Key Takeaways:
Morgan Stanley identifies energy and data center capacity, not chip design, as the next binding constraint on AI hardware expansion.
The next bottleneck for AI hardware is shifting from chip design to energy supply and manufacturing capacity, according to Morgan Stanley, as the industry's 8.6% productivity gains drive further infrastructure investment.
"AI model efficiency improvements will further incentivize companies to increase AI infrastructure spending, providing sustained support for semiconductor demand," the Morgan Stanley analysts wrote in the report.
The survey found tech hardware and semiconductor companies recorded net productivity gains of 8.6% and 8.2%, respectively, over the past 12 months, while facing net job losses of about 5% to 8%. Key constraints cited include infrastructure and energy limitations, data center capacity, legacy system integration, and a shortage of AI-skilled talent.
The shift has direct implications for investors. As AI evolves from generative models to agentic AI, demand for CPUs, semiconductor equipment, and memory will intensify, putting pressure on foundry capacity at TSMC and advanced packaging supply chains. JPMorgan's 2026 technology budget of about $200 billion includes roughly $23 billion — or 25% — allocated to AI, a spending trajectory that Morgan Stanley said will compound the strain on physical infrastructure.
European semiconductor companies are already adjusting. ASML, Infineon, and Aixtron all announced layoff plans in 2026, though management did not directly attribute the cuts to AI. A clearer case is ams OSRAM, which in February 2026 said it would eliminate more than 2,000 positions over three years, explicitly citing AI automation in its European operations as a reason while shifting some labor to Asia.
The report's core thesis is that AI efficiency gains create a positive investment cycle: the more companies improve productivity through AI, the more they spend on AI infrastructure, which in turn drives chip orders. Across all industries surveyed, the average net productivity gain was 9.6%, with banks and software companies among the top adopters.
Supply chain constraints tighten as 2nm ramp begins
TSMC, the dominant foundry for AI chips, is already signaling the strain. The company reported Q2 2026 revenue of $40.2 billion, up 33.7% year over year, and raised its full-year capital spending to a range of $60 billion to $64 billion — a 15% increase from its prior target of $52 billion to $56 billion. About 70% to 80% of that spending is allocated to advanced process nodes including 2nm and 3nm, while 10% to 20% goes to advanced packaging such as CoWoS, which remains a bottleneck for AI chip supply.
The 2nm node (which packs more transistors per square millimeter, improving performance per watt) is expected to dilute gross margins by 3 to 4 percentage points during its initial production ramp, according to TSMC. The company also announced a further $100 billion multiyear expansion of its Phoenix, Arizona facility, bringing its total US investment to $265 billion.
Investor implications: who wins, who loses
For investors, the Morgan Stanley report reinforces a bullish long-term outlook for semiconductor and AI infrastructure stocks by framing energy, power, and capacity constraints as structural drivers of sustained capital spending. Companies that provide the physical layer of AI — data center operators, energy infrastructure firms, and semiconductor equipment makers such as ASML and Applied Materials — stand to benefit as the bottleneck shifts from design to production.
Conversely, chip designers that rely on easily scalable advantages may face margin pressure as rising energy and manufacturing costs eat into profitability. TSMC shares, which fell about 5% to below $400 after the earnings call as concerns over capital spending intensity weighed on sentiment, trade at roughly 18x forward earnings — a discount to the broader semiconductor index, according to analysts at Morgan Stanley and Barclays, who view the higher spending as a proxy for long-term revenue visibility rather than structural inefficiency.
This article is for informational purposes only and does not constitute investment advice.
