Since ChatGPT took the global tech world by storm, artificial intelligence has been advancing at an unprecedented pace, driving transformative leaps across multiple domains - from productivity and automation to real-time decision-making. These breakthroughs are reshaping national economies and even the very fabric of human society. It is no exaggeration to say that AI, in its rapidly accelerating evolution, has upended traditional systems and signaled the dawn of a fifth industrial revolution.

However, the rapid development of AI comes at a significant cost, creating a mounting resource crisis. A December 1 article on the World Economic Forum website described this as the AI-Energy Nexus, highlighting that the technology’s remarkable progress relies on massive consumption of energy, water, and critical minerals. Public spending on AI reached nearly $1.5 trillion in 2025 and is projected to hit $2 trillion in 2026. This staggering expenditure is largely fueled by the AI Arms Race, now an undeniable global reality. In terms of energy consumption, Gartner estimates that AI operations in 2025 required 93 terawatt per hour, a figure expected to soar to 432 terawatt per hour by 2030.

Water consumption is also a major concern. A study published in the journal Patterns estimated that AI operations could use between 312.5 and 764.6 billion liters of water annually, compared with the global annual consumption of bottled water, which stands at approximately 446 billion liters.

All of this presents us with a complex problem threatening the global energy grids' ability to accommodate the increasing demand, draining essential water supplies needed for agricultural lands and local communities.

The road ahead

After all that has been said, the pressing question remains: where is AI taking us in 2026?

Firstly, what we won't see in 2026 is Artificial General Intelligence - AGI or Superintelligence.

Some experts argue that artificial general intelligence (AGI) remains out of reach, or may even be impossible. They often reference philosopher John Searle’s “Chinese Room” argument, which distinguishes between syntax and semantics. According to this perspective, AI systems excel at processing symbolic content and manipulating linguistic syntax, but they lack true understanding or semantics - a capability considered uniquely human. Regarding superintelligence, a hypothetical stage that would follow AGI, AI would gain the ability to continually self-improve and evolve, reminiscent of “Andrew” in Isaac Asimov’s "The Bicentennial Man," whose journey of self-modification ultimately challenges and redefines the concept of humanity. While achieving this stage may be impossible, AI continues to advance rapidly, expanding its influence across numerous sectors.

AI Humanoid
In 2025, significant progress was made in “World Models,” AI systems designed to simulate the external world and how it operates, enabling artificial intelligence to gain a deeper awareness of its complex surroundings. These models aim to bring AI closer to human-like intelligence by giving it an understanding of its environment and the ability to handle complex, unpredictable tasks similar to those performed by humans. This advancement has had a major impact on the field of humanoid robotics, with expectations that humanoid robots will increasingly be used in settings like hospitals, where they can interact intelligently with their environment and perform non-repetitive, unprogrammed tasks resembling human work.

Specialized Language Models

Institutions are increasingly shifting from using general-purpose Large Language Models (LLMs) to Small Language Models (SLMs) or domain-specific language models (DSLMs), tailored for specialized fields such as law or medicine.

These language models are trained on data related to that specific field, making them highly precise and focused. This data restriction aims to achieve two main goals:

(1) Improving model accuracy and output by reducing "hallucinations" present in general AI models.

(2) This shift allows AI to achieve a deeper and more precise understanding of specialized domains, including the nuanced terminology unique to each field. Notable examples include Med-PaLM 2 in medicine and BloombergGPT in finance.

Sovereign AI
Sovereignty has emerged as one of the key themes in AI this year, and it is expected to grow rapidly in the coming years. Sovereign AI, a priority for many governments, represents a strategic goal for states seeking digital independence. This approach involves controlling the full lifecycle of AI - from locally owning the infrastructure, such as GPU clusters and national or regional cloud computing, to ensuring that data remains within national borders and complies with local privacy regulations, such as GDPR.

Countries are increasingly seeking to ensure that AI systems reflect their cultural identities, languages, and unique ethical standards by developing local talent and creating models trained on domestically sourced data.

This focus on sovereignty has become a crucial element of national security and economic stability, helping countries reduce dependence on foreign tech giants and potential rival powers. By transforming AI from an externally sourced service into a domestic foundation, states can foster local innovation, safeguard national data privacy, and maintain their competitive advantages. At the same time, this approach allows countries to preserve their cultural uniqueness without compromising digital sovereignty or information security.