Welcome to 2026. This is the year that AI continues its systematic journey towards pragmatism.
There will be no AI bubble in 2026. Regardless of how people interpreted MIT's State of AI in Business 2025 report on ROI, there's nothing wrong with the ROI of AI. (How do you measure the ROI of a proof of concept? And why measure the ROI of a POC?)
As a reminder, AI is still in its infancy, and this is especially true for enterprise AI. The majority of companies have less than three years of experience with AI. This means that in most organizations, the structure, culture, processes, and lifecycle management of AI are not fully formed or mature.
Companies that have successfully applied AI (for example, companies like Adobe, Salesforce, ServiceNow, and SAP) are a relatively small group of AI innovators with one thing in common. It's that they started working on AI five to 10 years ago and have approached the technology as a long-term investment. During that time, these organizations learned what was needed for structure, culture, processes, and lifecycle management. Once generative and agentic AI became a reality, they quickly adapted and accelerated.
With all that said, research from Omdia, a division of Informa TechTarget, repeatedly shows that companies are enthusiastic about AI. In 2026, organizations will accelerate their investments and efforts to continue the tedious-sounding but important task of operationalizing AI within their organizations.
Is this a prediction? No, it's a trend. It has been going on for several years and will continue for years to come. Let's move on to some predictions.
Free ChatGPT is shut down or severely restricted. Users migrate to Gemini, Copilot
OpenAI cannot keep ChatGPT free unless it moves to an ad-supported revenue model. That is unlikely, or has already happened. Google and Microsoft look at their Gemini and Copilot apps differently. These primarily complement adjacent revenue streams such as search and Microsoft 365. That's why we can offer these apps for free or at a discounted rate. What will happen if OpenAI starts losing ChatGPT users? It will lose its influence in the market and will no longer be able to control the AI stock market trends.
The advent of physical AI
In 2026, the market ecosystem will consolidate and accelerate the adoption of a new field: physical AI. HCL Technologies has a definition suitable for our purposes. “At its core, physical AI integrates cognition and mechanics. It combines AI models, simulations, and real-time perception with robotics and sensors to create systems that learn from experience, adapt to uncertainty, and work seamlessly with humans. …Traditional AI operates in the digital realm, generating predictions or automating software workflows. In contrast, physical AI It operates in dynamic and uncertain real-world environments. It is designed with resilience in mind, and the diversity of objects, environments, and human collaboration allows companies to move beyond an advisory role to an autonomous operational capability.”
This week at CES, NVIDIA announced the availability of a comprehensive physical AI stack that includes robot foundation models, simulation tools, and edge hardware.
Physical AI is not just robots for manufacturing and logistics. It also has applications in areas such as energy, transportation, construction, public safety, and general field service management.
Sovereign AI sparks AI innovation in Europe and India
Sovereign AI refers to the ability of a given country or region to maintain autonomous control over its AI ecosystem, including data governance, AI model development, infrastructure ownership, and regulatory frameworks. This includes securing domestic AI capabilities, protecting jurisdiction, and reducing dependence on foreign AI technologies and services.
This concept emphasizes national autonomy in AI capabilities, often for reasons of data sovereignty, security, resilience, and economic competitiveness.
Corporate adoption of sovereign AI in the EU, UK, and India is expected to increase significantly in 2026. Since these markets have independently addressable markets, there is a good chance that the domestic AI ecosystem will gain traction and displace some US and China-based vendors. That replacement is less likely to occur in GPUs or CPUs, but rather in services and software. AI models that are deep dived in languages other than English come to mind. To some extent, the smaller the market size (e.g. Denmark, Finland, Slovenia), the more likely regional vendors are to prosper. When companies have access to better-tuned models in their own language, results will theoretically become more accurate, accelerating AI initiatives into production.
Apart from this, India presents a huge independent market opportunity comparable to the US and Chinese markets. Sovereign AI regulation there could boost potential players in India's AI ecosystem, giving them an opportunity to innovate and thrive in India, and perhaps even expand to take on AI businesses outside India.
Mark Beccue is a principal analyst at Omdia, covering artificial intelligence.
Omdia is a division of Informa TechTarget. The company's analysts have business relationships with technology vendors.
