Where is the next core bottleneck for AI hardware?

```

The AI wave is reshaping the global workforce structure, but for the technology hardware and semiconductor industries, this transformation brings net gains rather than shocks—the sustained expansion of computing power demand is pushing structural constraints on the supply side to the forefront.

According to ZF Trading Desk, citing Morgan Stanley's latest AI Adoption and Future Work Survey Report, the technology hardware and semiconductor industries recorded net productivity improvements of approximately 8.6% to 8.2% over the past 12 months, while also facing net job losses of about 5% to 8%. However, Morgan Stanley analysts clearly point out that the sector is a "net beneficiary" of the AI wave—not only reducing costs through internal operational efficiency improvements, but also directly benefiting from surging orders for chips and related equipment due to the booming demand for external AI infrastructure.

Morgan Stanley analysts believe that continued improvements in AI model efficiency will further spur companies to increase AI infrastructure investments, thereby continuously supporting semiconductor demand. The real bottleneck in the next phase of AI hardware will shift from chip design to energy, electricity, data center capacity, and manufacturing capacity; these supply constraints may become key limitations to AI infrastructure expansion.

Hardware and Semiconductors: The "Picks and Shovels" of AI

Morgan Stanley's report characterizes the technology hardware and semiconductor industries as the core suppliers providing "picks and shovels" in the AI ecosystem.

Survey data shows net job losses of about 5% (technology hardware) and 8% (semiconductors) over the past 12 months, but these figures do not represent harm to the industry—layoffs are concentrated in early stages, and the overall impact is positive.

From specific application scenarios, European semiconductor companies have quite diversified AI adoption. ASML uses AI to enhance precision and efficiency in computational lithography solutions; STMicroelectronics applies it for defect classification and yield optimization. Additionally, 23% of technology hardware companies surveyed in the UK and Germany reported they have deployed digital twin technology in manufacturing facilities, with suppliers including EDA firms Cadence and Synopsys, aiming to bridge simulation and reality gaps.

In terms of workforce restructuring, European semiconductor companies are still in the early stages of layoffs.

ASML, Infineon, and Aixtron all announced layoff plans in 2026, but management did not directly attribute these to AI. The clearer case is ams OSRAM, which announced in February 2026 plans to cut more than 2,000 positions in the next three years, explicitly naming AI automation in European business as one of the reasons for layoffs, while shifting some workforce to Asia.

AI Adoption Accelerates Expansion in Chip Orders

The core logic of Morgan Stanley's report is: the more pronounced the efficiency benefits from AI, the stronger the willingness of companies to invest in AI infrastructure—which in turn creates a positive cycle for semiconductor demand.

Survey data shows that the average net productivity improvement across the industry is 9.6%, with productivity gains especially prominent in high-intensity AI adoption sectors like banking and software. Continued investment from these industries will directly convert into rigid demand for computing power.

From a capital expenditure perspective, Morgan Stanley cited data showing JPMorgan’s technology budget for 2026 is nearly $20 billion, with around $2.3 billion (about 25% of total investment budget) directly allocated to AI investments. Such enterprise-level AI spending ultimately flows down the supply chain to chip design, wafer manufacturing, and semiconductor equipment segments.

Morgan Stanley analysts further point out in the report that as AI evolves from generative models to agentic AI, CPUs, semiconductor equipment, and memory will become key beneficiaries. Large-scale deployment of agentic AI demands higher computing density, inference efficiency, and memory bandwidth, meaning existing hardware architectures face new rounds of iterative pressure and foreshadow the emergence of the next supply bottleneck.

Energy, Computing Power, and Supply Chain Constraints

Despite a bright outlook for demand, Morgan Stanley’s survey also reveals multiple structural obstacles constraining AI hardware expansion.

Among software industry respondents’ major challenges over the next 12 months, "Infrastructure and Energy Constraints (computing capacity, electricity availability)" are clearly mentioned. This issue also applies to hardware suppliers providing underlying support for AI workloads.

From the supply chain perspective, the survey shows "legacy system integration" and "data readiness" are common pain points across industries for AI deployment, posing extra challenges for hardware companies needing to embed AI into existing manufacturing processes. At the same time, "AI skill shortages" rank among the top three barriers in multiple industries—the semiconductor sector’s reliance on specialized engineering talent puts it under greater pressure in talent competition.

Observations from the Australian software industry offer another perspective: as AI adoption moves toward broader deployment, demand for data centers and computing infrastructure will grow significantly, with capital density and electricity intensity rising in tandem. This assessment provides direct reference value for global semiconductor and hardware investors—the next core bottleneck may be less about chip design itself and more about energy supply and manufacturing capacity needed for large-scale AI inference support.

For investors, Morgan Stanley’s survey provides a clear framework for industry positioning. Technology hardware and semiconductors, as core suppliers of AI infrastructure, benefit from persistent demand caused by accelerating cross-industry AI adoption, and their fundamental logic is relatively solid.

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

The above content is from ZF Trading Desk.

For more detailed interpretation, including real-time analysis and frontline research, please join [ZF Trading Desk ▪ Annual Membership]

Risk Warning and DisclaimerThe market involves risks, investment must be cautious. This article does not constitute personal investment advice, nor does it consider specific investment objectives, financial situations, or needs of individual users. Users should decide whether the opinions, viewpoints, or conclusions in this article fit their particular circumstances. Investment is at your own risk. ```