AI bubble worries and market concentration risk

In the first half of 2026, global stock markets have continued their strong performance driven by AI, but internal market signals show a marked rise in uncertainty. Artificial intelligence and related companies now account for about 45% of the market capitalization of the S&P 500, far exceeding their weighting in the Nasdaq, creating an extreme concentration. This kind of narrow, growth-led drive is common in past technology cycles, but it also points to potential adjustment pressure. In this environment, investors should prudently assess risks, consider taking profits and using moderate hedges, while also keeping an eye on long-term opportunities in hard assets and key commodities.

AI mania: bubble traits and historical parallels

The AI wave shows striking similarities to the internet bubble at the end of the last century. Both were driven by transformative technological breakthroughs, drawing massive capital inflows and causing valuations to inflate rapidly. Universities such as Stanford—among the talent pipelines of Silicon Valley—witnessed the paradigm shift from the internet to AI. Today, AI infrastructure buildout relies on analysts’ optimistic forecasts for capital expenditures (CapEx): related spending in 2026 is expected to reach about $800 billion, rising further to $1.1 trillion in 2027. But these projections are based on idealized Excel models that ignore real-world physical constraints.

Insiders have observed that even the tech elite have begun to de-risk significantly. Some early beneficiaries have dramatically reduced technology positions in their portfolios, rotating into defensive assets such as U.S. Treasuries. This “insider de-risking” phenomenon is worth watching: in the late stage of a bubble, those who truly understand the industry often act first rather than waiting for a clear top signal. Similar to the 1997–2000 internet cycle, the market needs actual testing to reveal the turning point. Currently, AI trading may be at a stage similar to around 1998, but it remains uncertain how long it will take for a full correction; even so, valuations have already lost fundamental support.

Physical constraints are becoming the main bottleneck for AI expansion. Data center construction faces shortages of land, permits, water resources, and power supply—along with hardware constraints such as chips and capacitors. In multiple places, construction has been paused or delayed. For example, in Reno, Nevada, the city has paused approvals for new data centers mainly due to tight electricity and water resources. Institutions such as Blackstone have recently exited U.S. projects for the largest data centers, further underscoring the difficulty of execution. Even with abundant capital, actual buildout plans may be cut from the forecast 100% down to 75–80%, leading to a reassessment on Wall Street and a repricing of sectors.

In addition, the depreciation characteristics of data centers differ from those of past infrastructure. Once fiber optics or rail lines are built, they remain usable for a long time, whereas AI chips face technical obsolescence in roughly three years, with upgrade costs of about two-thirds of new builds. This will affect the profitability of companies in the 2.0 era and amplify cyclical risks. Semiconductors are a highly cyclical industry; even if today’s hot market may be hard to sustain, the narrow, concentrated trading pattern further amplifies system fragility.

Market concentration and macroeconomic outlook

Core stocks such as the Magnificent Seven have shown divergent performance, and some have already seen adjustments this year. The lack of market breadth has become more apparent. While the AI boom is lifting indexes, if hyperscalers continue to lag and sub-sectors such as semiconductors cannot provide a follow-through, the risk of capital outflows could spill into the broader market. Analysts expect that over the remainder of 2026 there may be a 10–15% pullback, and potentially stage-by-stage corrections similar to those during the 1998 Russian debt crisis.

On the macro side, the risk of decoupling between economic performance and market performance is rising. Even though K-shaped economic traits are visible—consumer debt rising, the savings rate falling, and credit defaults increasing—an initial “capital flood” from AI CapEx and government deficit spending is supporting economic growth in the short term. Interest expense in 2026 has already exceeded $827 billion, and full-year totals could break $1.1 trillion; the long-term trajectory points to challenges in fiscal sustainability. If inflation expectations heat up alongside a potentially hawkish stance from the Federal Reserve, market volatility will likely be amplified further.

If wage growth can keep pace with inflation, consumption resilience can be maintained; otherwise, pressure accumulating in the lower tiers may lead to social consequences. History shows that when K-shaped divergence intensifies, the share of lower-tier voters is often higher, which may push policy to shift. The victory of candidates with socialist leanings in some regions reflects this dynamic. Over the long run, declining labor force participation combined with AI and robots’ double substitution of white-collar and blue-collar jobs will magnify the risk of structural unemployment. If technological unemployment is not accompanied by new job creation, it could trigger a social crisis beyond the economic sphere.

Social structural change and Western institutional resilience

Income inequality and an excess of elites are potential sources of instability. If highly educated groups cannot share in prosperity, they may become a catalyst for change. Historical cases show that Marxist-style ideologies often begin spreading in elite universities before trickling down. In parts of today’s education systems, some narratives emphasize “deconstruction” rather than inheritance, which may weaken identification with the values of Western civilization. External forces could also use this to amplify internal fractures, creating long-term strategic pressure.

The way forward is to return to truly free-market capitalism: allow creative destruction, clear mismatched investments, and simultaneously build a fair regulatory framework. Some East Asian economies have achieved rapid poverty reduction by combining long-term planning with market mechanisms, offering a reference for mixed governance. But the core success factors are still rule of law, protection of property rights, and incentive compatibility—not command-style economics. The West needs to repair the crony-capitalism problem in existing models rather than switching to alternative paths with a poor historical track record.

While combining AI with robots accelerates labor substitution, it also creates new opportunities—such as modular data centers and micro-reactor technologies—which can improve deployment flexibility. Future employment structure will shift toward high-skill, entrepreneurial roles, but coverage will be limited. Policy should plan ahead for retraining and social safety nets to cushion the shock of transition.

Hard assets and commodity allocation opportunities

In an uncertain environment, precious metals show defensive value. After gold and silver surged early in 2026 on speculative inflows and then pulled back, gold is currently hovering around $4,000–$4,150 per ounce, while silver is in the $58–$62 per ounce range. Falling oil prices ease some pressure, and combined with long-term expectations of debt monetization, precious metals may have already completed de-bubbling and now show bottoming characteristics.

Industrial metals such as copper have excellent long-term prospects. AI data centers, electrification, and renewable energy transitions will drive strong demand. The U.S. list of critical minerals has been updated to include copper, silver, uranium, and others, highlighting the importance of supply-chain security. Even if the pace of AI buildout construction growth slows in the short term, infrastructure expansion will still require vast amounts of copper resources. Investors may consider a diversified commodities basket—including energy, critical minerals, and related physical assets—as a tool to hedge fiscal and geopolitical risks.

Investment strategy recommendations

For the second half of 2026, a defensive allocation approach is recommended:

• Take profits and hedge—reduce exposure modestly in AI-heavy sectors, and consider volatility products or defensive assets.

• Tilt toward hard assets—increase exposure to precious metals, copper, and broad commodities to address potential currency depreciation.

• Diversification—focus on real estate that can be repurposed for data center use, as well as areas that benefit from the energy transition.

• Long-term perspective—AI transformative potential remains, but it must pass through bubble corrections and the infrastructure digestion period.

Tightening geopolitical conditions (such as the situation in the Middle East) and the fiscal trajectory jointly point to a volatile environment. Investors should evaluate independently, avoid chasing hot themes, and focus on fundamentals and risk management. Historical experience suggests that when turning points in technology cycles coincide with macro pressures, prudent allocation often outperforms concentrated bets. Over the next few years, the key to success lies in balancing innovation opportunities with system resilience and adapting to structural transitions.