Morgan Stanley warns: AI may experience an "intelligence explosion" by 2026, with power shortages and rising unemployment posing new challenges

March 13 News, Morgan Stanley’s latest research report indicates that artificial intelligence technology may achieve a critical breakthrough in the first half of 2026, while global infrastructure and economic systems are not yet fully prepared to handle this change. The report suggests that major AI laboratories in the United States are rapidly accumulating unprecedented computing power, which will drive large models into a new development stage.

The report specifically mentions Elon Musk’s previous comments in an interview, stating that if the computing power used to train large language models increases by about 10 times, the overall intelligence level of the models could nearly double. The industry generally still believes this “scaling law” remains valid, meaning that as computing power continues to grow, AI capabilities may increase exponentially.

Meanwhile, industry progress has already begun to surpass market expectations. Morgan Stanley notes that OpenAI’s recent release of the GPT-5.4 “Thinking” model achieved an 83.0% score on the GDPVal benchmark test, approaching or even surpassing the level of some human experts in several economically valuable tasks. The firm believes this capability leap could accelerate in the coming years.

However, the growth in computing power also brings significant infrastructure pressure. Morgan Stanley’s “Smart Factory” model shows that by 2028, the U.S. power grid could face a net shortfall of 9 to 18 gigawatts, accounting for 12% to 25% of the electricity needed for AI-related systems. As AI data center construction accelerates, energy supply will become a critical bottleneck.

To address power shortages, some tech companies and infrastructure operators are exploring alternatives. For example, converting Bitcoin mining facilities into high-performance computing centers, deploying natural gas generators, and fuel cell systems to support the continuous operation of AI data centers. The industry has also proposed a “15-15-15” model, which includes 15-year data center leases, approximately 15% return on investment, and about $15 per watt of net electricity value.

The report also points out that AI could have a profound impact on the labor market. Since AI tools can replicate certain human workflows at very low costs, some companies have already improved efficiency and reduced staff through automation and AI systems.

OpenAI CEO Sam Altman has envisioned a future where small companies of 1 to 5 people, powered by AI, can accomplish business scales that previously only large corporations could achieve. Additionally, Jimmy Ba, co-founder of xAI, believes that self-improving AI systems may emerge by early 2027, pushing technology into a new development phase.

Morgan Stanley concludes that as computing power continues to increase, AI could become a key driver of future economic structural changes, with power, energy, and data center infrastructure becoming the core resources in the next phase of global competition. (Fortune)