The State of AI 2026: From Adoption to Real-World Impact

Twelve months ago, the question was whether artificial intelligence would reorganise the global economy. That question is settled. The one that remains is far more interesting: who is organised to capture it, and how fast?

It is my privilege to introduce The State of AI 2026, the official publication of the Dubai AI Festival. This has been a year of AI breakthroughs arriving at breakneck speed, which has made keeping up a challenge from a publishing standpoint. A report written in one month, may already be outdated in the next.

Let’s start with deployment. Nine in ten organisations now use AI regularly, and nearly one in five working-age adults worldwide use it. But breadth is not depth. Only 44% of organisations say AI is scaling across the enterprise, just 37% see any effect on earnings, and a mere 6% attribute 5% or more of EBIT to it. The gap between AI activity and AI impact is the defining story of 2026, and it is widening by company size: 40% of billion-dollar firms are now scaling AI agents, against 22% of everyone else.

The agents are coming regardless. On SWE-bench Verified (a standard test of whether AI can do real software engineering work), the best systems went from solving 4% of real-world software bugs in late 2023 to nearly 80% by late 2025, and 97% by September 2026, while the cost of GPT-3.5-level output fell more than 280-fold. When capability rises and cost collapses at the same time, autonomy stops being a demo and becomes a procurement decision.

Then, you must follow the money, which is flowing upstream. Global private AI funding reached US $217 billion in 2025; the first quarter of 2026 alone delivered US $226 billion. Amazon, Alphabet, Microsoft and Meta are guiding to more than US $720 billion in capital expenditure this year, nearly five times what they spent in 2023. Anthropic is now valued at US $965 billion, OpenAI at US $852 billion and NVIDIA at roughly US $5.4 trillion. AI is no longer being financed like software; instead, it is being financed like power grids and railways.

Next there is the energy equation, the least glamorous and most decisive variable of all. Data centres consumed around 485 TWh of electricity in 2025, up 17% in a single year, and the IEA expects that figure to reach 950 TWh by 2030. In Ireland, data centres already absorb 23% of metered electricity; in PJM, America’s largest grid market, capacity prices jumped more than ninefold. It’s a bottleneck already showing signs of trouble in some markets, as models ship in months, while grids take years to build.

Megawatts, not just model weights, now decide who can scale — and in 2026, so does the ability to physically protect them. 

Against that backdrop, nations are moving at radically different speeds. More than 80 countries now have national AI strategies, but their playbooks are diverging: Europe has pushed its high-risk AI rules back to December 2027, Washington is prioritising speed and export controls, and Beijing is ordering foreign chips out of state-funded data centres. The United States holds roughly three-quarters of global AI compute, while China holds about 15%. Usage is also varied by geography: more than 70% of working-age adults use generative AI in the UAE, 63% in Singapore, 31% in the United States — and under 10% across much of Sub-Saharan Africa.

And then there is all that is happening here in Dubai. In the final days of April 2026, His Highness Sheikh Hamdan bin Mohammed bin Rashid Al Maktoum, Crown Prince of Dubai, began issuing directives for Dubai’s government — and its entire private sector — to transition to agentic AI within two years. Over roughly six weeks, the UAE committed to moving 50% of government services to agentic AI, launched an 80,000-person training programme, and set out plans to support 295,000 Dubai companies to adopt AI, develop 100 specialised AI assistants and establish 50 agentic AI companies. Where others set ambitions, Dubai set deadlines.

The UAE appointed the world’s first Minister of State for AI in 2017. The Dubai AI Campus at DIFC set a target of attracting over 500 AI companies by 2028, a target that had been achieved by October 2026. The Dubai AI Seal has made trust a condition of public procurement. Regional data-centre capacity is set to more than triple by 2030, much of it built to AI-era specifications from day one. Dubai is not trying to out-build Silicon Valley on frontier models. It is betting on something harder to copy: orchestration — of capital, regulation, energy, talent and demand, at speed.

Inside this report, you will find the data behind every one of these shifts: the kingpins redrawing the industry, ten charts that define where AI stands, the frontiers of agentic AI, healthcare, finance and cybersecurity, and a candid account of the challenges ahead. If 2023 was the year of awakening, 2024 the year of acceleration, and 2025 the year of AI agents, then 2026 is the year of reckoning. The race is well underway.

Welcome to The State of AI 2026.

 

 

Unpack the massive financial flows and the looming infrastructure shifts defining the industry in The State of AI 2026 report.