Morgan Stanley Assesses AI Market Reaching Turning Point: Shifting From Computing Power Construction to Industry Diffusion, Barbell Strategy Becomes Main Allocation Theme

According to news from Zhitong Finance APP, Morgan Stanley has released a thematic research report titled "Decoding the Pace of the AI Revolution: From Computing Power Construction to Industrial Diffusion". Based on its sixth round of an AI stock mapping database covering approximately 3,600 stocks worldwide, the report puts forward a core viewpoint: AI investment has moved past the stage of simply betting on upstream computing power enablers, and has entered a new era of 'barbell' dual-track deployment. Investors should both seize structural opportunities brought by computing power bottlenecks and increase the allocation weight to AI application enterprises.

Morgan Stanley states that its core research theme for 2026 is technology diffusion. In this round of the transformation industrial cycle, a barbell-shaped pattern of excess returns is expected to emerge, with key infrastructure enabling companies running in tandem with early-stage software vendors and AI adoption firms.

The usual technological cycle unfolds with semiconductors achieving excess returns first, followed by the infrastructure segment, and then software and services. The bank believes that the investment logic of the AI industry is entering a new stage: shifting from highly concentrated transactions in enabling firms to a broader barbell pool of opportunities, covering both selected enabling enterprises and emerging AI adoption firms.

Since the start of this AI cycle, the semiconductor and infrastructure segments have already recorded 500-700 points of excess returns relative to the S&P 500 Index. We believe that now is the time to increase exposure to early-stage software enablers and AI application adopters.

However, at the same time, the wave of large-scale AI capital expenditures, combined with critical bottlenecks such as power supply, will lengthen the upward window for certain high-quality targets. Therefore, this computer cycle will not simply repeat the straightforward linear sector rotation leadership of previous cycles.

Morgan Stanley notes that this AI cycle is unique, with labor, electricity, and regulatory policy forming the three core constraints. According to their analysis, there will be a global power supply gap of 57GW for data centers during 2026-2028. Adding in local government approvals, labor shortages and other practical obstacles, the supply expansion of computing power is being significantly constrained, so the shortage of computing power will persist for many years, and there won't be a simple and clear sector rotation as in history. Capital expenditure growth by hyperscale cloud providers is projected to slow from 93% in 2026 to 14% by 2028. While hardware capital expenditure is being increasingly priced in as having peaked, certain bottleneck tracks will see continued prosperity.

Data shows that AI enablers (chips, computing power, hardware and other upstream suppliers) have doubled their expected EPS over the past two years and continue to maintain strong profit momentum. Meanwhile, AI adopters—companies in various industries leveraging AI to reduce costs and increase efficiency—are expected to see cumulative EPS growth of about 70% over the next 12 months, with profit inflection points already emerging.

Market consensus expects that high-AI-impact adopters will see their EBIT margins expand by 460 basis points in 2025‑2026, nearly double that of the MSCI ACWI index. However, the market has yet to fully price in the long-term revenue gains from AI-driven productivity, meaning there is still upside to long-term earnings forecasts. In terms of valuation, future PE ratios for high-AI-impact adopters have fallen to 18x, compared to 22x for enablers, greatly improving the risk-return ratio after valuation adjustment.

The bank’s proposed barbell investment strategy consists of two ends. On one end, continue to hold onto bottleneck assets: prioritizing investment in power-related sectors, including operational data center service providers, energy storage, grid equipment, and renewable energy; meanwhile, selectively investing in semiconductors and infrastructure by focusing on second-tier supply chain bottlenecks beyond electricity.

On the other end, expand new allocations: increasing positions in early-stage software enablers, with an order of priority for infrastructure software and cybersecurity, then selected application software; invest in AI adoption companies that have already achieved quantifiable business benefits, focusing on two key metrics: the substantial impact of AI on the company’s business, and the company’s pricing power. Only companies capable of retaining AI-driven cost reductions can genuinely translate these into profit realization. IT services, as a late-cycle beneficiary, are not considered a key allocation at this stage.

From a global perspective, the development path of AI differs significantly across markets. In North America, AI dividends are expanding from hardware to software and physical industries; in Asia-Pacific, benefits are still mostly concentrated in the upstream supply chain, but more and more listed companies are disclosing AI-driven revenue increases in their financial reports; in Europe, AI opportunities are mainly focused on application implementation in traditional industries.

The report particularly emphasizes that investing should not be about simply chasing the "AI concept"—the substantive impact of AI on a company’s investment logic is the key to outperforming. Companies with AI as the core of their business logic enjoy far greater excess returns than those with only secondary exposure; conversely, companies whose core business is threatened by AI will significantly underperform even compared to peers facing moderate impact.

At the same time, there are counter risks in the market. If computing power demand continues to exceed expectations and labor and power bottlenecks persist, the prosperity of semiconductors and infrastructure sectors could last much longer than the market currently anticipates—historically, Wall Street has repeatedly underestimated the growth potential of the tech industry. Overall, the AI boom has evolved from simple “computing power stacking” speculation to a new stage of validating commercial ROI. A dual focus on upstream bottleneck assets and application-based companies realizing tangible benefits is now a more suitable allocation strategy.