The AI boom hit new heights in 2025, capturing close to 50% of all global investment, up from 34% in 2024. Its dominance doesn’t look set to disappear any time soon, but the impact of the recent ‘doomsday’ report from Citrini Research, which caused tech and financial stocks to tumble demonstrates how precarious the current situation is.
Alongside all the hype and excitement comes a huge dollop of caution and uncertainty, which smart investors must be wary of. JP Morgan’s Jamie Dimon noted recently that there are people ‘doing dumb things’ in the name of AI, and it seems likely many of these will rise to the surface in the not-too-distant future.
But that doesn’t mean there aren’t fantastic investment opportunities for those who look beyond the hype. So, as we close out Q1 of 2026, here are a few AI trends that many have overlooked.
LLM bubble to deflate as commoditisation takes over
The cyclically adjusted price earnings ratio shows equity valuations are near the highest level since 1880, suggesting a correction is on the horizon. The AI boom has delivered some fantastic businesses, but fundraising has been too easy, given that many companies are ultimately competing for the same pie including trying to match Google’s AI-first strategy and stratospheric levels of investment.
As a result, 2026 will be dominated by the fight for infrastructure dominance between the big LLMs – GPT, Claude, Gemini, Grok etc. – and the winners will achieve a global scale that will be hard to surpass. For customers, the upshot will be a rapid fall in prices, as AI foundation models become increasingly commoditised, making them more accessible for a wider audience. For the less popular providers, it could mean a quick slide into irrelevance.
Added to this, too many of the VC-backed businesses raised at huge valuations are just wrappers for foundation models that are attempting to verticalise their technology for a particular sector. These businesses will struggle to exist in a world of LLM dominance, while competing with end-point solutions differentiated through proprietary data and sector expertise.
In short, a pop of the AI valuation bubble – or at least a partial deflation – seems inevitable later this year.
Value will flow down to the application layer
But, as the big LLMs fight it out and prices come down, this is driving exciting opportunities at the application layer specifically within vertical sectors where companies can leverage unique data sources. This is where much of the innovation and productivity gains will happen in 2026 and moving into 2027.
In 2025, a subset of vertical SaaS companies was busy building products in the background, experimenting, prototyping, and testing. Now they’re ready to come out of the shadows, sparking a wave of these solutions moving out of the sandbox and into real life.
We’re seeing this in our own portfolio companies, which, between them, have developed an AI pricing agent (PriceShape), an employee churn prediction product (Relesys), and a suite of AI tools to identify cyber risks and assess cyber readiness (SecureFlag), which have all gone live this year.
Unlike VC backed “AI native companies”, these have all been developed by existing SaaS specialists, with trust in the market, strong sector relationships, understanding of the context and the ability to match and influence real customer workflows. Earned over time, this understanding isn’t easily replicated by the AI-native newcomers, who lack the same background and expertise in the market. In fact, MIT research found that a lack of workflow knowledge was behind the failure of 95% of enterprise Gen AI pilots to generate value.
The other characteristic that marks out these businesses is that they’re still small and agile enough to respond quickly to the AI opportunities coming thick and fast. Many are also still led by the original founding team, giving them an extra depth of knowledge, not to mention vision, to drive through AI innovation and change effectively.
Plus, with their modern tech stacks, maximising data and developing new products isn’t the headache it would be for mid-market ‘frankenstacks’ built over numerous years of evolution and M&A. With high volumes of proprietary data, they can quickly maximise the power of the foundational LLMs, creating a valuable moat between them and new entrants.
AI attackers drive defence
Another AI risk turned opportunity is in cybersecurity, an inescapable theme of 2026, as CrowdStrike recently warned that AI-enabled cyber-attacks have nearly doubled in the last year. Businesses are having to ramp up investment to get ahead of the hackers, while cracking down on how employees are using AI tools.
Microsoft recently warned that 71% of workers have used unapproved AI tools at work and there are concerning reports about the impact of using LLMs for coding BaxBench, a benchmark from ETH Zurich, UC Berkeley, and INSAIT that evaluates LLMs on security-critical backend coding tasks, found that 62% of solutions generated by even the best models are either incorrect or contain security vulnerabilities. This demonstrates why using untested AI products or developing them in house with vibe coding is fraught with risk.
These trends bring clear opportunities for businesses that can help clients combat the AI cyber threat, through training and security solutions. The global cybersecurity market is predicted to almost triple from $22.4bn in 2023 to over $60bn in 2028.
We’re seeing this firsthand with our portfolio company, SecureFlag, a secure coding training platform, which has just launched a new programme on managing teams in an AI-driven workplace? Solutions like these are set to see huge demand in the months and years ahead.
Looking beyond the AI hype
It is easy to get lost amid the AI noise in 2026, but for those who don’t lose their heads, there are massive opportunities to be had. The key is cutting through the hype to identify solid businesses with real knowledge and experience, using AI to solve real problems. Or, alternatively, seeking out opportunities to help others navigate this unpredictable new world of AI – and the many threats that come with it.
