AI Industry Confronts Budget Overruns and Data‑Center Heat Issues

Spending on artificial intelligence has surged, with companies investing more in the past two years than on any technology in a generation. Yet roughly four‑in‑five organisations report no measurable business impact, and about 85% exceed their AI budgets due to unforeseen token consumption. Analysts predict many AI projects will be abandoned before reaching production, pointing to an “AI velocity gap” where rapid model advances outpace the adaptation of skills, governance and workflows. In the United States, 93% of jobs are already exposed to AI‑driven change, creating a trillion‑dollar opportunity to close this gap through both efficiency gains in existing work and the creation of new, agentic AI products and services.

At the same time, AI chips generate massive heat, creating a parallel challenge for data‑center operators. Nvidia’s next‑generation Vera Rubin chip consumes 120‑130 kilowatts per rack—equivalent to the annual electricity use of about 100 U.S. homes—and future versions may use five times more power. This power density demands upgraded electrical delivery and advanced cooling solutions, opening investment opportunities for companies that can mitigate the heat problem.