The Great Valuation Pivot: From the "Rule of 40" to the "Rule of Data" in the Era of Artificial Intelligence

Lloyd Price
2 minutes ago
16 min read

Executive Summary

The global technology and venture capital landscape is currently navigating a structural transformation of a magnitude not witnessed since the transition from on-premise software to cloud computing. As the industry approaches 2026, the financial heuristics that governed the Software-as-a-Service (SaaS) boom, most notably the "Rule of 40", are rapidly eroding in efficacy.

The catalytic force behind this obsolescence is the widespread commoditisation of generative artificial intelligence (AI) algorithms. With open-source foundation models now achieving functional parity with proprietary closed-source alternatives, the value proposition of pure software code has approached zero. In this new paradigm, enterprise value is decoupling from raw revenue growth and re-anchoring to a new asset class: the Proprietary Data Moat.

This report articulates the emergence of the "Rule of Data," a valuation framework that prioritises the defensibility, quality and exclusivity of the underlying data assets over the scalability of the software wrapper. We predict that in 2026, the only truly defensible data asset is high dimensional, proprietary data that cannot be scraped from the public web, with biological data emerging as the gold standard of this asset class. Through a rigorous analysis of market trends, financial disclosures from Q3 2025 and emerging architectural patterns, this report provides a roadmap for investors and operators navigating the shift from the "Rule of 40" to the "Rule of Data."

The Macro-Financial Shift: The Obsolescence of the "Rule of 40"

For over a decade, the "Rule of 40" served as the North Star for the software industry. It provided a simple, elegant heuristic: a healthy SaaS company’s combined revenue growth rate and profit margin (typically EBITDA or Free Cash Flow) should equal or exceed 40%. This metric governed boardrooms, dictated executive compensation, and determined valuation multiples. However, as we enter 2026, the utility of this metric has collapsed under the weight of Goodhart’s Law and the changing physics of AI economics.

The Historical Utility and Mechanism of the Rule of 40

The Rule of 40 was born in an era where the primary constraint on growth was sales and marketing efficiency. It offered a standardised way to compare companies at different stages of their lifecycle. A startup growing at 100% year-over-year could justify a -60% margin, while a mature incumbent growing at 10% needed to deliver 30% margins to be considered "efficient".

Strategic Trade-offs: The rule clarified the acceptable trade-offs between growth and profitability. It allowed investors to model how operational changes, such as reducing customer acquisition costs (CAC) or increasing upsells, would impact the firm’s overall health.
Valuation Correlation: Historically, companies that consistently exceeded the Rule of 40 commanded significant valuation premiums. In the public markets, high performers often traded at 12–15x EV/Revenue, compared to a median of roughly 6x for the broader SaaS cohort.Companies like Doximity (55%) and Datadog historically exemplified this elite tier.

The Collapse: Goodhart’s Law in 2025

By late 2025, the Rule of 40 began to fail as a predictive signal for long-term value. This failure is rooted in Goodhart’s Law: "When a measure becomes a target, it ceases to be a good measure".

Artificial Optimisation: To meet the 40% threshold during the capital-constrained environment of 2023–2025, many companies slashed Research and Development (R&D) budgets. While this improved short-term EBITDA margins, it hollowed out their long-term defensibility against AI disruption.
The Growth Decoupling: Data from SaaS Capital’s 2025 survey reveals a structural decline in Rule of 40 scores across the industry. The median score dropped to just 12% in Q1 2025, driven primarily by a slowdown in revenue growth that cost-cutting could not offset.
The "Lumpy" Reality of AI CAPEX: The Rule of 40 fails to account for the massive, irregular capital expenditures required for AI infrastructure. Training a foundation model requires upfront GPU investments that distort EBITDA margins for quarters at a time, making the metric noisy and unreliable.

The Rise of Capital Efficiency: The "Burn Multiple"

As the Rule of 40 fades, a sharper metric has emerged to assess the sustainability of growth: the Burn Multiple. Defined as Net Burn divided by Net New Annual Recurring Revenue (ARR), this metric isolates the capital efficiency of growth.

