Navigating AI Market Exuberance: A Disciplined Investment Framework That Builds Value
If you only have a minute, here’s what you need to know.
The rapid rise of artificial intelligence has reignited the debate around whether the market is experiencing a speculative bubble. While signs of exuberance are clearly present — including elevated valuations, unprecedented revenue growth, and massive infrastructure investment — we do not believe the current environment represents a systemic AI bubble comparable to the dot-com era. Unlike previous speculative cycles, today’s AI expansion is supported by real-world adoption, tangible revenue generation, and infrastructure investments largely financed through corporate cash flows rather than leverage.
AI adoption is occurring at an unprecedented speed and scale. Consumer platforms have reached hundreds of millions of users in record time, while enterprise adoption has accelerated across nearly all industries. At the same time, AI-native companies are redefining software economics: reaching meaningful revenue milestones in a fraction of the time required by traditional SaaS businesses and doing so with dramatically leaner teams. These dynamics point to a structural shift in how value is created, rather than demand driven purely by hype.
That said, risks remain: enterprise deployments often stall at pilot stage, foundation model leaders face significant cash burn with unclear near-term profitability, and the ecosystem shows increasing financial interdependence among hyperscalers, model developers, and infrastructure suppliers — all of which can amplify downside in a correction, especially in late-stage and foundation model segments.
Against this backdrop, Mindset Ventures’ strategy is intentionally designed to perform across market cycles. We focus on Seed and Series A AI-native companies with clear technical differentiation, capital-efficient business models, and defensible moats — particularly in vertical applications and software infrastructure layer that enables enterprise-scale AI deployment, such as security, observability, and orchestration. These companies benefit from AI adoption regardless of which foundation models dominate, while remaining insulated from the financial interdependencies of large incumbents.
Our approach is grounded in over a decade of AI investing experience, more than 25 AI-native investments, and a rigorous diligence process that combines deep business analysis with hands-on technical evaluation by seasoned AI practitioners. With a longer investment period in our new AI-native fund, we maintain valuation discipline, avoid forced deployment during overheated markets, and retain flexibility to invest aggressively when conditions normalize.
While we remain vigilant to evolving risks, our conviction is clear: AI represents a durable technological platform shift, not a transient bubble. Our investment framework is built to protect capital in downside scenarios while capturing the asymmetric upside of long-term AI value creation.
Navigating AI Market Exuberance: A Disciplined Investment Framework That Builds Value
Pedro Mesquita — Partner at Mindset Ventures
Recent months have seen intense debate over whether we’re in an AI bubble. While there are a lot of misconceptions and misinterpreted data going around, compelling arguments exist on both sides. Like everyone in the market, we cannot predict the future, but rigorously evaluating various perspectives helps us better gauge market dynamics and risk.
Regardless of the conclusion, our decade of AI investing experience has shaped a disciplined strategy: one designed to weather potential downturns while capitalizing on AI’s long-term value creation. With our new AI-native fund on the horizon, this approach becomes even more critical.
The arguments against a bubble
1)AI adoption is unprecedented in speed and scale
To evaluate any technology’s real value, we must examine adoption rates. By this measure, AI is rewriting the rules.
On the consumer side, ChatGPT reached 800 million weekly active users in just over two years, a scale that took the internet over a decade to achieve. This isn’t just hype; it reflects genuine utility and how existing infrastructure (smartphones, broadband, cloud computing among others) can dramatically accelerate the diffusion of breakthrough technologies.
Enterprise adoption tells an equally compelling story. As of 2025, 88% of organizations now use AI in at least one business function, up from 50% in the pre-ChatGPT era.
Adoption metrics matter, but the real test is revenue generation. Can companies actually capture meaningful economic value from this technology wave?
For leading AI-native companies, the answer is an unequivocal yes! Traditional SaaS companies took an average of 7 years to reach $100M in ARR. The best AI-native companies are reaching this milestone in just 1.5 years. Anysphere, the maker of AI code editor Cursor, achieved over $1B in ARR in November 2025, up from $1M in ARR just 2 years earlier. AI-native companies are rewriting decades of software economics.
Efficiency gains are equally striking. Historically, the best SaaS companies needed 300–500 employees to reach $100M in ARR (think Slack, LinkedIn, Calendly). AI-native companies are achieving the same milestone with fewer than 100 employees and often with teams of just 30–50 people. While most of these companies still burn cash due to computing and infrastructure costs, this operational efficiency signals fundamentally more scalable business models.
2) Valuations are elevated but remain well below dot-com bubble extremes
Valuation is also an important metric to measure market exuberance.
