How will AI change venture capital fund returns?

The advent of artificial intelligence is creating a period of radical variance in venture capital, with the potential for both unprecedented returns and catastrophic losses, fundamentally reshaping investment theses and financial models . Experts describe AI as the first major business model dislocation in over two decades, creating an opportunity for several new trillion-dollar companies to emerge by targeting massive markets like the $10 trillion U.S. labor sector through the replacement of knowledge workers [5, 9, 19, 23]. This potential for explosive, non-linear growth is rendering traditional SaaS evaluation metrics like "T2D3" obsolete [1, 4]. Consequently, some investors predict the performance gap between top-performing venture funds and other asset classes like private equity will widen significantly over the next decade . However, this opportunity is coupled with extreme risk, characterized as a speculative "industrial bubble" akin to the dot-com era . The consensus among many is that a massive extinction event is imminent for the AI application layer, with improving foundation models projected to "swallow" or nullify the functionality of over 80% of current startups in the space [5, 19, 27]. This has led to predictions of impossibly large losses for VCs backing thinly-wrapped AI applications without durable moats .

The unit economics of AI-native companies present a significant challenge to traditional venture evaluation, forcing a re-evaluation of financial metrics. These companies are structurally lower gross margin businesses due to high inference and compute costs, with some funding rounds seeing as much as 70% of capital spent on compute [1, 14]. This has created a VC-subsidized market where a large portion of investment flows directly through startups to foundational model providers and hardware companies, raising questions about long-term viability [20, 24]. In response, some investors are shifting their focus from gross margin percentages to absolute gross profit dollars per customer, arguing that larger contract sizes can justify the different cost structure [6, 29]. The bull case suggests that while gross margins are lower, AI companies may achieve higher terminal operating margins by leveraging AI to create significant efficiencies in sales, engineering, and G&A . This economic uncertainty is compounded by practices such as "circular deals," where large tech companies invest in startups that then use the capital to purchase their cloud services, potentially inflating valuations and obscuring true financial health .

In response to these dynamics, venture capital investment theses are adapting rapidly. Capital is shifting away from traditional SaaS and toward foundational models and defensible AI-native applications . The new diligence framework centers on whether a company's value increases or decreases as underlying foundation models improve . The investment focus is narrowing to companies that can pass a "$100B public company test," indicating massive total addressable markets and multi-product potential . A key point of debate is whether incumbents or startups will be the primary beneficiaries. One view posits that established companies will leverage AI to enhance productivity and capitalize on existing distribution, reinforcing their market power . The opposing view argues that AI will enable entirely new categories of companies that were previously impossible, leading to widespread disruption [26, 30]. This uncertainty affects fund strategy, with some asserting that mega-funds cannot produce top-tier returns in this environment [6, 10], while others believe their access to pivotal companies like OpenAI and Anthropic provides a decisive advantage . Given that median historical VC returns are relatively modest at around 8% IRR [13, 21], the high-variance nature of the AI transition suggests that manager selection will be more critical than ever for limited partners seeking to capture the asset class's potential upside.