The New Standard: In the high-interest-rate environment of 2025 (with the Fed rate hovering around 4.0–4.25%), capital is no longer free. A Burn Multiple below 1.5x is now considered the "sweet spot," signaling that a company is generating significantly more revenue than it consumes in cash.
The AI Trap: Crucially, the Burn Multiple exposes "AI-wrapper" companies that appear to be growing fast (high revenue) but are merely reselling compute at low margins (high burn). A company might show 50% growth (satisfying the Rule of 40) but have a Burn Multiple of 3.0x due to high GPU costs, revealing it as a fragile investment.

The market has realized that high margins on a non-defensible product are temporary. This realisation is driving the shift toward the "Rule of Data," where the asset base—not just the P&L, determines value.

The Great Commoditisation: Algorithms in the Age of Open Source

To understand why data has become the primary defensible asset, one must first analyse the rapid depreciation of the algorithm itself. The year 2025 marked the "Commoditisation Event" for Generative AI, where the performance gap between proprietary models (like GPT-4) and open-source models (like Llama 3) effectively closed.

The Closing of the Performance Gap

In the early days of the generative AI boom (2023–2024), proprietary models held a significant advantage in reasoning, coding, and multimodal capabilities. Companies built business models around "access" to these superior models. By 2025, that advantage evaporated.

Open Source Maturity: Models such as Meta’s LLaMA 3 (70B) and Mistral’s Mixtral (Mixture of Experts) now rival proprietary giants in critical enterprise tasks like summarisation, classification, and code generation. The release of LLaMA 3 70B provided developers with a model that many argue is indistinguishable from GPT-4 for 90% of use cases, yet it is free to use and modify.
Specialised Dominance: Open models like Falcon (optimised for multilingual tasks) and Mistral 7B (optimised for speed and low latency) allow enterprises to deploy highly specialized agents that outperform generalist proprietary models on specific tasks.

The Economic Implications for "Wrappers"

This parity has devastated the "AI Wrapper" business model—startups that simply put a user interface on top of a third-party API.

Zero Switching Costs: The abundance of high-quality open models means that switching costs for the underlying intelligence engine have plummeted. Enterprises in 2026 are adopting "hybrid" architectures, routing simple queries to cheap, open-source models (the 80%) and only using expensive proprietary models for complex reasoning (the 20%).
Margin Compression: Companies that rely on third-party APIs have structurally lower gross margins (often 50-60%) compared to true software companies (80%+). As the underlying models become commodities, pricing power collapses, creating a "race to the bottom".

The Strategic Pivot to Sovereignty

For regulated industries like healthcare and finance, the "black box" nature of proprietary models is a liability. Open-source models allow for Data Sovereignty, companies can host the model within their own secure VPC (Virtual Private Cloud), ensuring that sensitive patient or financial data never leaves their control.This requirement for control further accelerates the adoption of open-source, rendering the proprietary algorithm less relevant than the secure, proprietary data it processes.

The New Sovereign: Defining the "Proprietary Data Moat"

In the vacuum left by the commoditised algorithm, the Proprietary Data Moat has emerged as the definitive metric for 2026. The "Rule of Data" posits that a company’s long-term enterprise value is strictly proportional to the volume, quality, and exclusivity of the data it possesses that cannot be accessed by public foundation models.

Anatomy of a True Data Moat

Not all data constitutes a moat. A database of public LinkedIn profiles or scraped web text is valueless because it is already included in the training sets of major models like GPT-5 or Claude. A true data moat must satisfy three rigorous criteria:

Exclusivity (The "Un-Scrapable" Test): The data must be inaccessible to web crawlers. This includes proprietary biological assays, private financial transaction logs, or internal enterprise workflows. If OpenAI can scrape it, it is not a moat.
Multimodality: The highest-value datasets in 2026 are those that combine disparate data types. For example, Tempus AI pairs genomic sequencing data with clinical outcome data. Neither dataset is unique on its own, but the linkage between them is extremely rare and valuable.
The Feedback Loop (The Data Flywheel): The product must be designed such that every user interaction generates new training data that improves the model. This creates a "Data Flywheel" where more users lead to a better model, which attracts more users.

