Capability and use

For several years the visible question in AI was capability: how far models could go, and how quickly. That question has not gone away. Industry produced more than 90% of notable frontier models in 2025, and on a widely used coding benchmark, performance moved from 60% to near the human baseline in a single year.

Use has spread with it. Stanford’s 2026 AI Index puts organisational AI adoption at 88% of surveyed organisations in 2025, up from 78% in 2024. Generative AI is used in at least one business function at 70% of organisations. Private investment in the United States reached $285.9 billion in 2025.

Adoption at that level is not the same as reliable use. The same Index finds that AI agent deployment remained in the single digits across nearly all business functions. A system can be available, and even impressive in a demonstration, while still being difficult to place inside an existing workflow, data environment or approval process.

What has to work in practice

The practical tests are ordinary. The system has to behave the same way on the next hundred tasks. It has to see the information the work depends on. Its output has to be checkable. The cost of running it has to fit the job. And the organisation has to be able to change the surrounding process without breaking it.

Those tests sit next to a physical constraint. The International Energy Agency estimates that capital expenditure by five large technology companies exceeded $400 billion in 2025, and expects that spending to rise by a further 75% in 2026. Electricity use by data centres is projected to rise from about 485 terawatt-hours in 2025 to about 950 terawatt-hours in 2030. Software can be rolled out in weeks. Power, chips and buildings cannot.

Through 2027, the useful measure is how much of the new capability becomes dependable in ordinary operations. Capability will keep moving. The constraint is implementation.

Sources

Stanford Institute for Human-Centered Artificial Intelligence, AI Index Report 2026, economy and research chapters. Organisational adoption 88% in 2025 (78% in 2024). Generative AI in at least one function at 70%. Agent deployment in the single digits across nearly all functions. US private AI investment $285.9 billion in 2025. Industry produced over 90% of notable frontier models in 2025. SWE-bench Verified rose from 60% to near 100% in a year.
https://hai.stanford.edu/ai-index/2026-ai-index-report

IEA, Key Questions on Energy and AI, executive summary. Capital expenditure of five technology companies exceeded $400 billion in 2025 and is expected to rise another 75% in 2026. Data-centre electricity consumption projected from 485 TWh in 2025 to 950 TWh in 2030.
https://www.iea.org/reports/key-questions-on-energy-and-ai/executive-summary