The term ‘AI frenzy’ may be overused, but it’s perfectly accurate when talking about venture capital investment. One in five new unicorns is building AI agents; the tech dominates investment capital, representing 53% of all new billion-dollar companies in 2025 so far. It’s contributed to the largest growth of wealth in history, with existing AI unicorns holding a combined valuation of $2.7 trillion.
But we can’t conflate valuation with real value. Data from Silicon Valley Bank shows that AI makes up almost half of all VC investment, with startups developing apps using the tech reaching unicorn status in just four and a half years, compared to eight years for application and infrastructure companies and six years for cybersecurity. This surge has coincided with a broader slowdown in unicorn exits and the birth of the ‘zombiecorn’: companies with weak revenue growth and poor unit economics. According to CB Insights, there are now a record 1,200 venture-backed unicorns yet to go public or find acquisition partners in 2025.
The momentum driving this investment is rooted in the assumption that these companies will continue to increase in value. But this belief has wavered in recent weeks with a stumble in tech stocks, which wiped more than $1 trillion from U.S. markets in just four days. The downturn was triggered by an MIT report revealing that roughly 95% of corporate generative AI pilot programs delivered “little to no measurable impact.”
AI thoroughbreds vs. AI unicorns
There is, certainly, immense value in AI. It will likely transform all major industries through large-scale automation, workflow and resource allocation optimisation, and by enabling faster, more accurate decision-making. However, successful investment requires a technical understanding of the space to identify genuine value creation in startups.
Firms with these strong fundamentals could be described as ‘thoroughbreds’ – the term coined by Saul Klein for ventures generating at least $100m in annual revenue.
Unlike unicorns, which are defined by valuation alone, thoroughbreds demonstrate sustainability through consistent revenue growth at reasonable investment levels. Klein goes further in sketching a spectrum of company maturity. Startups with potential to grow into thoroughbreds, ventures at $25m in annual turnover, are colts.
This framing is especially relevant in AI, where pilot programs and inflated valuations often mask weak revenue models and potentially high churn. A focus on thoroughbreds shifts attention back to factors such as returns and market fit that show a company is equipped to endure market cycles or the panicked sell-offs we’re seeing currently.
Avoiding the wrapper-ware trap
So how do you spot a thoroughbred, colt, or at least a company with the potential to become one? Investors with technical backgrounds know to avoid the wrapper-ware trap: startups that don’t build core technology but simply wrap existing tools like ChatGPT into new interfaces.
Today, folding an LLM into your product is enough to claim an ‘AI badge.’ That’s perfectly natural – and often trivial to implement – but it won’t deliver durable returns. These products can look impressive in pitch decks, but they solve shallow problems and are easily replicated. More troubling, they’re crowding out investment in deeper R&D that could address real structural challenges.
Understanding the fundamentals of AI infrastructure – open-source technologies, data infrastructure, foundational models – allows VCs to distinguish genuine value from hype. The key is backing companies that offer real solutions to industry-specific issues in overlooked markets, focusing on the long-term trajectory and investing in AI’s less glamorous aspects: infrastructure, automation, and vertical-specific software that provides high value for users.
The perfect market fit
The most promising AI investments target sectors suffering from labour shortages and operational inefficiencies. AI-enabled automation of physical work through robotics in industries like social care, manufacturing, and logistics offers enormous potential for productivity gains. These aren’t flashy consumer apps, but they address genuine economic needs that traditional software couldn’t solve. Firms that can identify and build these solutions with a clear market-fit have the potential to mature into AI thoroughbreds.
We’re also seeing AI-enabled services that can fully acquire and operate entire firms. Rather than building software to sell to service industries, startups are automating operations and buying up traditional firms in order to roll them up. This represents a fundamental shift from selling tools to becoming the business itself.
It mirrors what we’ve seen in financial services, where banks increasingly look to external vendors for AI implementation rather than building in-house. The most successful partnerships emerge when technology companies understand specific regulatory and operational constraints of their target sectors, rather than imposing generic solutions.
Europe’s investment strategy as an advantage
Europe is uniquely positioned to lead this more disciplined approach to AI investment. The continent’s venture firms are more interested in the hard numbers that reflect sustainable business models, thanks partly to more stringent financial-disclosure rules that European firms must follow.
It’s true that US tech startups are 40% more likely to have raised VC funding after five years than their European counterparts. But this capital abundance may actually be a disadvantage when it comes to building sustainable businesses. According to PitchBook’s Q2 2025 European Venture Report, when rising interest rates and inflation in 2022 put the brakes on VC investing globally, Europe emerged as a safer haven for capital.
This resilience reflects a fundamental difference in investment philosophy. European investors, constrained by smaller fund sizes and stricter regulatory oversight, have been forced to develop better due diligence practices and focus on companies with genuine revenue potential rather than just growth metrics.
It’s a more measured approach that may prove to be the winning strategy for returns on AI investment. By focusing on ‘thoroughbreds’ building business tools with the right market fit, investors can separate lasting value from hype.
This investment strategy goes hand in hand with revenue generation, rather than the excessive value inflation seen across the Atlantic. While Europe’s cautious, revenue-driven approach may not grab headlines like Silicon Valley’s billion-dollar bets, it does offer a blueprint for sustainable returns.
Tom Henriksson is a general partner at OpenOcean
Read more: Are you ready for the AI risks in your supply chain?