The "1-10-100 Rule" of Data Quality

The "Rule of Data" is not just about quantity; it is obsessively focused on quality. In an AI-driven system, bad data does not just cause a reporting error; it causes hallucinations and model drift, which can destroy the product's utility.
The Metric: The 1-10-100 Rule has become a standard diligence framework for AI investors. It posits that verifying a record at the point of entry costs $1. Cleaning it after it has been stored costs $10. But if bad data feeds into an automated AI decision making process (like a drug target prediction or a loan approval), the failure cost is $100.
Operational Consequence: Companies are now valued on their "Data Hygiene." Investors look for automated data governance platforms (like Alkymi or Moody’s tools) that ensure data is "audit-ready" and semantically consistent.A company with 10 petabytes of "messy" data is a liability; a company with 1 petabyte of structured, labeled data is an asset.

Semantic Consistency and Governance

For a data moat to be actionable, it must have Semantic Consistency. In large enterprises, "revenue" might be defined differently by Sales, Finance, and Marketing. AI models cannot reason across these contradictions.Therefore, the "Rule of Data" demands a unified semantic layer where data definitions are standardised. Companies that have achieved this "semantic coherence" can deploy agents that actually work, creating a defensible barrier against competitors who are still struggling with data silos.

The Architecture of Defensibility: AI-Native vs. AI-Enabled

The "Rule of Data" has catalysed a bifurcation in the software market, dividing companies into two distinct architectural classes: AI-Native and AI-Enabled. This distinction is not merely semantic; it drives a massive divergence in valuation multiples.

Comparative Analysis of Architectures

The following table synthesises the structural differences between these two business models:

Feature	AI-Native (The New Premium)	AI-Enabled (The Legacy SaaS)
Core Strategy	AI is the foundation; the business cannot exist without it.	AI is a feature added to existing workflows (e.g., a "Copilot").
Data Flow	Data collection is intrinsic to product design; every click trains the model.	Data is siloed; limited feedback loops; often relies on third-party APIs.
Valuation Multiple	20x - 50x Revenue	5x - 10x Revenue
Gross Margins	90% (Automation replaces human labor/COGS)	70-80% (Standard SaaS Hosting)
Moat Source	Proprietary Data + Continuous Learning Loops	Brand + Distribution + Workflow Lock-in
Revenue/Employee	~$3.5M (High operational leverage)	~$200K (Standard SaaS leverage)
Scaling Dynamics	Unlimited scaling via compute.	Linear scaling via headcount (Sales/CS).
Examples	Midjourney, Recursion, Perplexity	Salesforce Einstein, Microsoft Copilot

The Economics of the AI-Native Firm

The most striking metric in the "Rule of Data" era is Revenue per Employee. Traditional SaaS companies average around $200,000–$300,000 in revenue per employee. AI-Native companies, by contrast, are achieving numbers upwards of $3.5 million per employee.

Midjourney serves as the archetype: with a team of fewer than 100 people, it generates hundreds of millions in revenue. This hyper-efficiency is possible because the "product" is generated by the AI, not by human service delivery.
Valuation Implications: Investors pay a premium for this leverage because it implies that as the company scales, costs will grow linearly (compute) while revenue grows exponentially. This breaks the traditional linear constraints of the "Rule of 40," creating a "Rule of Data" where the asset (the model/data) does the work.

Sector Case Study: The Industrialisation of Biology (TechBio)

The prompt identifies Proprietary Biological Data as the "only defensible asset." This assertion is grounded in the fact that biological data is high-dimensional, incredibly expensive to generate, and legally protected. Unlike code or text, you cannot "hallucinate" a correct biological interaction; you must observe it in the physical world.

