Morgan Stanley: The market is underestimating next year's potential "major AI catalysts," but key uncertainties remain.
```
Author: Long Yue
Source: Hard AI
An AI capability leap driven by computing power may be brewing.
According to Hard AI, Morgan Stanley stated in a recent report that the market may be seriously underestimating a major boon in the field of artificial intelligence likely to arrive in 2026—a “nonlinear” leap in model capabilities driven by exponential growth in computing power.
According to the report written by analysts including Stephen C Byrd, several large U.S. language model (LLM) developers plan to increase the computing power used for training frontier models by around 10 times by the end of 2025. This unprecedented investment in computing power is expected to yield results in the first half of 2026, constituting an “underappreciated catalyst.”
The report quotes Tesla CEO Elon Musk as saying that a 10x increase in computing power could double the “intelligence” level of models. It points out that if the current “scaling laws” continue, the consequences could be “seismic,” broadly impacting asset valuations ranging from AI infrastructure to the global supply chain.
However, this optimistic outlook is not guaranteed. The report stresses that the core uncertainty faced is whether AI development will hit a “scaling wall.” This refers to a disappointing outcome where, despite massive computing power input, the improvement in model capabilities quickly diminishes.
Tenfold Increase in Computing Power May Spur AI Capability Leap
The report believes investors need to prepare for the possibility of a stepwise upgrade in AI capabilities in 2026.
It describes the upcoming scale of computing power: a 1,000-megawatt data center made up of Blackwell GPUs will exceed 5,000 exaFLOPs (500 quintillion floating-point operations per second). By comparison, the U.S. government’s Frontier supercomputer only slightly exceeds 1 exaFLOP. This magnitude of growth in computing power is the core basis for the market's expectation of a nonlinear leap in AI capabilities.
The report states that while many LLM developers generally agree that more computing power leads to better capabilities, skeptics believe that intelligence, creativity, and problem-solving in frontier models may have an upper limit.
The “Scaling Wall” Debate: Key Uncertainty in AI Progress
Despite the exciting prospects, the report also clearly points out a key risk—the existence of a “scaling wall.”
This concept refers to the sharply diminishing, even disappointing, improvements in intelligence, creativity, and problem-solving when computing power input crosses a certain threshold. This is currently the largest uncertainty in the field of AI. Many skeptics believe that simply adding computing power may not continuously bring about leaps in intelligence.
However, the report also mentions some positive signs. A recent research paper, “Demystifying Synthetic Data in LLM Pre-training,” jointly published by teams from Meta, Virginia Tech, and Cerebras Systems, found no foreseeable performance degradation (“model collapse”) within anticipated scales when using synthetic data for large-scale training.
This finding is encouraging, as it suggests models may continue to improve with increased computing power, and the risk of hitting a “scaling wall” may be lower than expected.
In addition, the report lists other key risks, including financing challenges for AI infrastructure, regulatory pressure in places like the EU, power supply bottlenecks facing data centers, and the potential abuse or weaponization of LLMs.
How Will Global Asset Valuation Be Reshaped?
If AI capabilities do achieve a nonlinear leap, how will asset values be reshaped? The report believes investors should begin to assess the multifaceted impact on asset valuation and outlines four core directions:
First is AI infrastructure stocks, especially those that can alleviate data center growth bottlenecks. The report believes that if AI can solve more problems for the global GDP at lower costs and higher performance, the infrastructure supporting this value creation will also appreciate sharply.
Second is the China-U.S. supply chain. The intensifying AI race may spur the U.S. to accelerate “decoupling” in key minerals and other fields.
Third are AI Adopter stocks with pricing power. The report analyzes that AI applications will create about $13 trillion to $16 trillion in market value for the S&P 500 index. However, not all companies will benefit equally. Those with strong pricing power can turn efficiency gains and cost savings from AI into tangible profits, thereby capturing most of the gains.
Finally, in the longer term, hard assets that cannot be cheaply replicated by AI, such as land, energy, and certain infrastructure, may see their relative value rise.Assets with physical scarcity: Such as waterfront real estate, land in specific locations, energy and power assets (especially power plants that can host data centers), transportation infrastructure (airports, ports), minerals, and water resources.Assets with regulatory scarcity: Such as various protected licenses and franchises.Proprietary data and brands: Strong IP portfolios and unique brand images.Unique luxury goods and human experiences: Such as sporting events and musical performances.
This article is from WeChat public account “Hard AI”. For more AI frontier information, click here
Risk Reminder and DisclaimerThe market carries risk, and investments should be made cautiously. This article does not constitute individual investment advice, nor does it take into account individual users’ specific investment objectives, financial situations, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article are suitable for their particular circumstances. Investing accordingly is at one’s own risk. ```