The SaaS funding freeze that H1 just punched through

Pre-AI era SaaS startups are getting starved of capital right now. Investors have largely moved on, chasing AI-native tools and foundation model bets while leaving older software businesses to justify their existence without a compelling answer to the question everyone is asking: can an LLM just do this instead?

Against that backdrop, H1’s $40 million raise from CVS stands out. H1 is a nine-year-old healthcare data platform that sells detailed physician information to pharma companies, hospital systems, and health insurers. CVS is not a venture fund hunting for moonshots — it’s a $370 billion healthcare conglomerate making a deliberate infrastructure bet. That distinction matters. When a strategic incumbent cuts a $40 million check into a nearly decade-old SaaS company, it’s not funding a feature. It’s buying access to data that would take years and enormous resources to replicate independently.

H1 co-founder and CEO Ariel Katz makes the case bluntly. Workflow SaaS, he argues, is now vulnerable — “you could vibe code that,” he told TechCrunch. What you cannot replicate with a prompt is a proprietary dataset built over years of grinding collection and curation. Katz says he has no anxiety about Anthropic’s Claude displacing what H1 does, because the underlying data — granular, structured intelligence on physicians — simply doesn’t exist anywhere an AI model can reach.

That’s a self-interested argument from someone running exactly the kind of company that benefits from it being true. But the CVS deal suggests at least one major industry player agrees with the logic.

The real story here isn’t the fundraise headline. It’s what the deal reveals about how capital is being selectively deployed. Traditional industries with complex, regulated data environments — healthcare chief among them — are not abandoning SaaS. They’re narrowing down to SaaS companies that own something irreplaceable. H1 cleared that bar. Most of its peers are still waiting to find out if they can.

The ‘vibe coding’ argument: bold claim or convenient narrative?

H1 CEO Ariel Katz made a pointed claim when his company closed a $40 million round led by CVS: workflow SaaS is now replicable by AI through tools like vibe coding, but data-centric platforms are not. The distinction matters because Katz isn’t speaking abstractly — he’s describing the exact fault line investors are using to separate fundable companies from ones facing slow obsolescence.

The self-serving element is real and worth naming. H1 sells detailed physician data to pharma companies, hospital systems, and health insurers. When Katz says AI can’t replicate what his company does, he’s simultaneously pitching his own business model as the template for SaaS survival. Readers should hold that context in view.

But the underlying argument holds up under pressure. Vibe coding tools — AI-assisted development environments that let users generate functional software through natural language prompts — genuinely compress the time and cost required to build workflow automation. A scheduling tool, an approval routing system, a reporting dashboard: these can be approximated faster and cheaper than they could two years ago. The switching costs for workflow SaaS were always thinner than vendors admitted, and AI is exposing that.

Data assets don’t follow the same logic. Curated, proprietary datasets — built through years of sourcing, cleaning, credentialing, and maintaining relationships — can’t be vibe coded into existence. H1’s physician database reflects nine years of accumulation. Claude can’t train its way to that. Neither can a well-prompted competitor.

Katz said directly that he doesn’t worry about Anthropic’s Claude replicating H1’s capabilities. That confidence isn’t arrogance — it’s a structural argument. The moat isn’t the interface or the workflow layer sitting on top of the data. The moat is the data itself, and that distinction is driving a visible shift in how investors evaluate SaaS durability in 2025.

What most coverage is missing: the ‘data moat’ thesis isn’t new, but the stakes just got higher

The data moat concept has been a cornerstone of competitive strategy theory since long before ChatGPT entered the public vocabulary. What’s changed is the consequence of getting it wrong. When generalist AI models couldn’t generate coherent paragraphs, a SaaS company with a modest proprietary dataset had a reasonable buffer. Now that buffer is gone for most players — and the line between “defensible data business” and “workflow tool that GPT-4o can replace” has become a survival threshold.

Most coverage of H1’s $40 million raise from CVS treats it as a proof point that pre-AI SaaS can still attract institutional capital. That framing buries the more important question. H1 CEO Ariel Katz told TechCrunch that AI cannot easily replicate what his company does, because H1 is a data provider at its core rather than a workflow product. He was direct about the contrast: “If you’re a workflow SaaS company, you could vibe code that.” He also said he doesn’t worry about Claude doing what H1 does.

That’s a self-interested position from someone whose nine-year-old company depends on selling detailed physician data to pharma companies, hospital systems, and health insurers. But the logic holds — and the healthcare context is where the real story sits.

Healthcare provider data isn’t defensible because it’s large. It’s defensible because building it requires navigating HIPAA compliance, credentialing systems, relationship-driven sourcing with hospital networks, and state-level regulatory variation that changes constantly. No foundation model trained on publicly available text can shortcut those structural barriers. Claude cannot cold-call a health system, negotiate a data-sharing agreement, or maintain the compliance infrastructure required to handle sensitive provider records at scale.

That’s the missing layer in most funding coverage: the moat isn’t the data itself, it’s the institutional friction required to acquire and maintain it. For healthcare specifically, that friction is high enough to make replication genuinely expensive and slow — exactly the conditions that make a data asset defensible when AI commoditizes everything else.