Recursion Pharmaceuticals: The 65-Petabyte Moat

Recursion Pharmaceuticals (RXRX) is the standard-bearer for the "Rule of Data" in biotech.

The Asset: Recursion has industrialized the wet lab, using robots to conduct millions of experiments per week. This has generated a proprietary dataset exceeding 65 petabytes of biological and chemical data. This includes "phenomics", high-resolution images of cells reacting to various chemical compounds.
The Flywheel: This data feeds their foundation models, MolE (chemistry) and Molphenix (phenomics). These models predict how a new drug candidate will interact with a target in silico (virtually), allowing them to screen billions of compounds without touching a pipette.
Financial Reality: In Q3 2025, Recursion reported revenue of $5.18 million, a significant miss against the expected $16.95 million. However, the stock did not collapse to zero because investors are valuing the platform, not the quarterly drug sales. The company holds $785 million in cash, providing a runway through 2027.
Validation: The moat was validated by a $30 million milestone payment from Roche/Genentech for the delivery of a "whole genome neuromap." This transaction proves that the data itself—not just the drug—is a monetizable asset.

Tempus AI: The Clinical-Genomic Nexus

Tempus AI (TEM) illustrates the power of clinical data aggregation.

The Asset: Tempus has built a library of over 9 million patient records and 4 million genomic profiles. The moat lies in the linkage: they know the patient's genetic mutation (genomics) AND how they responded to treatment (clinical outcome). This paired dataset is estimated to be 60x larger than public datasets like The Cancer Genome Atlas.
Business Model & Valuation: Tempus operates a "data-enabled" business model. It runs diagnostic labs (low margin) to generate data, which it then licenses to pharma companies (high margin) for drug discovery. In 2025, this model drove revenue guidance to $1.26 billion (82% growth), with the company trading at ~10.5x 2025 Sales. This multiple is significantly higher than traditional diagnostic labs, reflecting the premium placed on its proprietary data moat.

Absci: Generative AI for "De Novo" Design

Absci (ABSI) represents the frontier of "Generative Biology."

The Technology: Unlike Recursion (which screens existing compounds), Absci uses generative AI to design new antibodies from scratch ("de novo").
Strategic Pivot: In late 2025, Absci executed a strategic pivot, reallocating resources to its internal pipeline (ABS-201) and away from lower-value services. Despite reporting negligible revenue of $0.4 million in Q3 2025, the company maintains a cash runway into 2028.
The Thesis: Investors are betting that Absci’s data, derived from screening billions of antibody interactions, will allow it to "solve" antibody design, reducing development times from years to months. The valuation is almost entirely derived from the optionality of its data platform rather than current cash flows.

Emerging Players: CardiaTec and Meliora

The ecosystem is expanding beyond the giants:

CardiaTec is building a multi-omics dataset specifically for cardiovascular disease, utilising human heart tissue data that is impossible to replicate without a massive clinical network.
Meliora Therapeutics is building a "mechanism of action" atlas for oncology, using machine learning to correct mischaracterised drugs. Their "molecular fingerprint" method relies on a proprietary data engine that serves as a discovery flywheel.

Beyond Biology: Data Moats in Other Sectors

While biology offers the starkest example of the "Rule of Data," the principle applies to any sector where data is scarce, private, and complex.

Financial Services: The Sovereign Data of Moody’s

In the fintech sector, Moody’s demonstrates how legacy incumbents can pivot to become AI powerhouses.

The Moat: Moody’s possesses decades of proprietary financial data, credit ratings, and risk assessments that are not available on the open web.
The Application: By training AI models on this private corpus, Moody’s created an "AI Research Assistant" that reduces financial analysis time by 30%. Because the model is grounded in proprietary, verified data, it avoids the hallucinations common in generic finance bots, creating a defensible "high-trust" moat.

Private Markets: Unstructured Data as a Moat

Companies like Alkymi and 7 Chord are capitalising on the opacity of private markets.