The valuation of the top public tech companies is currently high but significantly lower than the respective group of companies prior to the dot-com bubble. In 1999, leading tech companies traded at forward P/E ratios of ~67x earnings, and many had no earnings at all. Today, the top tech companies trade at an average of 28x forward earnings, elevated but backed by actual profitability and strong cash flow generation.
In the private market, we can see a significant increase in median pre-money valuations, especially at later-stage rounds (Series B+). This has been mainly driven by megadeals ($5B+ rounds) going to a small set of high-profile companies that have contributed to a high degree of concentration in the VC market with 0.05% of deals in 2025 attracting half of all VC dollars throughout the same period. On the other hand, early-stage deals remain closer to historic medians.
3) CAPEX investments are high but mostly financed by the companies’ own cash
Hyperscalers (massive cloud data providers like Google, Microsoft, Amazon, and Meta) are all expected to increase their CAPEX investments to support anticipated AI demand. These investments include AI infrastructure such as GPUs, servers, and data centers, which together represent hundreds of billions of dollars in annual investment. However, unlike the dot-com era, this capital is coming from companies’ own balance sheets rather than debt, which drastically decreases the financial risk of such transactions for the corporations backing them.
Furthermore, looking at AI demand in hyperscalers’ earnings releases, the current CAPEX investment appears justified. By analyzing current and announced future investments in AI infrastructure, we can estimate how much AI demand these projects aim to absorb. Comparing this future AI capacity with current demand growth rates allows us to evaluate whether these big tech companies are being overoptimistic. Current data suggests they are not.
Arguments supporting a bubble
1) Adoption is strong, but we have yet to see transformational value
Enterprise AI adoption is rising sharply, but the vast majority of these initiatives remain prototypes, experiments, or limited-function deployments. According to McKinsey, only 7% of enterprises that report using AI have deployed it at full scale across their operations. Some sources also point to a stagnation in enterprise AI adoption across all industries. This apparent loss of momentum and gap between experimentation and production deployment is what bubble skeptics point to.
However, we view this stagnation differently: as evidence of untapped opportunity rather than failed promise. Enterprises aren’t avoiding AI adoption because they don’t see value. They’re constrained by a lack of software infrastructure that allows them to deploy AI to its fullest potential while maintaining compliance, security, and operational standards. The gap between pilot and production isn’t a demand problem; it’s a tooling problem.
This is where our AI-native fund thesis focuses. We’re targeting companies building the critical infrastructure layer that will enable enterprise-scale AI deployment: cybersecurity solutions for AI systems, observability platforms for model monitoring, and orchestration tools for multi-model workflows. These are the picks-and-shovels companies that will enable the next wave of adoption — and they’re largely underinvested relative to foundational models-layer companies.
2) Revenue growth is unprecedented, but profitability remains elusive
As mentioned earlier, revenue growth is unlike anything we’ve seen in history, but most AI leaders are still losing substantial amounts of money. OpenAI, for example, reportedly projects a $115 billion cumulative cash burn by 2029 while only achieving breakeven that same year. Anthropic, despite strong revenue traction, is similarly unprofitable with billions in annual losses.
This matters because revenue growth without a path to profitability was what characterized dot-com bubble companies. The fundamental question is whether today’s losses represent necessary investment in a winner-take-all market (similar to Amazon’s early years) or unsustainable unit economics (like Pets.com or Webvan). Foundation models face particularly challenging economics: training costs in the hundreds of millions, inference costs that scale with usage, and intense price competition that has already driven API pricing down over 90% since 2023.
However, this unprecedented cash burn is largely exclusive to a select few foundation model companies. The application and software infrastructure layer tell a different story — one we’ve witnessed firsthand through Fund IV. We’ve identified and invested in AI-native companies achieving exceptional growth while maintaining disciplined cash burn. These companies leverage existing foundation models rather than building them from scratch, resulting in capital-efficient business models with clear paths to profitability.
That said, not all application-layer companies are created equal. The AI application space is crowded with “wrapper” companies, startups that simply add a user interface on top of existing models without any real technical differentiation or defensible moat. These companies are vulnerable to commoditization as foundation model providers themselves move up the stack or as competitors easily replicate their functionality.
Our investment criteria are designed to filter out these wrappers and identify companies with genuine defensibility. We focus on vertical applications where founders bring deep industry expertise, established existing/potential customer relationships, and access to proprietary field data. Our two technical partners, both with decades of AI experience, evaluate whether companies have real technical sophistication: custom fine-tuning approaches, proprietary data pipelines, domain-specific model architectures, or unique integration capabilities that create switching costs. These characteristics not only provide defensibility but also make these companies attractive acquisition targets for foundation model developers looking to expand into specific industries.
