2 2026, according to AIBD's analysis of Companies House data. Sounds like a lot. It isn't. That figure represents a 41.2% collapse versus the prior period. The vehicles that typically absorb and deploy fintech VC are forming at less than half their previous rate. And yet the cheques keep getting larger.

Something is bifurcating. Badly.

Goldman Picks a Side

On June 24, New York-based Taktile closed a $110M Series C led by Growth Equity at Goldman Sachs Alternatives, with participation from Balderton Capital, Index Ventures, Tiger Global, and Y Combinator. Total raised: $184M. The company builds AI decisioning infrastructure for financial institutions - credit underwriting, fraud detection, loan approvals - the kind of regulated, audit-facing workflow where a model error doesn't just embarrass you, it generates an enforcement action.

Notably, Taktile did not disclose its valuation. In a market that normally treats private marks as a marketing asset, that's a conscious choice. As the June 24 funding roundup from TechStartups noted, some of the strongest companies are "choosing to emphasize deployment metrics and institutional logos over vanity numbers." Goldman's name on the cap table is the valuation signal. Everything else is noise.

Let's Do the Maths

Pull back to the macro. Q1 2026 saw $300 billion in global venture investment across roughly 6,000 startups, per Crunchbase - up more than 150% year-over-year, an all-time record. AI captured $242 billion of that, or 80% of total global VC. Four rounds alone - OpenAI ($122B), Anthropic ($30B), xAI ($20B), Waymo ($16B) - absorbed $188 billion, or 65% of the entire quarterly total.

Now remove those five largest deals. PitchBook and NVCA calculated that doing so would slash deal value by 73.2%. Seventy-three percent. The headline number is a statistical illusion generated by a handful of frontier lab mega-rounds. The market underneath it is considerably thinner.

This is the 2000 dot-com bubble in negative photo. Back then, Cisco peaked at 131x trailing revenue in March 2000 and the Nasdaq lost 78% over 30 months. Today's concentration isn't sector-wide euphoria - it's a two-tier system where the top five firms captured 73.1% of new VC fund commitments. Capital abundance at the crown. Scarcity almost everywhere else.

Fintech Is the Canary

Global VC into fintech startups totalled $12 billion across 751 deals year-to-date through early April, per Crunchbase. That's not a boom number. It's a triage number. The theme threading through 2026 fintech deal flow, per tracker data, is vertical AI applied to regulated finance workflows: tax preparation, Medicare navigation, commercial insurance underwriting, fraud detection, B2B procurement.

Taktile fits that thesis precisely. Financial institutions are no longer funding horizontal AI tooling for general model deployment. They're funding domain-native control layers that can withstand audits. The difference matters enormously. A generic LLM wrapper breaks under regulatory scrutiny. A purpose-built decisioning engine with explainability and compliance logging does not.

Then there's the structural signal from Companies House. 4,208 new SIC 64.20 holding and investment vehicle registrations in Q2 2026 - down 41.2% on the prior period. These are the entities through which fintech capital is structured, deployed, and recycled. When their formation rate craters, it means the institutional plumbing of fintech is contracting even as the headline funding numbers look strong. Charlie Munger used to say invert, always invert. Invert the funding boom and you find a formation drought.

The IPO Queue Is the Other Shoe

Anthropic filed its draft S-1 on June 1 at a $965 billion valuation - on roughly $47 billion in annualised revenue, a 21x multiple. OpenAI followed on June 8 at $852 billion, targeting a public listing above $1 trillion. SpaceX began trading on June 12 under SPCX and rose above $2 trillion in market valuation, the largest IPO in history.

The combined pipeline demand from the five largest AI listings alone exceeds the entire 2025 US IPO market. Index providers are scrambling: Nasdaq implemented a Fast Entry rule effective May 1, 2026, allowing mega-cap IPOs to join the Nasdaq-100 within 15 trading days if they rank among the 40 largest components by market cap.

Public-market stress-testing is coming. Anthropic's listing in H2 2026 will be the first real test of whether 20-30x revenue multiples survive contact with quarterly earnings calls, short sellers, and retail holders who bought at IPO and need a story to tell at Christmas. As Wedbush analyst Dan Ives put it, if the marquee AI listings stumble, "there is no sugar coating on that."

Polymarket currently prices an actual OpenAI public listing at roughly 40% probability by end of December 2026. The market is treating the confidential S-1 as a distant option, not an imminent event.

The Figure AI Problem

The real bubble risk isn't in the frontier labs, where at least revenue is scaling fast enough to compress multiples. It's in the application layer. AI robotics companies are trading at nearly 400x revenue on demo hype. Thin AI wrapper startups without proprietary data or distribution are raising Series As on the ambient warmth of the Anthropic glow.

Paying restaurant prices for a microwave meal. That's the application-layer AI trade right now.

Taktile's decision to skip the valuation disclosure is actually a blueprint. In a market where the number itself has become the product, building something that regulators, compliance teams, and Goldman's credit committee will trust is a different and harder pitch. So is not bragging about the multiple.

The Prediction

By Q3 2026, the 41.2% drop in SIC 64.20 company formations will show up in downstream fintech deal flow as a visible funding gap - particularly at Series A and B, where the holding-company infrastructure matters most. Simultaneously, at least one of the marquee AI IPOs will price below its last private round valuation. When it does, the application-layer wrapper trade corrects sharply. The infrastructure layer - Taktile, decisioning engines, compliance-native AI - holds. Because Goldman didn't buy the hype. Goldman bought the audit trail.