The SaaS sector saw one of its sharpest repricing cycles in 2024–2025, with hundreds of billions in market cap erased across horizontal software companies driven by AI disruption. The reset isn’t temporary; it’s structural. Horizontal SaaS — tools that every company uses — is being commoditised by foundation models. Vertical AI — tools that specific industries use — is emerging as the winner. Three tiers are crystallising, and only one of them has venture economics.
The Three Tiers
Commodity tier ($20–50/user/month): Copilot, Gemini, and every wrapper product that sits on top of a foundation model API. These are not venture-scale businesses. They’re features being bundled into productivity suites or absorbed into platform subscription tiers. Economics: high user growth, zero defensibility, compression to near-zero margin within 24 months. The graveyard is substantial — dozens of AI email tools, code review tools, and general-purpose assistant products, most dead or acquired at distress valuations.
Professional tier ($2K–15K/seat/year): This is the inflection point. Harvey (legal AI) at $14.4K/seat/year (per Sacra research). PitchBook (investment intelligence) at $12K–70K/seat/year. Dedicated vertical intelligence platforms in investment, compliance, and clinical decision support. These products are vertical, defended by proprietary data and methodology, and sticky — switching cost is material. Gross margin runs 55–75% when you control the cost of delivery. This is the venture tier.
Enterprise tier ($50K+/engagement): McKinsey, Bloomberg Terminal, Goldman Sachs, and in-house AI teams at mega-cap companies. Full white-glove implementation, regulatory compliance, integration into closed systems. Margin: 70–85% gross, but customer concentration risk is extreme. Growth depends on enterprise sales excellence, not product velocity. Venture outcomes here are possible but rare.
Why the Commodity Tier Failed
The repricing happened because the commodity tier was never viable at scale. Founders convinced themselves that horizontal AI products — write better emails, review code faster, summarise documents — were defensible. They weren’t. The moment foundation models became stable and cheap, every horizontal product became a wrapper. Wrapping is a feature, not a moat.
The $800B repricing is what happens when the market collectively recognises this.
Three Sub-Patterns in the Professional Tier
Founders and capital are now fighting for position in the professional tier. Three sub-patterns are emerging:
Vertical AI (industry-specific): Product focused on one industry. Defensibility comes from domain expertise and proprietary training data. Pricing is 5–10x higher than horizontal products because the ROI is 5–10x higher. Harvey (legal), Flume (water utilities), Crusoe (energy-intensive AI) are winning. The key: you need to understand the industry well enough that your AI agent actually outperforms a human practitioner on domain tasks, not just generic tasks.
Process AI (workflow-specific): Product focused on one broken process — hiring, customer onboarding, financial close, supplier qualification. Defensibility comes from owning the workflow so completely that switching would require redesigning the process itself. Pricing is subscription per process, not per user. Eightfold and Paradox in recruitment, Domo-era financial close tools — these are winning.
Intelligence AI (decision support): Product that helps humans make better decisions in a specific domain — investment, acquisition, customer segmentation, clinical triage. Defensibility comes from proprietary methodology, persistent knowledge graph, and calibration loops that improve with every decision made on the platform. Pricing is per decision or per analysis unit. Palantir is the mature version of this pattern. Earlier-stage players in investment intelligence and clinical decision support are building toward it. These are winning slowly because the sales cycle is long, but the retention is exceptional.
The Investment Frame
For founders in the commodity tier: migrate upstream or shut down. You cannot compete on features with a $100B foundation model company. The question to ask is simple — what happens when OpenAI ships your product as a feature? If the answer is “we’re done,” you’re in the commodity tier.
For investors evaluating AI SaaS: the question is not “is this an AI product?” It’s “which tier is this company in?” Commodity tier companies with high growth and low margin are not venture outcomes unless they’re acquisition targets. Professional tier companies with lower growth and high defensibility are. Most venture returns in the next five years will come from professional tier companies — that’s where profitability and scale intersect.
For enterprises evaluating AI vendors: you’re facing a reverse auction on the commodity tier. Use the foundation models directly; they’re cheaper and equally capable for horizontal tasks. Invest in professional tier products where the ROI is domain-specific and measurable. The commodity tier product with more features is still a commodity.
The Charaka View
Horizontal AI is over as a venture category. Vertical AI and intelligence AI are what remains. The $800B repricing cleared the field. The founders who understood this early — who built vertical depth before the wave broke — are now in a strong position. The founders who didn’t are looking for acqui-hire offers.
The next decade of software value creation will be built on domain expertise encoded in AI, not on AI encoded in general-purpose tools.