Just over a month into 2026, the artificial intelligence landscape looks significantly different from the era of unchecked scaling and generalised models that defined the 2021-2024 period. The past year has brought a clear shift in direction, showing that brute-force computing and ever-larger models are no longer driving meaningful progress.
2025 taught us that simply adding more GPUs no longer guarantees proportional improvements in AI performance. Research shows that each new generation of LLMs (large language models) now delivers incremental gains rather than the breakthroughs seen in the early years of development, and the economics reinforce this reality with McKinsey estimating that companies will need to invest £4.1 trillion into data centres by 2030 to meet global demand. The cost of progress is rising, and the need for efficiency is urgent.
This marks a fundamental change in the economics of AI development. Organisations once expected heavy investment in compute to translate directly into stronger models. Today, achieving comparable gains requires far greater resources. More investment no longer means more output.
The rise of deeper, specialised language models (SLMs)
Perhaps the most important development of 2025 was the realisation that value in AI comes not from trying to solve every problem, but from building expertise in specific domains. Generalised models attempting to cover all operations often act as a jack-of-all-trades and a master-of-none. In contrast, specialised language models deliver stronger results by focusing on getting the narrower, more specialised needs fulfilled.
This trend is expected to accelerate in 2026. Funding for vertical AI startups grew by more than 70% in 2025, led by healthcare, legal technology and industrial automation, and recent data shows that healthcare and financial services already account for 58% of total vertical AI value.
The shift is particularly obvious in highly regulated sectors such as finance and investing, where accuracy, transparency, and explainability are fundamental. Companies using vertical AI models report up to 30-40% lower loan delinquency rates.
Financial services providers are increasingly recognising that regulatory-compliant, explainable AI built specifically for equity and fund analysis delivers far more value than general-purpose tools. Bessemer Venture Partners projects that the market capitalisation of vertical AI could grow 10 times compared to traditional LLMs.
Regulatory clarity as a USP
Another key takeaway from 2025 was that regulatory compliance and explainability are no longer box-ticking exercises; they are essential to stay competitive. Teams that embedded compliance and transparency from the outset moved faster and built greater trust than those that treated regulation as an afterthought.
The European Union’s AI Act, which came into force in August 2024, has set a global benchmark for AI governance. Its risk-based framework imposes strict requirements on high-risk systems, with fines of up to €35 million for non-compliance. High-risk AI systems will be required to meet these standards from 2027, making 2026 a critical year for preparation.
The limits of AGI hype
Artificial general intelligence will dominate headlines throughout 2026, but maintaining a realistic outlook is essential. Without major breakthroughs in hardware architecture, most claims are likely to outpace what is technically possible.
The gap between hype and real-world capability is widening and while excitement around AGI is understandable, investors must remain realistic about its impact. 2026 is still unlikely to be the year AGI delivers any profound change, rather it will be a year of incremental progress towards that becoming a possibility.
A tighter competitive race
Competition among leading LLMs providers is growing. In 2025, Google’s Gemini increased its market share from 5.4% to 18.2%, while OpenAI’s ChatGPT fell by 19.2%. By early 2026, ChatGPT’s share had declined further, and Gemini surpassed 650 million monthly active users, adding 200 million since July 2025.
Google’s growth has been driven by deep integration across more than three billion Android devices and Google Workspace products. Following Google’s launch of Gemini 3, OpenAI CEO Sam Altman reportedly declared an internal “code red” situation, reflecting just how competitive the market has become.
Rather than a single dominant player, 2026 is shaping up to offer more balance, with SLMs consolidating specialised-industries.
The surge of vertical-first AI companies
These market shifts create major opportunities for specialist, vertical AI companies. As scrutiny of AI increases, trust and compliance are becoming essential for success and in sectors such as finance, law, and healthcare, models that cannot explain their decisions or meet regulatory standards risk market exclusion.
2026 is likely to be the year when vertical-first AI companies truly come into their own. From wealth technology to legal and health technology, organisations that combine deep domain expertise with AI capabilities are offering value that generalised models struggle to reach.
The investment intelligence sector illustrates this clearly. Partnerships between specialist AI providers and major financial institutions, including stock exchanges and securities firms across Asia, Europe, North America, and LatAm, reflect strong demand for transparent, regulatory-compliant AI built specifically for investment analysis.
Navigating the year ahead
For investors and institutions, 2026 presents both risk and opportunity. Success will depend on distinguishing genuine innovation from hype, and depth from breadth.
The organisations most likely to thrive are those that recognise AI’s next chapter is not about doing more, but about doing the right things well. Vertical AI and SLMs, applied to clearly defined problems, supported by strong domain expertise, robust regulation and a commitment to transparency, will decide who comes out on top in 2026.
Author: Gaby Diamant, CEO and Co-founder of BridgeWise
