AI in 2026: What it can do and where it will realistically deliver value

After over two years of hype, experimentation and big expectations, artificial intelligence (AI) enters 2026 at a more sobering —and consequential— note.

 

The question is no longer what AI can do, but where will it realistically deliver value.

 

While the hype around artificial general intelligence (AGI) has waned a little, AI agents are now reshaping how work gets done.

 

Legal battles over data are inching toward negotiated frameworks. The fight for AI supremacy is shifting from models to chips and infrastructure. And governments, including that of India, are asserting greater control over where data, compute and intelligence reside.

 

From agents and chips to sovereignty and security, here are the AI developments that are likely to define 2026— and those that will remain over the horizon for time to come.

 

AGI, still a long way ahead

 

AGI or artificial general intelligence is an AI system that can apply knowledge to solve any problem or complete an intellectual task as humans do.

 

Till last year, many wondered and also bet that AGI will be available by 2026. Some of the people who predicted this include Elon Musk, Anthropic’s Dario Amodei among others. The reason for the optimism was the speed with which AI systems were progressing.

 

However, closer to the end of 2025 that narrative changed. AI heavyweights such as Andrej Karpathy and Demis Hassabis believe that this could be at least a decade away.

 

Karpathy in a podcast with Dwarkesh Patel said, “We have some very early agents that are extremely impressive and that I use daily — Claude and Codex and so on— but I still feel there’s so much work to be done. My reaction is we’ll be working with these things for a decade.” 

 

Google DeepMind’s Demis Hassabis, in a media briefing, has also said that AGI will start to emerge in the next five to 10 years.

 

As Karpathy in an X post list out the reasons for such a system to come, “…there is still a lot of work remaining (grunt work, integration work, sensors and actuators to the physical world, societal work, safety and security work -- jailbreaks, poisoning, etc.) and also research to get done before we have an entity that you’d prefer to hire over a person for an arbitrary job in the world. I think that overall, 10 years should otherwise be a very bullish timeline for AGI, it’s only in contrast to present hype that it doesn’t feel that way.”

 

Assuming AGI is a decade away, 2026 clearly will focus on execution.

 

…but agents are here:

 

While AGI is still in the works, AI agents have quietly entered the workforce.

 

Aparna Chennapragada, Microsoft’s chief product officer for AI experiences, sees 2026 as a new era for alliances between technology and people. If recent years were about AI answering questions and reasoning through problems, the next wave will be about true collaboration, Chennapragada says.

 

“The future isn’t about replacing humans,” she says. “It’s about amplifying them.” Her views are part of Microsoft blog ‘What’s next in AI: 7 trends to watch in 2026’.

 

Chennapragada envisions a workplace where a three-person team can launch a global campaign in days, with AI handling data crunching, content generation and personalisation while humans steer strategy and creativity. She predicts organisations that design for people to learn and work with AI “will get the best of both worlds,” helping teams tackle bigger creative challenges and deliver results faster.

 

In India, we are seeing more IT services firms adopt Copilot’s and agentic AI across workflows.

 

Satya Nadella, chairman and CEO Microsoft on his recent India visit, said that IT firms TCS, Cognizant, Infosys and Wipro will deploy 50,000 Microsoft Copilot licenses, collectively surpassing 200,000 licenses, setting a new benchmark for enterprise scale AI adoption.

 

LLM majors & content creators to call a truce?

 

Even as content creators such as music companies and newsrooms across the globe have filed a host of cases against makers of large language and artificial intelligence models for the unfair use of their content to train these models, 2026 could see some of these cases settled at the negotiating table rather than in courtrooms.

 

In India, the Department for Promotion of Industry and Internal Trade (DPIIT) kick-started this process by releasing a working paper that proposed an arrangement under which AI companies can make certain payments to content creators.

 

In its working paper, the DPIIT has suggested that technology companies that use content creators’ data to train their respective AI and LLMs pay royalties to these content creators at a rate set by the government or the court. The new framework has also suggested that the payment for this blanket licence be collected by a unified industry body, which will then be responsible for distributing the proceeds to content creators.

 

Elsewhere in the world, countries such as Japan and Singapore, and even the European Union, have introduced laws for copyrighted content, but with exceptions for text and data mining.

 

On the other hand, US copyright law grants the owner of a copyright exclusive rights in relation to reproduction, preparation of derivative works, distribution of copies or phono records of copyrighted works, public performance, and public display.

 

With the introduction of AI models that have been trained on massive datasets of content creators, the courts in the US have also relied on whether the AI systems developed were “highly transformative” in nature.

 

In countries such as Japan, which have granted a text and data mining exception, a crucial factor is whether AI training on copyrighted work will “impede the potential sales channels of the copyrighted work in the future.”

