Morgan Stanley Projects Major AI Capability Increase by June 2026

Morgan Stanley published a report this month projecting a significant AI capability increase between now and June 2026, driven by an unprecedented accumulation of compute resources at major U.S. AI labs.

Basis of the Projection

The report cites increased GPU capacity, new data centers, and expanded power infrastructure coming online in the first half of 2026. Morgan Stanley's analysis is based on scaling laws — the observed relationship between compute applied to model training and resulting model capability. The bank states these scaling laws continue to hold, with no evidence of diminishing returns.

The report references the concept that applying 10x the compute to model training approximately doubles a model's capability. Morgan Stanley's analysis concludes that the compute conditions for such a step-change are being met in the current period.

Current AI Performance Benchmarks

GPT-5.4, released earlier this month, scored 83% on GDPVal, a benchmark measuring performance on tasks that would previously require human experts. If the next generation of models is trained with 10x the compute and scaling laws hold, the resulting models would perform at approximately twice the current capability level.

Infrastructure and Capital Projections

Morgan Stanley projects a net U.S. power shortfall of 9 to 18 gigawatts through 2028 to run AI infrastructure. The firm estimates nearly $3 trillion in AI-related infrastructure investment will flow through the global economy by 2028, with more than 80% of that spending yet to occur.

These are capital flow and logistics projections from a team that tracks infrastructure investment as a core function.

Potential Implications

Current AI models are already used at scale for legal research, code review, content strategy, and data analysis. A model with significantly greater capability could reduce the human oversight required for these tasks, potentially accelerating automation of knowledge work.

A capability increase could also enable new applications that current models cannot perform reliably, including improved medical diagnosis, large-scale scientific research, and more effective personalized education.

Readiness Gap

Morgan Stanley's report states that most institutions, workers, and regulators are not prepared for a step-change in AI capability.

Regulatory frameworks for AI are behind current capability levels. Companies are deploying AI faster than their legal, HR, and compliance functions can adapt. Workers face uncertainty about which skills to develop as capabilities change rapidly.

Current AI models retain significant limitations, including hallucinations, unreliability on novel problems, and compliance risks in regulated industries. Human judgment remains a factor in most consequential decisions.

Morgan Stanley is advising its institutional clients to monitor the convergence of infrastructure conditions over the next 90 days.