CVS as investor: strategic bet, not just a check

CVS leading H1’s $40 million round is not a passive financial bet. CVS operates one of the largest pharmacy networks in the United States, runs MinuteClinic, and has been aggressively building out its healthcare services infrastructure through acquisitions like Aetna and Oak Street Health. A company with that footprint does not write a lead check into a physician data platform out of portfolio diversification instinct.

H1 aggregates detailed profiles on doctors — their clinical focus, publication history, prescribing patterns, and institutional affiliations — and sells that intelligence to pharma companies, hospital systems, and health insurers. That is exactly the kind of structured, hard-to-replicate data that CVS needs as it pushes deeper into care delivery and pharmacy benefit management. CVS gets direct value from knowing which physicians are driving prescribing decisions, which specialists anchor referral networks, and where clinical influence actually concentrates.

Strategic corporate rounds rarely arrive without strings. These deals typically include data licensing agreements, preferred partnership terms, or rights to co-develop products — arrangements that never appear in the headline funding number. The $40 million figure almost certainly understates what CVS actually extracted in return.

For SaaS founders watching this, the lesson is structural. Venture capital is pricing AI replaceability into every deal right now. A workflow tool that automates scheduling or surfaces reports sits in genuine danger of being vibe-coded out of existence. But an incumbent like CVS cannot build nine years of physician relationship data from scratch, and it cannot buy that asset cheaply on the open market. H1’s CEO Ariel Katz put it plainly: no AI model is going to replicate what H1 has collected on physicians. CVS apparently agreed enough to lead the round.

The funding pathway this opens is real. SaaS startups sitting on proprietary industry data should be identifying which incumbents are most competitively exposed without it — and approaching them as potential lead investors, not just customers.

The playbook for SaaS startups that want to survive the AI era

H1’s $40 million raise from CVS gives SaaS founders a concrete framework for stress-testing their own defensibility. Apply two questions in sequence. First: could a well-prompted AI model replicate your core function within six months? Second: do you own data that would take years of domain relationships and ground-level access to reconstruct? If the answer to the first is yes and the second is no, the business is exposed.

H1 CEO Ariel Katz draws the line between workflow tools and data platforms. “If you’re a workflow SaaS company, you could vibe code that,” he told TechCrunch. Workflow software — scheduling tools, approval chains, reporting dashboards — is functionally a set of instructions that AI coding tools can now reproduce cheaply and fast. A data asset built over nine years through relationships with pharma companies, hospital systems, and health insurers is something different. Claude cannot cold-call physicians and accumulate verified clinical and professional data at scale. The moat is not the software layer; it is the corpus underneath it.

The urgency here is real. As AI coding tools mature, the cost to replicate a workflow product drops toward zero. Founders who have not yet built a proprietary data layer are not facing a future problem — they are facing a narrowing window. Every month spent iterating on features without asking what data only your company can own is a month of compounding exposure.

Investors are already running this test. Founders who cannot articulate a data defensibility story with the same precision they bring to a product roadmap will struggle to raise. H1’s round is evidence that capital is still available for companies that pass the test — not as workflow vendors with a data side project, but as data businesses that happen to have built excellent software on top. That is the posture pre-AI SaaS companies need to adopt, and the time to do it is before the next fundraise, not during it.

What this means for the broader SaaS market in 2025

H1’s $40 million raise from CVS won’t reverse the broader SaaS funding slump, but it hands the market a working template: go niche, own your data, and plant yourself inside a regulated industry where customers can’t easily leave. That combination — not another AI-powered dashboard — is what serious investors are now stress-testing before writing checks.

The growth-metrics era is effectively over. For most of the last decade, venture capital flowed to SaaS companies that could show steep user acquisition curves and expanding ARR. Defensibility was a talking point, not a prerequisite. That calculus has shifted. Investors evaluating SaaS deals in 2025 are asking a harder question: who actually owns the underlying data asset? A great user experience built on someone else’s data is a product, not a moat.

That distinction exposes a specific class of startup to serious risk. Companies that layered compelling interfaces on top of data controlled by platform partners — CRMs pulling from LinkedIn, analytics tools dependent on Google or Meta APIs, productivity apps tethered to Salesforce objects — are structurally vulnerable. The moment a platform partner or a well-funded AI tool decides to compete directly, those startups lose their core value proposition overnight. They own the design; they don’t own the asset.

H1’s position is structurally different. The physician data it sells to pharma companies and hospital systems took years to compile and verify. Ariel Katz, the company’s co-founder and CEO, makes the point bluntly: an AI model like Claude can replicate a workflow, but it cannot replicate nine years of proprietary data collection in a compliance-heavy industry. That’s not a feature gap — it’s a structural barrier.

The startups that survive the current shakeout will share H1’s core characteristics. They will operate in verticals where data is hard to collect, regulated, and deeply embedded in customer operations. They will be the source of record, not a layer on top of one. Every other category faces a credible replacement threat — from AI tools, from platform incumbents, or from a competitor who simply builds the same workflow faster and cheaper.