According to a Bain report, global demand for AI computing power is growing at an annual rate of 4.5 times, far exceeding the efficiency improvement pace of Moore’s Law, which doubles every two years. To meet this demand, an additional 200 gigawatts of computing capacity will be required globally by 2030. For companies to truly profit from advanced AI agents, they must modernize their outdated core IT infrastructure.

A 2025 technology report from Bain & Company has unveiled a harsh yet clear reality behind the AI revolution: the growth in demand for computing power is spiraling out of control, while corporate infrastructure has fallen significantly behind.

On October 20, Barclays analyst Raimo Lenschow analyzed the report, concluding that it highlights two irreversible trends: the growth in demand for computing power driven by AI is far outpacing Moore’s Law, and companies seeking to capitalize on this trend must modernize their outdated core IT infrastructures.

Computing power demand far exceeds Moore’s Law

The report’s most striking insight is that AI-driven demand for computing power is growing at an annual rate of 4.5 times, more than double the pace of Moore’s Law (which estimates that chip efficiency doubles approximately every two years).

This indicates that technological innovation can no longer keep pace with demand. This imbalance will trigger massive investments in data center infrastructure.

According to Bain’s analysis, global demand for computing power could reach 200 gigawatts (GW) by 2030, with the United States alone accounting for 100 GW. Building such capacity would require approximately USD 500 billion in capital expenditure annually, corresponding to USD 2 trillion in annual revenue.

Barclays noted that while government subsidies are unlikely to bridge such a vast funding gap, the private sector is taking more aggressive actions. Moves like OpenAI signing trillion-dollar-level computing power contracts suggest that future investment levels will only continue to rise.

This forecast is being validated by market actions. According to Reuters, an investment group backed by Blackrock and NVIDIA recently acquired Aligned Data Centers, a data center operator with nearly 80 sites, for $400 billion. Meanwhile, Morgan Stanley estimates that tech giants including Alphabet, Amazon, and Meta will spend $400 billion this year on AI infrastructure.

Enterprise Modernization: The Only Path to Harnessing AI

Beyond the race for computing power, another core focus of the report is the urgent need for enterprise IT modernization.

As AI evolves from simple information retrieval (Level 1 agents) to executing complex workflows (Level 2 and Level 3 agents), enterprises are finding themselves unprepared. Corporate clients remain cautious about fully relying on autonomous AI agents without human oversight.

To truly harness AI, particularly the more advanced ‘Agentic AI,’ enterprises must undergo comprehensive IT modernization. Agentic AI can think, reason, and act based on user instructions, far surpassing simple information-retrieval-based chatbots.

Bain & Company believes the solution lies in transforming the architecture surrounding these agents. The report recommends, “Companies must modernize their core IT infrastructure and embed observability features to provide a suitable ecosystem for these agents.”

For investors, this means that in the next phase of AI, the battleground will shift from pure computing hardware to software and service providers that can help enterprises navigate this painful but necessary transformation. This is not merely a technological upgrade but a battle for survival.

This article is from the WeChat public account “Hard AI.” For more cutting-edge AI news, please visit here.

Editor/Lee