Web Summit 2026 made it clear that the AI investment landscape is moving past its gold rush phase and into an era of cautious recalibration.
Beneath the usual fervor, there was a notable shift from the broad “AI is big” sentiment to the specific investor need for quality.
Several announcements highlighted the forthcoming potential for massive mainstream adoption, and partnerships in innovation announcements from governments and academia in spaces including quantum computing, life science and cleantech secured the narrative of AI becoming increasingly embedded in everyday life.
That said, not every part of the stack is equally investable, and after years of exuberant spending, investors are looking for where the value will stick in a crowded market, and the market is demanding signs of real economics, not just narrative momentum.
Web Summit 2026’s venture conversations pointed to a market that is getting more disciplined, more specialized and less forgiving.
A more selective investment market
Across panels on the state of VC’s role in tech, the message was clear: there is still a huge opportunity in AI, but the market is becoming more selective, more specialized and more demanding of both investors and founders.
A common thread amongst panelists is that venture is becoming more specialized.
Joe Ross, a partner at Entrepreneur First, argued that the AI era has pushed growth investing and early-stage investing further apart, making it harder for large multi-stage firms to credibly do both well.
“When you think about the sort of capital-intensive rounds that folks like the LLM providers are raising … you can only really do that if you’re a large platform that has a ton of capital to deploy,” Ross said.
Panelists contrasted specialist seed firms with large growth platforms, suggesting that early-stage and later-stage companies are best served by different kinds of investors.
That also raised the question of what it takes to win funding now, especially as revenue expectations rise for AI-focused companies.
During a discussion that centered on the implications of mega seed rounds, Ross said he has seen seed rounds grow from around US$400 million to over US$2 million throughout his career.
“There’s a continued trend upwards,” he said. “Now, yes, we see one or two absolutely huge deals, but I think they are fairly unique. I think they are really based on a very specific team thesis for that. And I think that if you look in our portfolio historically over the last 10 years, the size of the round does not absolutely correlate with the eventual outcome.”
Catherine Ouellet-Dupuis, a general partner at White Star Capital, echoed the sentiment and noted that smaller rounds allow for flexibility and agility.
“I don’t know where the world will be in one year and two years from now, and the founders don’t know it either. So if you raise a lower valuation, then there’s more flexibility to pivot and go after the opportunities,” she said.
Finding the real value in the AI stack
The consistent message was that the value lies in what a company owns, what data it controls and what makes it hard to replace.
For Salil Deshpande, a general partner at Uncorrelated, the AI market is a three-tier stack with infrastructure and hardware at the base, model providers in the middle and applications at the top. He said he is moving down the stack to find value.
The crowded middle layer, according to Deshpande, is the primary “trouble spot” for investors, a core issue being recurring costs of model retraining aggressively cannibalizing profits. The result has been low or even reverse gross margins for several companies.
Simultaneously, the rapid advancement of open-source models is stripping away proprietary advantages, leaving model providers with little room for long-term differentiation.
At the top layer, both Deshpande and David Cohen, founder and CEO of Techstars, remain skeptical of so-called thin wrappers, simple applications that sit on top of general‑purpose models without proprietary data or workflows.
By their view, the clearest answer to “where is quality?” is in the infrastructure, chips, power and cooling, and in sectors like healthcare and space, where regulatory barriers and physical-world complexity create real protection.
Deshpande and Cogen also said they expect a shift toward pay‑for‑performance and pay‑for‑utility models over seat‑based pricing as agents replace software tools.
The bottom line
For investors, the takeaway is that the smart money is looking for more than ambition or AI branding; it needs a real product, a real moat and a plan for turning capital into proof.
Don’t forget to follow us @INN_Technology for real-time news updates!
Securities Disclosure: I, Meagen Seatter, hold no direct investment interest in any company mentioned in this article.