Alkymi uses AI to extract data from unstructured private equity documents (PDFs, emails). The moat here is the access to these private documents and the proprietary ontology built to understand them. A generic LLM cannot interpret a "capital call notice" with the precision required for financial settlement without this specialised training data.

Marketing Tech: Zeta Global’s Data Cloud

Zeta Global provides a counter-example in the marketing space.

The Asset: Zeta manages a proprietary data cloud of consumer intent signals.
Valuation: Trading at a discount to high-growth peers, Zeta is viewed by some analysts as an "undervalued gem" because its data moat allows for precise targeting that survives the death of the third-party cookie. The model projects a 49% return based on this data-driven efficient growth.

The Investor’s Dilemma: Valuation Frameworks in 2026

As the market pivots from the Rule of 40 to the Rule of Data, the toolkit for valuing companies is being rewritten. The following frameworks summarise how investors are pricing assets in 2026.

The New Due Diligence Checklist

Investors are moving beyond simple P&L analysis to audit the asset base itself.

Data Provenance & Exclusivity: Is the data proprietary, or is the company a "wrapper" around public data? Investors look for the "Data-to-Model Ratio", how much unique internal data is used to fine-tune the model vs. base training data.
Burn Multiple: As a proxy for product-market fit. A Burn Multiple > 2.0x suggests the company is "buying revenue" rather than growing through data network effects.
Revenue Quality: A shift from Gross Merchandise Value (GMV) to "Revenue Quality." AI-driven revenue (high margin, automated) is valued higher than human-driven revenue (low margin, service-heavy).

Valuation Multiples by Category

The market has stratified into distinct valuation tiers based on defensibility.

Category	Valuation Multiple (EV/Revenue)	Defensibility Profile
Traditional SaaS	3.9x - 6.0x	Low. Commoditized by AI code gen.
AI-Enabled SaaS	5.0x - 10.0x	Medium. Incremental efficiency gains.
AI-Native / TechBio	12.0x - 30.0x	High. Proprietary data creates a "hard" moat.
Sovereign Data Platforms	Premium (Outliers)	Extreme. National security / Health critical assets.

VC Sentiment and Deal Structure

The "Flight to Quality" is real. In Q3 2025, venture funding concentrated heavily in late-stage, proven winners. Eleven "mega-deals" accounted for a significant portion of capital, with AI funding representing 46% of all VC dollars. Investors are willing to pay massive premiums for companies that have proven their "Data Flywheel," while seed-stage funding for "wrapper" startups has dried up.

The Counter-Thesis: Distribution and Reasoning

No strategic shift is without its detractors. A significant counter-narrative exists that challenges the supremacy of the "Data Moat," arguing instead for Distribution and Reasoning.

The "Distribution is King" Argument

Skeptics argue that in a world where intelligence is a commodity, the entity that owns the customer wins.

Workflow Lock-in: If a company like Salesforce (CRM) or Microsoft (Office) integrates a "good enough" open-source model into their existing workflow, they can crush a superior AI-native competitor. The "switching cost" of moving data out of Salesforce is higher than the benefit of a slightly better AI model.
The "Verb" Moat: Being the default option (e.g., "Google it", "Slack me") creates a brand moat that technical superiority cannot breach. Investors betting on this thesis prioritise companies with massive install bases over those with unique data.

The "Reasoning" Overhang

A more technical counter-argument comes from AI researchers like Bob McGrew (formerly of OpenAI), who argue that Reasoning (System 2 thinking) is the next frontier, not just data scale.

Less Data Needed: As models get better at "reasoning," they may need less training data to perform tasks. "Zero-shot" learning could theoretically allow a generic model to perform a specialised task (like reading a contract) without needing a proprietary dataset of 10,000 contracts.
Commoditization of Agents: If reasoning becomes solved, then AI agents themselves will become commodities priced at the cost of compute. In this scenario, the only value capture is in the physical world (Robotics/Bio) or in "Sovereign" data that requires trust (Healthcare/Finance).

