By Global Business & Technology Desk
Published in partnership with Crunchbase Insights and Products That Count
Executive Overview
For nearly a decade, the pitch deck formula for artificial intelligence startups has remained remarkably consistent: open with a grand vision, highlight a proprietary dataset, and drop the magic phrase—"We use AI."
According to SC Moatti, founding managing partner of Mighty Capital and board chair at Products That Count, that introductory phrase is no longer a differentiator. In fact, it is an administrative error.
Data compiled from a recent analysis of 576 venture-backed, AI-driven B2B companies that have raised rounds exceeding $50 million since early 2025 reveals a staggering reality: 97% of products nominated for this year’s Products That Count Product Awards are already deeply integrated with artificial intelligence. The era of AI as a unique selling proposition (USP) is officially over. AI has graduated from a business model to basic infrastructure—no more remarkable than electricity or cloud hosting.
+-----------------------------------------------------------------+
| THE EVOLUTION OF THE AI PITCH DECK |
| |
| [ 2018 - 2023 ] ---> "We use AI" = Revolutionary USP |
| [ 2025 & Beyond ] ---> "We use AI" = Baseline Infrastructure |
+-----------------------------------------------------------------+When building code, deploying models, and generating synthetic data become virtually free, what separates a fleeting application from a lasting enterprise? To answer this question, Moatti and her research team layered proprietary Crunchbase data over Hamilton Helmer’s legendary 7 Powers framework, synthesizing insights from a community of over 600,000 global product leaders.
The empirical findings offer a stark warning to founders and venture capitalists alike. When foundational models can be spun up overnight by well-funded competitors, traditional concepts of defensive value—such as proprietary data or sheer code volume—crumble. Out of the classic strategic moats analyzed, only two consistently command premium market multiples without requiring a startup to out-spend hyperscalers like OpenAI or Anthropic. Two others act as dangerous traps, and one remains structurally out of reach for 99% of early-stage builders.
Detailed Chronology: The Demise of the "Model-First" Era
To understand why traditional venture moats are failing in the current market cycle, one must examine how the venture ecosystem arrived at this saturation point.
Phase 1: The Foundation Model Gold Rush (2022–2024)
Following the public release of generative transformer models, capital flooded into the market. Startups built "wrapper" applications—thin user interfaces layered on top of third-party foundational models. During this window, simply demonstrating that an API could automate a manual workflow was enough to command sky-high valuations. Investors prized velocity over structural defensibility, assuming that early market entry and initial feature sets would create self-sustaining momentum.
Phase 2: Democratization and Cost Compression (Late 2024–Early 2025)
As open-source models (such as Meta’s Llama series and various Mistral releases) matured, the cost of deploying state-of-the-art intelligence plummeted. Simultaneously, inference costs dropped by orders of magnitude. Startups found that capabilities which previously required millions of dollars in custom training runs could now be replicated in a weekend by any developer with an internet connection and a cloud budget.
Phase 3: The Reckoning (2025 and Beyond)
This brings us to the current landscape. An analysis of 576 venture-backed B2B AI firms reveals a bifurcated market. Companies relying purely on model superiority or static data assets are seeing their pricing power eroded by aggressive, well-capitalized new entrants. Conversely, companies that engineered structural defensibility into their business models—independent of the underlying AI—are commanding median enterprise value-to-raised capital multiples north of 4x to 5x.
Supporting Context & Metrics: The 7 Powers Framework Analyzed
The Crunchbase-Mighty Capital dataset evaluated companies across Helmer’s 7 Powers framework to determine which structural advantages actually yield enterprise value in an AI-saturated market.
+---------------------+-------------------+-----------------------------------+
| Strategic Power | Dataset Prevalence| Median Enterprise Value Multiple |
+---------------------+-------------------+-----------------------------------+
| Counter-Positioning | 5% | 5.3x |
| Network Economies | 5% | 4.2x |
| Switching Costs | 37% | 4.0x (High Capital Intensity) |
| Cornered Resources | 44% | 2.6x (Failing Moat) |
| Scale Economies | Minimal* | 3.2x* (*Skewed by OpenAI/Anthropic|
+---------------------+-------------------+-----------------------------------+1. Counter-Positioning: The Ultimate Power (5% Prevalence | 5.3x Multiple)
Counter-positioning occurs when a newcomer adopts a business model so fundamentally misaligned with an incumbent’s legacy revenue streams that the incumbent cannot copy it without destroying their own core economics.
- The Historical Paradigm: Netflix versus Blockbuster. Blockbuster could technically launch a DVD-by-mail or streaming subscription service, but doing so would instantly cannibalize the high-margin late fees that kept their physical franchise stores solvent. By the time they tried, it was too late.
- The AI Era Manifestation: Only 5% of companies in the dataset leverage counter-positioning, yet investors price this scarcity at a median enterprise value of 5.3x per dollar raised—the highest multiple in the study.
- Real-World Application: This power appears clearly in vertically integrated AI insurance platforms selling directly to employers, a direct-to-consumer motion that traditional insurance brokers cannot replicate without alienating their existing distribution channels. It also emerges in AI-native revenue management systems that would gut the lucrative, billable-hour consulting revenues legacy software vendors rely upon if they attempted to match them.
