Over the past year, headlines have been dominated by venture financings at valuations that would have seemed unimaginable only a few years ago. Artificial intelligence (“AI”) companies are raising billions of dollars at valuations measured in the tens, and in some cases hundreds, of billions of dollars.
Historically, periods of elevated venture valuations have often been followed by weaker investment returns as expectations eventually converge with fundamentals. We saw this during the dot-com bubble, in the years leading up to the Global Financial Crisis, and likely during the venture boom of 2021 and 2022. This naturally raises the question: are today’s venture valuations, particularly in AI, justified?
The answer is more nuanced than history’s negative implications alone might suggest. There are credible reasons to believe the economics of company building are changing. AI has the potential to enable businesses to scale faster, operate more efficiently, and reach meaningful scale with fewer resources than in prior technology cycles. At the same time, there is considerable uncertainty to these utopian forecasts as we remain in the infancy of this new AI paradigm. While these changes may ultimately justify higher valuations for a select group of companies, history reminds us that periods of rapid innovation are often accompanied by excessive optimism.
Importantly for our clients, we believe our venture strategy is well positioned across a range of outcomes. If today’s premium valuations are ultimately supported by exceptional future growth, our portfolio provides exposure to that upside. If markets are ahead of themselves and valuations retrench, our emphasis on finding truly exceptional managers and investing at the earliest stages of company formation positions us well to navigate a more challenging environment.
Why Current Valuations May Be Justified
The increase in valuations has been remarkable. Nowhere has this been more apparent than in early-stage venture, where one could argue valuations have increased by fivefold since 2017.
The easy conclusion from this chart would be to head to the beach and declare venture too expensive to generate attractive future returns. The reality is more nuanced. The economics of building AI companies appear fundamentally different than those of prior technology cycles.
For starters, companies are reaching meaningful revenue levels much earlier and growing at unprecedented rates. Anthropic provides the best example of this phenomenon, with some reports suggesting gross annualized recurring revenue has approached $65 billion by the end of July, up from single-digit billions at the start of the year. That pace of growth would have seemed unimaginable only a few years ago. With companies compounding at unprecedented rates, does it follow that investors should be willing to underwrite valuations that also sit well above historical norms?
One important counterbalance is the extraordinary cost of building these companies. AI models require enormous amounts of compute, both in the form of specialized chips and the electricity required to power them. Unlike many previous software businesses, today’s leading AI companies often require billions of dollars of capital simply to meet customer demand. A true war chest is required to compete. This dynamic is raising the barriers to entry and concentrating the race among a relatively small number of well-capitalized firms.
AI has also intensified the competition for talent. The industry’s best engineers increasingly gravitate toward companies they believe have the highest probability of success. Large financings and high valuations become powerful recruiting signals, making it easier for category leaders to attract exceptional employees. In AI, where a relatively small number of individuals can create enormous value, those recruiting advantages compound over time.
The venture ecosystem itself has also changed dramatically. Capital has become increasingly concentrated among a small number of firms. In 2025, the 10 largest venture funds raised $22 billion, representing 32.9% of all U.S. venture fundraising, up from just 13% in 2021 (NVCA). By the first quarter of 2026, just six venture firms accounted for more than three-quarters of all venture commitments (Pitchbook).
With capital concentrating into a handful of firms, it is not surprising that financing rounds have become substantially larger. Today’s largest venture firms are sitting at the poker table with dramatically larger chip stacks than the rest of the market, and they are playing a fundamentally different game. Rather than simply funding companies, they increasingly use capital itself as a competitive weapon. Massive financings create capital moats, allowing portfolio companies to recruit the best talent, absorb enormous compute costs, and outlast competitors with fewer financial resources.
As fund sizes have grown, the economics of venture investing have changed alongside them. A $1 billion exit, once considered extraordinary, has a limited impact on a multi-billion-dollar fund. Increasingly, the largest firms require portfolio companies capable of producing outcomes measured in the tens of billions of dollars. That reality naturally pushes capital toward a smaller number of perceived category winners and helps explain why today’s valuation environment looks so different from anything venture investors have experienced before.
Why We Continue To Tread Lightly
It is easy to be excited about venture capital and the future. We may very well be living through the next industrial revolution, or perhaps more appropriately, the AI Revolution. We firmly believe our clients should have exposure to that possibility. At the same time, history reminds us that transformative technologies do not eliminate investment risk. If anything, they often amplify it.
The reality is that we remain in the very early innings of this new technology cycle. The ground beneath our feet seems to shift almost weekly. Perhaps the best example is the foundation model race itself. A year ago, it appeared almost inevitable that OpenAI would emerge as the long-term leader. Within a matter of months, Anthropic became the perceived frontrunner as enterprise adoption accelerated around its coding capabilities. Then, in July, Kimi, an open-source model from China, surprised the industry by demonstrating coding performance comparable to Anthropic. The lesson is not which company will ultimately win. Rather, it is that the competitive landscape remains remarkably fluid, and today’s perceived leader can quickly become tomorrow’s challenger.
Beyond the pace of technological change, there are also important questions surrounding revenue quality and long-term business economics. While the public markets generally rely on well-defined GAAP revenue metrics, the revenue figures commonly discussed in private markets often deserve closer scrutiny. In many cases, reported annualized recurring revenue is calculated by annualizing the most recent month’s revenue, regardless of how sustainable that level of activity ultimately proves to be. It also remains unclear how much of this revenue is truly recurring, how durable customer retention will be, and whether commercial arrangements between AI companies may temporarily inflate reported results.
Aggressive revenue reporting is not unusual during periods of technological enthusiasm. Over time, the market has a way of separating durable businesses from promotional narratives. The more important question, in our view, is not whether AI companies can grow quickly. Many clearly can. Rather, the key question is whether they ultimately exhibit an attractive combination of growth, retention, pricing power, and profitability. Today’s valuations largely assume that they will. While AI has the potential to significantly outperform historical software companies on growth, it remains far less certain whether it will ultimately match them on margins and durability.
Our Perspective
Valuations matter. They always have, and they always will. But we should remain open-minded that “high” is not always synonymous with “expensive.” As we have seen, there is compelling industrial logic behind many of the valuations we are seeing across AI. This time could truly be different.
Our perspective is to be excited by the potential AI Revolution that is unfolding before us. We may be witnessing one of the most important technological shifts in modern history. At the same time, history has a way of rhyming, and there are warning signs lurking in the corners. We sit in the control room for our families’ multi-generational legacy. It is our responsibility to position them to benefit from the AI Revolution if it unfolds as many expect, while protecting them should optimism have run ahead of reality.
Sean Warrington is partner and head of private investments at Gresham Partners.
