Guidewire Paid $1.7 Billion for the Pricing Models Underwriting Agents Will Call

On October 27, Guidewire agreed to buy hyperexponential (hx), a London pricing-software company, for $1.7 billion in cash plus up to $400 million more if hx's recurring revenue hits targets after closing. Guidewire sells the core systems behind much of property-and-casualty insurance: the software that issues policies, bills customers and handles claims for insurers covering buildings, fleets and business liability. Until now, its largest disclosed acquisition price was about $260 million, for Cyence in 2017.

What Guidewire bought is a model library. hx is where specialty insurers keep their pricing logic as versioned, approved Python code, meaning the actuarial models that set what an unusual commercial risk should cost. The agent features hx markets on top of that library are mostly in preview. The models themselves are in production at named insurers, and rebuilding them somewhere else is a long actuarial project that also needs fresh internal approval. So the library is both the asset and the switching cost. High confidence.

The terms, and closing language that doesn't reconcile

Late 2029 gets crowded. Guidewire had about $1.2 billion in cash and investments at July 31, so it borrowed. The term loan matures December 2, 2029, with most of the principal still outstanding, and Guidewire's $690 million of 1.25% convertible notes come due the month before. Two refinancings a month apart make another deal this size harder to do before then. Moderate confidence.

hx's shareholders carry the earnout, and what they're betting on is Guidewire's sales force. The earnout pays on recurring revenue. Nothing disclosed ties it to anyone staying employed; retention money sits in its own pool inside the base price. hx's investors (Highland Europe and Battery led its rounds; a16z joined the Series B) collect the last $400 million only if revenue grows while Guidewire's salespeople are the ones driving it.

The base price works out to roughly 20 to 40 times revenue, for a business that still loses money. An analyst on the call floated about $80 million in annual recurring revenue. Management didn't confirm it, but take it at face value and you get about 21x. hx's UK statutory accounts show £34.8 million of revenue for the year ended January 31, 2026, up 42% and mostly subscription, with a £6.2 million operating loss. At roughly $1.30 to the pound, that puts the multiple in the high 30s. These are different measures at different dates, so read the range as calibration, not a precise number.

The filing and the call disagree on closing. The 8-K says the deal "is not subject to any conditions to closing." On the acquisition call, Cooper said closing was "subject to customary closing conditions," which normally means regulatory sign-off and no material adverse change at the target. The investor deck says only that there's no financing condition, which doesn't settle it either way. Under the 8-K's wording, Guidewire is bound once it signs. Under Cooper's, the deal can still stop. None of the three documents mentions US antitrust review or UK national-security review. The merger agreement should arrive with the quarterly report for the period ending October 31. Until you've read it, treat this deal as signed, not done.

Where hx sits in specialty underwriting

Underwriting means deciding whether to insure a risk and at what price. Most of Guidewire's history is in what management calls flow business: small-business and auto policies, priced by rating tables run against thousands of similar applications.

hx's customers write the other kind. Think an aviation fleet, a specialty liability account, a risk standard insurers won't touch. That business goes to the London market around Lloyd's, or to excess-and-surplus (E&S) carriers, which operate under looser rate and form rules. Guidewire CEO Mike Rosenbaum called these risks "almost completely bespoke per risk." An underwriter prices each one against an actuary's model, and that model has usually lived in a spreadsheet. hx pitches its Actuarial Agent partly as a way to convert spreadsheet tools into Python, which tells you the spreadsheet is still its main competitor.

In place of the spreadsheet, hx runs a governed pipeline:

• Authoring. Actuaries write pricing and appetite logic in Python (appetite meaning which risks the insurer wants). Each release is versioned and approved.
• Deployment. An approved model becomes a calculation engine that underwriters, portals, APIs and agents call. The system records which version priced which risk.
• Workflow. Workflow Builder sorts submissions, adds outside data, checks appetite, prices, and routes anything beyond an underwriter's authority to someone senior. Approval rules and human checkpoints are configured step by step.
• Portfolio. Dashboards track rate change, concentrated exposure and loss ratios (claims paid as a share of premium), and those feed back into pricing.

The evidence for the models is strong. Aviva disclosed 23 live hx models wired into its Touchstone underwriting portal. Guidewire also names Allianz, Sompo, Beazley, Markel, HDI, Cincinnati Financial, Convex and Bowhead. Be careful with its "$75 billion in commercial premium associated with underwriting decisions." That figure counts the insurance business flowing through hx. It is not hx revenue.

