The Golden Age of AI Chips Meets an Old Industry Reality
The semiconductor industry has rarely been out of the spotlight, but the arrival of generative AI has ignited a historic surge in demand. Graphics processors, custom AI accelerators, and memory chips have become the lifeblood of data centers powering ChatGPT, autonomous systems, and enterprise AI deployments. Valuations of major players like Nvidia, AMD, and TSMC have reached staggering levels, with Wall Street racing to price in an endless AI revolution.
Yet, a warning from Morningstar suggests investors may be underestimating the oldest and most reliable law of semiconductors: the boom-bust cycle. Despite AI demand acting as a turbocharger, the chip industry’s cyclical nature remains intact — and history indicates the next downturn could arrive sooner than expected.
Morningstar’s Red Flag: Cycles Still Dominate Chips
Morningstar analysts note that the semiconductor sector typically runs in four-year cycles, dictated by supply and demand imbalances. The firm argues that while AI has stretched the current upcycle, the fundamentals haven’t changed.
- Billings Growth is Slowing: Semiconductor billings, a leading indicator for sector health, are already tapering off, suggesting the current expansion may be peaking.
- AI Spending Peak in 2025: The report projects AI-related capital expenditure will peak by mid-2025, with risks of softening demand in 2026.
- Macro Risks Ahead: Global economic uncertainty, weak consumer electronics demand, and high interest rates could amplify the downturn when AI momentum slows.
This analysis echoes previous cycles: after each surge in demand — whether for PCs in the 1990s, smartphones in the 2010s, or cloud in the late 2010s — oversupply eventually triggered sharp corrections.
AI as a Double-Edged Sword
The AI boom is undeniably unique in scale. Since late 2022, hyperscalers like Microsoft, Amazon, Meta, and Google have poured billions into AI training clusters. Nvidia’s GPUs and Broadcom’s ASIC accelerators are on backorder, while TSMC is running at full tilt to meet global demand.
But Morningstar warns that market expectations may be overextended. Unlike smartphones or PCs, which generated massive consumer-driven cycles, AI demand is heavily concentrated among a few mega-cap firms. If these players temper spending or consolidate workloads, the chip market could face oversupply shockwaves.
Even within AI infrastructure, not all chips are equally positioned. Cutting-edge GPUs and HBM (high-bandwidth memory) are in short supply, while legacy DRAM, NAND, and non-AI semiconductors are already seeing sluggish orders due to falling smartphone and PC sales.
Historical Parallels: Every Boom Has a Bust
The semiconductor industry’s past provides cautionary lessons:
- Dot-Com Bubble (2000): Internet infrastructure demand sent chip stocks soaring, only to collapse once spending outpaced adoption.
- Smartphone Supercycle (2010–2014): Apple and Samsung’s rapid expansion fueled record chip demand, but the eventual saturation led to years of weaker growth.
- Crypto Mining Surge (2017–2018): GPU demand skyrocketed for mining rigs before collapsing when crypto prices crashed.
AI could follow a similar trajectory: an overbuild phase followed by a reset, where excess supply pressures pricing until a new demand wave materializes.
TSMC and Foundries: Buffers, but Not Immune
Morningstar highlights Taiwan Semiconductor Manufacturing Company (TSMC) as relatively resilient due to its advanced process leadership and heavy U.S. investment push. The company’s 2nm and upcoming 1.4nm fabs have locked in commitments from AI giants, creating some cushion against downturns.
Still, even TSMC isn’t invulnerable. In previous cycles, foundries saw utilization rates plunge when macro conditions shifted. The difference now is that governments — from the U.S. to Japan and the EU — are pumping subsidies into chip sovereignty, potentially muting the severity of the next downturn but not eliminating it altogether.
Investor Reality Check: AI Hype vs. Profits
The broader question is whether AI’s hype translates into sustained corporate profits.
Goldman Sachs notes that while a record share of S&P 500 companies mentioned AI in earnings calls, only a handful quantified meaningful earnings impacts. In other words, AI is inflating capital expenditure and valuations, but its real economic footprint is still emerging.
Adding to the complexity, government data may be understating AI’s role in GDP, since semiconductor spending is often buried as “intermediate inputs.” That could mask the true size of AI’s economic contribution until broader adoption takes hold.
The Path Forward: Preparing for Volatility
For investors, the takeaway is clear: AI may delay the bust, but it won’t erase it. As 2025 approaches, risks of oversupply and tempered AI spending loom large. Key factors to watch include:
- AI Capex Announcements from hyperscalers (Microsoft, Amazon, Meta, Google).
- Semiconductor Billings Trends, which historically lead cycle shifts.
- Consumer Electronics Recovery — or lack thereof — to balance non-AI demand.
- Geopolitics and Tariffs, which could distort supply chains and accelerate volatility.
Final Thoughts: The Illusion of a Perpetual Boom
Morningstar’s warning serves as a reminder that the semiconductor industry, despite AI’s transformative promise, remains cyclical at its core. Investors betting on a perpetual AI-driven boom risk overlooking structural realities.
Yes, AI chips are scarce, valuations are high, and central banks’ policy shifts may keep fueling safe-haven buying of tech stocks. But history shows every upcycle has its limit. Whether in 2026 or beyond, a reset in the chip market is inevitable.
For long-term investors, the smarter play may be to balance enthusiasm for AI leaders like Nvidia and Broadcom with exposure to resilient foundries like TSMC — and a healthy respect for the volatility baked into the industry’s DNA.
Reference : Huileng Tan