Three AI Trends to Watch in 2026

In 2026, the AI industry’s focus is moving beyond demonstrations of model capability. The bigger questions are whether AI can be deployed at scale, how efficiently it can run, and whether organisations can meet increasingly concrete governance requirements.

1. AI agents move into everyday business workflows

AI is evolving from answering questions and generating content to acting as an agent: using tools and carrying out multi-step tasks. Customer support, software development, operations and knowledge-intensive work are among the most likely areas for adoption.

Large companies are leading the push. Among respondents from organisations with annual revenue above $1 billion, 40% said they were scaling AI agents, up from 27% a year earlier.1 The practical test is no longer simply whether an agent can complete a task, but whether it can do so reliably, securely and with appropriate human oversight.

2. Inference, chips and data centres become the competitive battleground

As commercial AI use grows, running models in production — known as inference — is becoming as strategically important as training them. Companies are paying closer attention to the cost, latency, energy use and reliability of every model call.

Ways to reduce inference costs include model optimisation, advanced chip packaging, custom silicon and co-packaged optics.2 At the same time, physical infrastructure is becoming a key constraint: electricity supply, GPU and specialised-chip availability, and data-centre capacity will shape where AI can expand. PwC projects $31.6 trillion in AI infrastructure capital expenditure through 2050, with power expected to be a decisive factor in investment decisions.3

3. AI governance turns from policy principle into compliance work

For businesses, responsible AI is increasingly an operational requirement rather than a high-level commitment. That means building transparency, data governance, risk assessments, labelling of AI-generated content, human oversight and auditability into products and internal processes.

The EU AI Act became broadly applicable on 2 August 2026, when enforcement began for the rules then in scope, including transparency requirements.45 Some requirements for high-risk systems have longer transition periods, but the Act’s effects will extend beyond Europe to global businesses offering AI products or services in the EU market.

References

• 1 McKinsey, The State of AI: Global Survey 2026.
• 2 McKinsey, Frontiers of compute: The technologies to reduce AI inference costs.
• 3 PwC, Global investment in AI infrastructure to hit US$31.6 trillion through 2050.
• 4 European Commission, AI Act | Shaping Europe’s digital future.
• 5 EU AI Act Service Desk, Timeline for the Implementation of the EU AI Act.