The evidence for the AI layer is thin:

hyperoperator's design is the clearest explanation of why the library was the purchase. It orchestrates specialist agents, but the price comes from a deterministic calculation engine, a versioned model an actuary approved. The language model gathers, extracts and explains. Configured rules decide what proceeds without a human. If you've designed agent permissions, you'll see four separate states: the agent's recommendation, the model's execution, the authority delegated to the underwriter or agent, and the bound policy, which is the point where the insurer owes claims. hx owns the second and third. An underwriting agent without a governed price to call can assemble a file, but it can't quote.

Why Guidewire bought a partner it overlaps with

Guidewire already had products here. UnderwritingCenter, still in Early Access, targets mid-size, large commercial and E&S accounts with agents that clean up submissions, check appetite and hand off to PolicyCenter, Guidewire's policy system, for quoting. PricingCenter, launched October 2025, builds statistical, machine-learning and rating models without code. Quantee, bought in April 2025 for about $28 million net, added dynamic pricing.

Rosenbaum conceded "there's definitely some overlap in terms of the vision." President John Mullen drew the line: Quantee and PricingCenter for flow, hx for complex risk. The line holds up, because the in-house products couldn't supply three things:

1. Models already running in the target segment. No product launch migrates Aviva's 23 models.
2. The actuary as the user. PricingCenter is no-code. hx is code-first, and that's how specialty pricing teams already work.
3. A tested relationship. Guidewire announced hx's addition to its partner program in October 2024, and hx has a Marketplace connector. Marcus Ryu, Guidewire's cofounder, has sat on hx's board since January 2024 through Battery's Series B. Guidewire also bought ProNavigator out of its startup program in November 2025. Partnering first and buying later is now a Guidewire pattern.

What the partnership proved was commercial fit, more than integration. The connector links hx to PolicyCenter, not UnderwritingCenter, and no production customer using it has been named.

That makes the integration problem a design problem. Two products now claim the specialty underwriter's working day: UnderwritingCenter, and hx's workflow and triage tools. After close, watch which surface underwriters actually live in, and watch where approval rules get configured. Those rules are an agent's permission to act, so the product that holds them will lead Guidewire's agentic underwriting, whatever the roadmap slides show.

What premium-based pricing does to the combined business

Guidewire's 10-K says core subscriptions are generally priced on direct written premium (DWP), the premium insurers write on policies the software manages, typically on five-year initial terms. Rosenbaum described hx's model as premium-based too, not per seat. No hx price list is public.

• Agent productivity raises revenue. Per-seat revenue would shrink as agents cut underwriter workload. Premium-based revenue grows when insurers write more with the same staff. Rosenbaum compared the coming shift to what AI coding tools did to software teams, and this pricing lets Guidewire sell that shift without cannibalizing itself. It also lets a DWP customer add hx on a metric it already pays. High confidence.
• More exposure to the insurance price cycle. Specialty rates swing between hard markets (rising) and soft (falling), and premium-linked revenue swings with them. The 10-K names DWP changes as a revenue factor without sizing it. A specialty-heavy acquisition adds to that exposure. Moderate confidence.
• The earnout may ride the market. If hx revenue tracks premium, hitting three-year targets depends partly on specialty rates, which neither side controls. Speculative; the mechanics are undisclosed.

Which vendors' positions shift

The underwriting workbench, and the intake layer that feeds it, are moving inside bigger platforms. Applied Systems bought Cytora in September 2025. Duck Creek bought Send in July 2026. Both connect to hx: Cytora sends it submissions to price, and Send's workbench calls hx pricing. After close, hx belongs to Guidewire, Duck Creek's most direct core-systems rival.

Check partner neutrality rather than assuming it. In a piece on Bayer's radiology platform, I argued a marketplace stays neutral only while the owner's core revenue sits outside the catalog. Applied here, that would predict pressure on Earnix and Akur8. But Bayer kept adding partners after its acquisition, and no departure there traces clearly to ownership. So watch the evidence instead of forecasting. Do Earnix and Akur8 keep their connectors and their slots at Guidewire's Connections conference? Does Duck Creek's hx partner page survive the close?

Where the capital moved and who to watch

Capital moved toward the pricing and decision layer for complex commercial risk. Agentic underwriting needs a governed, versioned price and configured approval rules to call, and a core vendor paid on premium wants that layer inside its own product, where agent productivity raises revenue instead of eliminating seats. Guidewire bought models already deployed at specialty insurers, plus a three-year bet that its own channel can grow them.

Watch Duck Creek, which needs a new answer for bespoke pricing. Watch Akur8, an independent pricing vendor that is already a Guidewire portfolio company and an hx partner. Watch Earnix, whose chairman has said a strategic sale remains possible alongside an eventual IPO. And watch Sixfold and Kalepa, whose route through Guidewire customers just narrowed. Then read the merger agreement before you assume any of this closes.