 

On the horizon: AI chip wars

 

Nvidia may be the undisputed world leader when it comes to making the AI chips that power the large language models (LLM) for training and inferencing, but a serious competition is emerging. 2026 may just be the year when hyper-scalers, Google Cloud, Amazon Web Services (AWS) and Microsoft Azure, may get into building their own chips to threaten the juggernaut of Jensen Huang’s company.

 

Early trends are already on the horizon. Late November, when reports emerged saying Meta Platforms may use Google’s in-house Tensor Processing Units (TPUs) for its upcoming data centres, it triggered a sharp reaction, knocking Nvidia’s stock down nearly 3 per cent and reigniting the debate over whether the long-standing GPU leader is finally facing credible competition.

 

Nvidia was quick to respond, publicly welcoming Google’s progress while asserting that its own chips remain “a generation ahead of the industry.”

 

TPUs are specialised processors designed by Google for AI applications. Google’s TPUs are ASICs built for specific tasks. These chips are also important when it comes to inferencing. Besides using TPUs for Gemini, Google has proven the capability of these chips.

 

OpenAI, in fact, has already been using Google’s TPUs to power its products though it remains one of the largest clients of Nvidia. This year may give some glimpse to analysts and investors whether the battle for AI supremacy shifts from building LLMs to building the chips that support them.

 

Sovereign AI

 

According to Stanford University Human-Centered Artificial Intelligence (HAI) year-end predictions, AI sovereignty will gain huge steam this year. AI infrastructure is fast becoming a strategic asset, on par with energy and defence.

 

“AI sovereignty will gain huge steam this year as countries show their independence from AI providers and the US political system,” says James Landay Stanford HAI Denning Co-Director.

 

The big tech companies are pouring billions of dollars profusely across countries to set up data centres and this year may see such investments touching new records, unless of course the AI sentiment catches a cold.

 

Morgan Stanley estimates global data centre capacity will need to grow six-fold by 2035 to meet the growing demand of cloud computing and AI. That is equivalent to $3 trillion investment in infrastructure between 2025 and 2028.

 

Asia’s third largest economy now boasts of a load capacity of 1.5-1.7 GW in 2025, which is expected to expand five-fold to 8 GW by 2030, according to research firm Jefferies. Almost 70 per cent of it is now being driven by hyper-scalers.

 

The huge demand, the JLL report says, will be driven by growing adoption of AI technologies as AI servers consume five to six times more power than traditional setups and require advanced liquid cooling systems. In addition, regulatory developments such as the Digital Personal Data Protection (DPDP) Act, 2023, and the RBI’s data localisation guidelines are driving enterprises to host and process data within India.

 

But it’s not just India.

 

Google, Meta, Microsoft and AWS are investing across the world to set up data centres in regions like UAE, South Korea and Europe. South Korea is working with the UAE on the US-backed Stargate project to build a new AI data campus in the Gulf country as both nations aim to become regional AI hubs.

 

South Korea’s flagship companies like Samsung Electronics and SK Hynix have signed initial agreements to supply memory chips for OpenAI’s Stargate data centres.

 

The first phase of that project will be the 1 GW Stargate UAE project, built by state-backed UAE firm G42 in partnership with OpenAI, Oracle, Nvidia and Cisco Systems, besides Japan’s SoftBank. At the same time, concerns will likely mount on the use of electricity and water, two vital ingredients to run such massive centres, and if such indiscriminate usage is justified in many countries where water is scarce.

 

Cybersecurity

 

As AI becomes more sophisticated, so will be the attacks to the systems which only exacerbates the risks going ahead.

 

For India, cybersecurity needs to be top priority for the government and organisations. India saw more than 265 million cyberattacks in 2025, according to a report by Seqrite, the enterprise arm of security solutions and services firm Quick Heal; Trojans and file infectors accounted for 70 per cent of all attacks.

 

According to the ‘State of Cyber Security in India, 2025’ report by Check Point Software Technologies, Indian organisations faced more than 2,000 cyberattacks per week per organisation. Across industries, Indian organisations faced 2,011 cyberattacks per week in 2025, significantly higher than the global average.

 

Data from the National Cyber Crime Reporting Portal (NCRP) highlights that cybersecurity incidents rose from 10,29,000 in 2022 to 22,68,000 in 2024. This reflects the growing scale and complexity of digital threats in India. 

 

Cybersecurity will continue to remain a top priority for enterprises as they integrate more digital tools into their operations.

 

“Threat actors will predominantly shift to launch malicious campaigns through vibe coding -- exacerbating the speed and delivery of execution. They will increasingly use AI as a teacher or trainer to help them do reconnaissance, but not because they don’t know how to launch a low-level attack. This reconnaissance will enable them to gather critical information about a target, and create the specialised tools needed for scanning and exploitation,” says Grant Bourzikas, Cloudflare’s chief security officer.

 

While 2025 was filled with AI’s contributions to basic malicious activities like social engineering, deepfakes, business email compromise and more, 2026 is expected to be the year of real attacks.  2025 roundup on AI startups in India