The excitement hasn’t disappeared, and neither has money. But the tone of investment has changed – investors are no longer impressed by the size of your model or the number of GPUs you’ve secured. They want to know one thing: Is your AI actually creating value – and is it generating revenue?

Because the market finally started asking harder questions, the discussion around the “AI bubble” has become a lot more interesting. Are we still in a hype-driven reality, or have we crossed into a period where AI has to prove that it can hold its ground?

Let’s try and unpack this.

The hype phase is fading – and that’s a good thing

Through 2023 and 2024, simply attaching “AI” to your pitch deck was generally enough to get the interest of investors and arrange a meeting. In 2025, that no longer works.

For example, OpenAI’s $40 billion raise in Q1 this year is often attributed to the late-2024 momentum. But Anthropic’s funding round in September showed more modest performance, with only $13 billion.

Investor pacing is now slower and more cautious, and AI firms face heavier scrutiny over their revenue models and regulatory alignment. This isn’t a crash, exactly – more of a correction in attitude.

Major VC funds, especially those active in Europe and Singapore, now openly talk about “AI fatigue”: not fatigue with the technology itself, but with AI-washing, where companies exaggerated their capabilities without a viable product behind them.

At the same time, large AI infrastructure players – AWS, Google Cloud, and even regional providers – have started tightening compute access and raising prices. AI workloads come with ever-higher processing requirements, so providers have to be stricter about allowing open-ended experimentation, and that, in turn, pushes startups to focus on efficiency.

The sentiment is clear: where before the AI sector celebrated scale, now it demands sustainability.

Global compliance pressures are forcing a market reset

With the introduction of the EU AI Act, it's impossible to ignore the fact that Europe is quietly reshaping what AI maturity looks like. AI startups can no longer just build fast and worry about regulation later. And compliance can no longer be treated like a mere checkbox.

And this isn’t just happening in Europe – the shift is echoing globally, with other jurisdictions also adopting similar stances. Even the US, which is a lot less restraining when it comes to AI innovation, is learning to ask European-style questions during due diligence: Where is your training data sourced from? How do you audit model bias? Can the product be deployed without compliance risk for enterprise clients?

AI startups now have to pay greater attention to their architecture and meet growing transparency and data governance requirements. All the while, investors are adopting new evaluation logic, treating compliance readiness as a sign of long-term viability.

Some might say that the growing regulation has slowed AI down, but it’s far more accurate to say that it’s filtering out noise, leaving players that can actually measure up to the upscale in standards.

Is there still a bubble? Maybe – but it won’t end like past tech crashes

There’s no denying that some AI valuations are still inflated. We see early-stage startups with limited revenue being priced at levels normally reserved for proven SaaS companies. Yes, a correction will come for that segment of the market.

But here’s what differentiates AI from past bubbles like VR or 3D printing: AI is already embedded in mission-critical workflows, especially in sectors like compliance automation, financial risk detection, cybersecurity, medical diagnostics, and logistics optimisation.

Banks are quietly deploying AI not just for chatbots – although there is that, as well – but also for things like regulatory monitoring and fraud detection at scale, which may be less noticeable but are far more impactful.

Healthcare systems in Germany and the UAE are using smaller, specialised medical AI assistants to reduce manual documentation times. And these are not overhyped demos, either. These are operational improvements that can deliver real, tangible impact.

AI isn’t deflating – it’s maturing. The market is learning to separate measurable performance from empty glitter. And that’s a healthy change.

Stepping beyond the hype

By this point, we have entered a post-hype era. And that is where the most serious breakthroughs are going to occur.

The companies that thrive in AI from now on are not the ones talking the most about disruption and innovation – they are the ones putting in real effort to reduce costs, cut manual workloads, and de-risk operations.

This is especially clear in Europe’s fintech and banking sector, where AI is no longer pitched as innovation – it’s pitched as an operational necessity, especially for mid-sized institutions that cannot scale teams as fast as regulatory demands increase.

Instead of asking, “Is the AI bubble going to burst?”, the more relevant question now is this: “Will your AI company still matter when the term ‘AI-powered’ stops being impressive and becomes the norm?”

The way I see it, 2025 was the year the air was let out of the bubble just enough for the technology to go back to solid ground. The hype cycle is over, and the impact cycle has begun in its place. And the companies that win in this new phase will not be the loudest, but the ones that quietly solve real market problems, comply with regulations, and deliver measurable outcomes.

That’s not a bubble bursting. That’s just technology finally growing up.

For more startup news, check out the other articles on the website, and subscribe to the magazine for free. Listen to The Cereal Entrepreneur podcast for more interviews with entrepreneurs and big-hitters in the startup ecosystem.