Interviews
Thomas Cuvelier, Partner, US & Europe, RTP Global – Interview Series
Thomas Cuvelier, Partner, US & Europe, RTP Global is a New York–based investor focused on backing founders across the United States and Europe who are building category-defining companies. He is particularly drawn to entrepreneurs driven by firsthand experience solving real-world problems rather than those who set out simply to start a business. With a background in Computer Science and Statistics, he invests in technically strong teams from Pre-Seed through Series A, typically deploying $1 million to $10 million. His investing experience spans the full company lifecycle, from early-stage venture to growth and late-stage buyouts, shaped by time spent across major global tech hubs including San Francisco, New York, London, Berlin, and Paris, and brings a strong cross-Atlantic perspective on technology, talent, and markets. He is also fluent in French, Mandarin, and Japanese.
RTP Global is a global early-stage venture capital firm that invests in technology-driven companies across North America, Europe, and Asia. Founded in 2000, the firm focuses on backing ambitious founders at the Seed and Series A stages, with a track record that includes companies such as Datadog, Delivery Hero, and Cred. RTP Global operates with a long-term investment philosophy, often supporting companies for many years, and reinvests capital from past successes into new ventures. Its global presence and network enable founders to scale internationally, while its sector focus spans areas such as AI, fintech, SaaS, and data infrastructure.
You’ve backed a wide range of AI-native startups, including workflow orchestration platforms like Kestra, robotics data platforms like Mecka AI, and vertical SaaS companies such as Archy and DualEntry. From your experience investing across these different sectors, what patterns are you seeing in the types of AI startups that are gaining real traction today?
An interesting pattern can be seen in the profile of founders behind successful AI startups. I’m increasingly seeing entrepreneurs motivated by, and building, AI-native solutions to pain points they’ve encountered
It’s also worth noting that traction is being generated across a broad landscape. Heavy industries that were once ‘invincible’ to disruption from startups, like robotics and government services, can now be disrupted and are viable for VC investment.
Recent launches from large model providers such as Anthropic have triggered debate about whether vertical AI startups still have defensible moats. How do you see the relationship evolving between foundational models and application-layer companies?
Anthropic’s launches do pose a threat to the startup ecosystem but we need to be precise. Companies that offer SaaS to SMBs and unregulated industries are vulnerable. They don’t have a moat that foundational model companies can’t breach with effective surface-level tooling.
But the appeal of surface-level tooling only stretches so far. Complex workflows in regulated industries are best served by AI applications with deep integration with industry-specific tooling that have been designed by teams with deep familiarity of their industry. The vertical AI startups behind such applications do have defensible moats and can weather the storm.
You’ve suggested that some Series B and C AI startups may now face structural challenges due to rapid advances in foundation models. What architectural mistakes did these companies make, and what lessons should younger startups take from that?
The problem that many AI startups at the Series B and C stage face is that their products are, at their core, dashboards to interface with LLMs that have only a handful of underlying integrations. These companies made quick traction with hobbyist and enterprise adoption over the past few years and raised large amounts. But, fast forward to today, and the foundational model players are well-placed to challenge their business model.
For new founders, the lesson is to avoid taking too much comfort from any perceived technology moat or historical customer lock-in. AI-assisted software development and business scaling makes today’s breakthrough vulnerable to becoming tomorrow’s commodity offering. Building moats around traits that are still hard to replicate – depth of integration, depth of industry-specific expertise
When you evaluate AI startups today, what actually constitutes defensibility? Is it proprietary data, workflow integration, regulatory complexity, distribution, or something else entirely?
I’d say, yes, all of these are important. I can’t stress the required depth of integration enough, however. We’re talking about integration with 100-plus industry-
As mentioned above, I do think that foundational model pushes into vertical industries can only stretch so far. Regulated industries and their critical workflows are too complex for foundational models and here lies the best gap for AI startups to fill and build defensible businesses.
Other than what you list, I’d also add that user retention and community building remain important building blocks of defensibility. Genuine user loyalty because a product is great and a pleasure to use is hard to build up but also hard to disrupt.
Many investors are now saying that the technology moat in AI is shrinking. Do you agree, and if so, what new forms of competitive advantage are emerging?
It’s a fact that the technology moat is shrinking. If your product can be built over a weekend, then you don’t have a technology moat – and the art of the possible from a weekend with AI keeps getting more sophisticated.
In a world where software is easy to replace, customer loyalty bolstered by amazing user experience becomes crucial.
From a venture capital perspective, what qualities are you looking for in founders building AI companies today? Are there specific traits, experiences, or ways of thinking that stand out when you decide to back a founder?
Personally, the #1 trait is crystal clear clarity on end customers. When founders know clients like the back of their hands, and how their product fits into the customer’s reality, then other important elements for building successful AI companies – like being guided by deep sector expertise – fall into place.
Other important qualities are a product-focused founder
You’ve invested in companies operating in regulated sectors such as insurance, healthcare, and financial services. Why do you believe regulated industries may offer stronger long-term opportunities for AI startups?
Industries like healthcare, pharmaceuticals, financial services, insurance and the public sector are ripe for disruption by AI startups that are built from the ground up with each industry’s needs and workflows in mind. The impact of LLMs and agentic AI is so revolutionary that it won’t just ‘pass by’ industries that have historically been slow to adopt technology.
These industries are so compelling for startups because they’re still dominated by legacy software incumbents that can’t easily pivot their products for AI. And, as mentioned, they’re less suited to disruption from foundational models.
There has been increasing discussion about a potential wave of mergers and acquisitions among mid-stage AI companies. What signals are you seeing that suggest consolidation could accelerate in the next couple of years?
The clearest signal is valuation compression meeting cash pressure simultaneously. A lot of mid-stage AI companies raised at 2021-2023 peak multiples and are now facing down-round territory if they go back to market, which makes a strategic acquisition suddenly more palatable than a humiliating re-price. On the acquirer side, the hyperscalers and large platform companies have spent the last two years integrating AI capabilities and are now realizing that building everything in-house is slower than buying differentiated distribution or proprietary data assets.
The second signal is the talent and data moat dynamic: as foundation model commoditization accelerates, the defensible value increasingly sits in proprietary training data and go-to-market rather than model architecture itself, which is exactly what acqui-hires and strategic roll-ups capture efficiently.
Finally, the regulatory environment is quietly becoming more permissive; the current US administration has signaled a lighter antitrust posture, which removes barriers for large-cap strategics who were previously cautious about deal scrutiny. Put it together and you have motivated sellers, motivated buyers, and a clearer regulatory runway: that combination historically precedes an M&A wave.
AI startups can reach early traction quickly, sometimes hitting their first million in revenue faster than previous generations of SaaS companies. But scaling beyond that seems to be where many struggle. What separates the companies that break through from those that stall?
The dividing line between AI founders hitting revenue ceilings and the founders that repeatedly smash through milestones is go-to-market velocity. Shipping features and products guided by a user community feedback loop. Not thinking twice about global growth. Making the right sales hires. Bulletproof distribution channels. These are crucial tenets of go-to-market velocity that set AI startups apart from the pack.
Looking ahead three to five years, what types of AI startups are you most excited to back right now, and which categories do you believe are already becoming crowded or vulnerable?
I’m feeling excited about possibilities for AI to disrupt healthcare and pharmaceuticals (full of complex and highly regulated workflows that can be disrupted), industrials (where technology disruption in general still has miles to run) and financials (particularly compliance and accounting/ERP).
The market for high-level, SMB-facing AI products is saturated and an area of startup vulnerability. It’s the same case for any AI startup with a product that can be easily replicated. Technology moats aren’t dependable anymore.