Synthesis: The "Data-Distribution Flywheel"

The most sophisticated investors in 2026 are looking for the intersection of both. The ultimate winner is the company that uses Distribution to acquire customers, which generates Data, which improves the product, which drives more Distribution.

xAI (Grok): Elon Musk’s xAI is cited as a prime example. It has "Distribution" (access to X/Twitter’s 600M users) AND "Data" (the real-time feed of global conversation). This combination creates a moat that neither a pure model company (like Anthropic) nor a pure social network can easily replicate.

Conclusion: The New Rules of the Road

The transition from 2025 to 2026 marks the end of the "Growth at All Costs" era and the beginning of the "Defensibility at All Costs" era. The "Rule of 40," while still a useful hygiene metric for financial solvency, has been superseded by the Rule of Data as the primary determinant of enterprise value.

Strategic Imperatives for 2026:

For Founders: Stop pitching the algorithm. Pitch the data. Demonstrate how your product captures unique, high-dimensional data that improves your model in a closed loop. If your data is available on the public web, you do not have a company; you have a feature.
For Investors: Audit the data pipeline. Apply the "1-10-100 Rule" to assess technical debt. Scrutinise the "Burn Multiple" to ensure that growth is organic and product-led, not fuelled by unsustainable ad spend.
For Enterprises: Embrace open source. The cost of intelligence is trending to zero. Use LLaMA and Mistral for 80% of your workflows to preserve margins, and reserve proprietary spend for the 20% of tasks where reasoning or security is paramount.

In the final analysis, the "Rule of Data" is a return to first principles. In a digital world where copying code is free, value accrues to the scarcity of truth. Whether that truth is the binding affinity of a protein, the creditworthiness of a borrower, or the intent of a consumer, the companies that own the Source of Truth will inherit the future.

Summary of Key Metrics for 2026

Metric	Definition	Benchmark
Rule of Data	Valuation based on proprietary data volume, exclusivity, and feedback loops.	Data Flywheel active.
Burn Multiple	Net Burn / Net New ARR	< 1.5x (Excellent)
1-10-100 Rule	Cost of Data Quality (Prevention vs Correction vs Failure).	Automated Governance in place.
Revenue/Employee	Total Revenue / Full Time Employees	>$1M (AI-Native Target)
Rule of 40	Revenue Growth + EBITDA Margin	> 40% (Baseline Hygiene only)

The "Rule of 40" is dead. Long live the Rule of Data...

Nelson Advisors > MedTech and HealthTech M&A

Nelson Advisors specialise in mergers, acquisitions and partnerships for Digital Health, HealthTech, Health IT, Consumer HealthTech, Healthcare Cybersecurity, Healthcare AI companies based in the UK, Europe and North America. www.nelsonadvisors.co.uk

Nelson Advisors regularly publish Healthcare Technology thought leadership articles covering market insights, trends, analysis & predictions @ https://www.healthcare.digital

We share our views on the latest Healthcare Technology mergers, acquisitions and partnerships with insights, analysis and predictions in our LinkedIn Newsletter every week, subscribe today! https://lnkd.in/e5hTp_xb

Founders for Founders > We pride ourselves on our DNA as ‘HealthTech entrepreneurs advising HealthTech entrepreneurs.’ Nelson Advisors partner with entrepreneurs, boards and investors to maximise shareholder value and investment returns. www.nelsonadvisors.co.uk

#NelsonAdvisors #HealthTech #DigitalHealth #HealthIT #Cybersecurity #HealthcareAI #ConsumerHealthTech #Mergers #Acquisitions #Partnerships #Growth #Strategy #NHS #UK #Europe #USA #VentureCapital #PrivateEquity #Founders #BuySide #SellSide #Divestitures #Corporate #Portfolio #Optimisation #SeriesA #SeriesB #Founders #SellSide #TechAssets #Fundraising #BuildBuyPartner #GoToMarket #PharmaTech #BioTech #Genomics #MedTech