Diagnostic Question for Founders: Could a well-resourced incumbent copy your business model if they wanted to? The correct answer is: "Technically yes, but executing it would cost them more in damaged legacy revenue than it would cost us to build."
2. Network Economies: The Self-Reinforcing Moat (5% Prevalence | 4.2x Multiple)
Network economies arise when a product’s utility increases for every user as more participants join the ecosystem.
- The Historical Paradigm: LinkedIn. Recruiters attract job seekers; job seekers attract professionals; professionals attract more recruiters.
- The AI Era Manifestation: Present in just 5% of the analyzed companies, network economies command a robust 4.2x valuation multiple, making them the most capital-efficient path to premium valuations.
- The B2B Variant: In modern B2B software, the network node is rarely an individual consumer; rather, it is an enterprise entity—connecting brands to factories, programmatic advertisers to niche audiences, or SaaS platforms to third-party integration partners. Every new participant enriches the proprietary feedback loops of cost structures, production cycles, and operational behaviors. Solving the "cold-start problem" in a two-sided B2B market is exceptionally difficult, but once achieved, it yields a fortress that a competitor with a superior base model cannot simply buy.
The Illusions: What Looks Like a Moat But Isn’t
Founders frequently misidentify operational hurdles or static assets as strategic moats. The data exposes two primary traps that consume excessive capital for substandard returns.
Cornered Resources (44% Prevalence | 2.6x Multiple)
Proprietary datasets, exclusive IP, and specialized talent pools make up nearly half (44%) of the dataset. Yet, they suffer from the lowest valuation multiple at 2.6x.
Why? Because foundation models and synthetic data generation have commoditized static information. If a startup’s data advantage does not compound dynamically through proprietary user workflows in a way that continuously outpaces public or synthetic data sets, it is not a moat. It is merely a depreciating asset.
Switching Costs (37% Prevalence | 4.0x Multiple)
Switching costs appear to be a formidable defense because customers rarely churn. Products featuring deep enterprise entanglement—embedded instrumentation, institutional memory, and complex rearchitecture risks—boast a respectable 4.0x multiple.
However, the capital required to build these switching costs is astronomical. Achieving deep integration necessitates notoriously expensive enterprise sales cycles, protracted onboarding, and heavy professional services. Unless a startup utilizes a product-led growth (PLG) motion to lower acquisition friction, or engineers user-to-user collaboration that gradually transforms switching costs into true network economies, the cash burn required to maintain this posture will bankrupt most early-stage balance sheets.
A Note on Scale Economies: Excluding market aberrations like OpenAI and Anthropic, the median multiple for scale economies collapses from 6.1x to 3.2x. Believing your unit economics will naturally improve with volume is not the same as possessing a structural scale moat. That gap is measured in billions of dollars of venture capital that 99% of startups will never secure.
Official Perspectives: Expert Commentary
Reflecting on the study’s implications, tech industry leaders and venture capitalists emphasize a return to classical economic fundamentals disguised in modern technological garb.
+-----------------------------------------------------------------+
| SC MOATTI ON DEFENSIVE STRATEGY |
| |
| "If you can't answer what survives a competitor starting |
| today with more capital and a better model, you're |
| building a product. The founders commanding 4x to 5x |
| multiples are building a power." |
+-----------------------------------------------------------------+"The market has matured past the point where a wrapper or a clever prompt can sustain a venture-scale valuation," notes the research synthesis from Mighty Capital and Products That Count. "When the underlying intelligence layer becomes a commodity utility, value migrates away from the model and toward business architecture. Founders must stop obsessing over fractional improvements in model accuracy and start engineering structural friction that repels well-funded copycats."
Industry analysts point out that venture capital deployment is increasingly shifting toward teams that understand economic game theory. Capital efficiency is no longer about spending less on marketing; it is about designing a system where the rules of the market make it irrational for incumbents to compete directly.
Future Outlook: The Blueprint for AI-Era Founders
As the venture ecosystem pushes deeper into the back half of the decade, the criteria for enduring success are crystallizing. The era of easy capital driven by hype cycles has yielded to an environment that rewards rigorous economic engineering.
Strategic Recommendations for Founders:
- Abandon the "AI First" Pitch: Pivot the narrative away from infrastructure capabilities. Investors and enterprise buyers assume your product uses AI; they want to know how your business model captures value that the model provider cannot extract from you.
- Audit Your Defensibility: Review your product through Hamilton Helmer’s 7 Powers. If your primary defense rests on "proprietary data" or "expert prompting," assume a competitor will replicate it within 90 days.
- Design for Counter-Positioning: Examine your target industry’s incumbents. Where are they structurally handcuffed by legacy revenue streams? Build a go-to-market motion that exploits those blind spots.
- Engineer Network Loops: If counter-positioning is unattainable, focus relentlessly on multi-tenant network dynamics. Ensure that every transaction, data point, or user interaction compounds the value of the platform for all other participants.
Ultimately, every sustainable company in the modern B2B landscape incorporates artificial intelligence. But the companies commanding premium 4x to 5x valuation multiples have built something profound underneath that AI layer—something a model cannot generate, and something a better-funded competitor cannot easily buy.
The question every founder must answer before stepping into a board meeting or a pitch room remains singular, unforgiving, and absolute:
What about your business would survive a competitor who starts tomorrow with more capital and a better model?
If you cannot answer that in a single sentence, you are not building a market power. You are building a temporary feature.
