The European applied AI market is in the middle of a significant boom. Agentic and generative AI startups on the continent have collectively raised €20bn since 2024, according to Sifted data, with €8bn secured this year already.
Companies leading the surge in GenAI include the likes of synthetic voice decacorn Elevenlabs, which raised $500m at the beginning of the year.
Meanwhile, a cohort of agentic AI startups are creating autonomous agents to streamline workflows. European companies leading in this space include Swedish legaltech Legora and Berlin-based workflow automation company N8n.
But as the market matures, companies must consider how they can sustain themselves in the long-term.
“Successful companies will build strong customer adoption, defensible market positions and develop a product that has lasting relevance,” says Dr Ling Ge, chief investment and strategy officer for the EMEA region at global technology and entertainment company Tencent.
In an interview with Sifted, Ge, alongside Archie Hollingsworth, cofounder of AI email assistant Fyxer, and Sauraj Gambhir, cofounder of AI foundation model developer Prior Labs, unpack how AI companies can prove lasting durability and what the next stage of the market has to offer.
Moving beyond the hype
The “honeymoon phase” of experimentation in the European AI market has concluded, according to Ge.
Investors and founders are now looking closely at how AI products can reliably generate value over time.
Tencent has been embedding AI across its products and services for nearly a decade, Ge tells Sifted. “We don’t necessarily invest in the companies jumping on the bandwagon or building what’s trending in that moment,” she says.
Instead, Tencent looks for founders with a long-term perspective."The best founders aren’t building for the next funding round; they’re building products customers will still depend on in five or 10 years."
It's not just about building the models, but how they connect back to real human communities.
Tencent’s investments in Europe span video games and digital entertainment companies as well as AI and emerging deeptech companies. These include Finnish mobile gaming giant Supercell and Horizon Quantum, which focuses on software infrastructure for quantum applications.
When navigating the changing market, the biggest challenge for Fyxer was proving its product could solve customers’ problems.
"Enterprise buyers want to know two things: is my data safe, and does this actually work? So we built around both: client data never trains third-party models, and we never send an email without a human reviewing it first."
With AI startups emerging globally at a rapid pace, to cut through the noise companies should aim to solve genuine problems with their products, says Ge.
“AI is increasing productivity, but it has also dramatically expanded the number of companies investors need to evaluate,” she says.
"Tencent is more interested in the companies making a lasting difference. It's not just about building the models, but how they connect back to real human communities."
Capital efficiency
Building AI products involves a long-term investment cycle, and investors are increasingly placing greater emphasis on capital efficiency.
Today, it’s common to see companies devote a large amount of their funding to compute, according to Ge. But sustainable growth comes from combining technology investment with customer value.
“One of the biggest risks is scaling before there is enough evidence of customer demand,” she says. “We often see companies invest aggressively before they fully understand where the long-term value lies or how customers will use the product in practice.”
Ge advises companies to first build a unique product based on proprietary data, domain expertise or workflow integration that can’t be easily replicated. Once product-market fit is established, they can then commit capital to scaling.
Those companies might also find that when they go to market to sell, they can find buyers in Europe, rather relying heavily on US customers, a situation industry insiders have wanted for a while.
“Enterprises are increasingly willing to buy from European providers rather than defaulting to US vendors,” Gambhir says.
‘Pilot pipelines won't be enough’: The next stage for the AI market
Startups shouldn’t simply assume foundational AI models will keep improving and they can build on top of them, says Hollingsworth.
We've spent years watching how professionals actually handle their inboxes. A foundation model built in a lab can't learn that.
The companies that will define the next era of the AI market are the ones that understand their customers better than any general-purpose AI ever will.
"We've spent years watching how professionals actually handle their inboxes. A foundation model built in a lab can't learn that." he says.
Alongside developing a moat around proprietary data and domain expertise, companies also need to rethink their core internal teams, adds Ge.
The most critical operational hire for developing AI companies is what she calls a ‘product empath’ — someone who can understand customer pain points and translate these into products that can solve them.
The companies that get ahead are also the ones combining both AI and human knowledge. For example, when building AI for scientific research, the goal is not to create an autonomous scientist, but rather a co-scientist, says Ge.
“The objective is not to replace the scientist but to reduce data-processing tasks so scientists can spend more time on hypothesis generation, experimentation and discovery.”
You need the best people who can hold both sides of the equation — pushing the science forward while staying focused on what actually matters to a customer.
While funding into European AI startups has increased, there’s also growing scrutiny from investors, adds Gambhir. “Exploratory budgets and pilot pipelines won't be enough. Expectations are shifting toward demonstrated enterprise adoption and genuine retention.”
The companies that last will be the ones with genuine research and the discipline to keep investing in it even as the landscape shifts, he adds.
“The field is moving fast enough that what worked six months ago might not work anymore. You need the best people who can hold both sides of the equation — pushing the science forward while staying focused on what actually matters to a customer.”
Gambhir predicts three types of applied AI companies will emerge over the next five years.
“Foundation model labs, as Europe has exceptional research depth, and a handful of those will scale into serious global players.
“Neo labs and companies with domain-specific post-training . These are businesses with genuine domain or modality expertise built on top of foundational capabilities.
“And AI-first companies in traditionally slower-moving sectors such as healthcare, energy, finance and manufacturing.”