3) There is a circular economy in play among the big AI players
The big AI players are all interconnected through multi-billion-dollar deals that create concerning circular dependencies. NVIDIA agreed to invest up to $100 billion in OpenAI, who will in turn use much of this capital to purchase NVIDIA’s own GPUs to power its data centers. Microsoft, Oracle, Meta, Google, Amazon, and other hyperscalers are part of similar loops — simultaneously investors, suppliers, and customers to each other.
While there’s no denying these companies’ astronomical revenues, this circular economy creates systemic risk. If one major player fails to meet market expectations or experiences a significant setback, it could trigger a downward chain reaction: reduced purchases from suppliers, writedowns on investments, and questioned valuations across the ecosystem. When everyone in this interconnected group is simultaneously betting on everyone else’s success, the margin for error shrinks considerably.
Our early-stage focus insulates us from these circular dependencies. The Seed and
Series A companies we invest in are building on top of commoditized foundation models and cloud infrastructure. They’re customers of this ecosystem, not participants in its interdependent capital flows. A correction in hyperscaler valuations or a slowdown in foundation model funding would not directly impact our portfolio companies’ ability to operate or grow. In fact, such a correction could benefit early-stage companies by reducing competition for talent and making cloud infrastructure more cost-effective.
Where do we stand in the bubble-no-bubble spectrum?
To go straight to the point: we don’t think we are currently in an overall AI bubble. Like any major technological shift, there has been a strong degree of exuberance driven by the enormous opportunity ahead. This inevitably leads to some assets being overpriced, but it doesn’t add up to the systemic overvaluation we saw in the dot-com era. The fundamentals, such as real adoption, actual revenue, and cash-financed infrastructure investment, are materially different from 1999.
That said, we recognize the uncertainty. As we enter 2026, we’re closely tracking the data points outlined above. Our position can change as evidence evolves, and we’re prepared to adjust accordingly.
But here’s what matters most: our strategy is designed to win regardless of who’s right.
We at Mindset Ventures have structured our investment approach to avoid major losses in a potential downturn while still capturing AI’s long-term value creation. After a decade of AI investing and over 25 AI-native investments, including navigating the 2022 market correction, we’ve developed a framework that’s proven resilient. Here’s how it works:
1) An extended investment period creates valuation flexibility
Our new AI-native fund has a longer investment period than any of our previous funds. This serves two critical purposes: first, it gives us more time to source the best deals in an increasingly competitive market. Second, and more importantly, it prevents overexposure to any single valuation environment. We can be patient during frothy periods and deploy more capital when valuations become rational. This temporal diversification is one of our most important risk management tools.
2) A decade-refined, hype-proof diligence process
Over ten years of AI investing, we’ve built a diligence process specifically designed to separate substance from hype. After identifying startups that fit our thesis, they go through clear stages with defined criteria. We know exactly which documents and data to request for deep business understanding.
More importantly than ever, we also have two technical partners who have been active in AI for decades and can evaluate whether a company’s technology is genuinely sophisticated or just marketing. They assess model architecture, training approaches, data quality, and defensibility of technical moats. This technical diligence has helped us avoid numerous “AI-washing” companies that would have looked compelling on surface metrics alone.
After going through our full process, we invest only in startups with real substance: demonstrated product-market fit, technical differentiation, and founders who deeply understand both the technology and the business.
3) Early-stage focus provides built-in downside protection
We remain committed to Seed and Series A deals, which offer natural insulation from market volatility. Public markets feel downturns first, with the impact cascading to late-stage private companies, then growth stage, and finally early-stage. Our positioning gives us time to read market signals and adapt our strategy before valuations at our stage are materially affected.
Early-stage deals also offer asymmetric upside. A successful Seed investment can return 30–50x, providing a substantial buffer against losses elsewhere in the portfolio. This risk-reward profile is fundamentally different from late-stage investing, where even successful outcomes may only return 3–5x. In a correction scenario, our portfolio construction protects us; in a continued boom, we capture disproportionate gains.
The Mindset advantage: a decade of AI investing at an inflection point
All in all, we are extremely excited about where we stand as a firm. We’re at the beginning of a major technological platform shift that will create more value than any previous wave. And we’re more prepared than ever to capitalize on it.
Our track record of over 25 AI-native investments means we entered this current wave with established pattern recognition, proven relationships, and battle-tested diligence frameworks. We’ve seen AI hype cycles before: the deep learning renaissance of 2015, the machine vision boom of 2018, the MLOps wave of 2021. We know what sustainable AI businesses look like versus what fades when the narrative shifts.
Our track record speaks to this: 40% higher MOIC on AI-native investments compared to our non-AI portfolio, with successful bets on companies that have become category leaders. We’re not newcomers trying to catch a wave; we’re specialists who’ve been building expertise while others were still skeptical of AI’s commercial viability.
The stars have aligned: transformative technology meeting mature investing discipline. It’s time for us to execute!
