Where will the main AI trend be next year? Morgan Stanley: The theme of "removing bottlenecks" will replace chips; optimistic about energy infrastructure.

After Nvidia has indisputably dominated the compute market, smart money on Wall Street is starting to look for the next big opportunity.

According to Chase the Trend Trading Desk, Morgan Stanley pointed out in its latest report that as AI compute demand grows nonlinearly, the market logic is undergoing a fundamental shift: investors’ focus in 2026 needs to shift from “chips” to “de-bottlenecking.”

Morgan Stanley again raised its data center power demand forecasts based on the latest sales growth expectations for Nvidia. From 2025 to 2028, the cumulative power gap for U.S. data centers is expected to reach 47 gigawatts (GW), higher than the previous forecast of 44 GW, equivalent to the total electricity consumption of nine Miami regions or fifteen Philadelphia regions.

The report further points out that even after subtracting various “rapid power supply” solutions, U.S. data centers will still face a 10-20% power shortfall, about 6-16 GW, with the gap becoming most severe in 2027.

Power Shortfall Surges: From Chip Crisis to Grid Crisis

Based on the latest chip sales data, Morgan Stanley again raised its power demand forecast, signaling a more severe energy crunch. The report shows that by 2028, the U.S. will face a power gap revised from the previous 44 GW to 47 GW. This supply-demand imbalance indicates that relying on the existing power grid alone is no longer realistic.

“Given the nonlinear advances in AI and the proliferation of use cases, we believe investors will shift their focus to alleviating ‘smart bottlenecks’ in 2026. Key ‘smart bottlenecks’ include: power, political support, labor, and various data center equipment.”

Pragmatic Solutions and Persistent Gaps

Faced with the long wait for grid interconnection, “Time-to-Power” has become a core metric for assessing asset value. Morgan Stanley listed four solutions to bypass power grid congestion: natural gas turbines, Bloom Energy (BE) fuel cells, site-adapted nuclear power plants, and conversion of bitcoin mining sites. Despite these workarounds, analysts remain cautious, believing that the gap cannot be completely filled.

“Even taking into account all kinds of ‘Time-to-Power’ solutions, the net U.S. power gap will still be 10-20% of the amount needed for data center construction (about 6-16 GW). In our view, this gap will be most severe in 2027, when chip demand grows rapidly while most turbine solutions have not yet been put into operation.”

Bitcoin Miners’ “Arbitrage” Moment

Among all energy solutions, cryptocurrency mining companies, thanks to existing power access permits, are evolving into a “fast lane” in the field of AI infrastructure.

Morgan Stanley highlighted two models: “New Neocloud” players like IREN directly leasing GPUs, and “REIT endgame” models like APLD, where shells are built and leased out to hyperscale enterprises. This asset transformation is leading to a revaluation of mining companies.

“We continue to believe that bitcoin sites provide AI participants with the fastest time-to-power and lowest execution risk . . . Given the recent weakness in many AI infrastructure stocks, we recommend focusing on the most promising ‘de-bottlenecking’ players.”

Nonlinear Growth in AI Capabilities Drives Surging Demand

What underpins this huge energy consumption is the exponential leap in AI capabilities. The report notes that continuous nonlinear improvement in AI capability and the emergence of more compute-intensive scenarios are the fundamental drivers behind repeated upward revisions of data center power demand. Morgan Stanley defines this trend as the “diffusion of intelligence.”

Specific evidence includes: Morgan Stanley analyst Brian Nowak forecasts Agentic Commerce to reach $190 billion to $385 billion in GMV by 2030, accounting for 10-20% of US e-commerce. Currently, 45% of US respondents use ChatGPT, 32% use Gemini, and 36% of ChatGPT users have made purchases via the platform in the past month.

In enterprise AI adoption, 24% of AI-adopting enterprises reported quantifiable returns in Q3 2025, up from 15% in Q3 2024. Morgan Stanley expects AI-driven efficiency will contribute 30 and 50 basis points of incremental net profit margin to S&P 500 constituents in 2026 and 2027, respectively.

More importantly, in the most challenging general intelligence test ARC-AGI-2, the latest cutting-edge model Gemini 3 Deep Think scored about 45%, while months ago, models of this category scored only 10-20%. Considering that the human average is 60%, and cutting-edge models in 2026 will train with roughly 10x more compute, the industry expects AI capability may surpass human level in multiple complex reasoning tests.

~~~~~~~~~~~~~~~~~~~~~~~~

The amazing content above is from Chase the Trend Trading Desk.

For more detailed interpretations, including real-time commentary and frontline research, please join [Chase the Trend Trading Desk ▪ Annual Membership]

Risk Warning and DisclaimerThe market has risks, investments should be made cautiously. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial conditions, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their particular circumstances. Investment decisions made based on this are at your own risk. ```