The AI Boom Is Real, but the Easy Money Is Over: Why Semiconductor Volatility Is Rewriting the Investment Playbook

Author’s Note and Disclaimer:
Zion Zhao Real Estate | 88844623 | 狮家社小赵 | wa.me/6588844623 |  https://linktr.ee/zionzhao

This post is for general information, education, and market literacy only. It does not constitute financial, investment, trading, legal, tax, accounting, or other professional advice, and is not an offer, solicitation, recommendation, or endorsement. Views expressed are personal, general in nature, and subject to change without notice. While reasonable care is taken, no representation or warranty is given as to accuracy, completeness, or reliability. Readers should conduct independent due diligence and seek professional advice. To the fullest extent permitted by law, no liability is accepted for any loss arising from reliance on this material. 

AI Spending Surges as Chip Stocks Sink, Exposing a Growing Valuation Divide

The Stock Market Is Confusing, but the Confusion Is Telling Us Something

Semiconductor earnings remain strong. Artificial intelligence infrastructure spending is accelerating. Cloud providers still report capacity constraints. Demand for memory, networking, cooling, electricity and data-centre construction appears substantial.

Yet many of the stocks positioned as the greatest beneficiaries of the artificial intelligence boom have suffered violent corrections.

This is not necessarily evidence that artificial intelligence has failed. It may be evidence that the market is beginning to distinguish between technological importance, corporate quality and investment value.

Those are three very different things.

Strong Fundamentals Do Not Guarantee Rising Share Prices

Markets do not price current earnings alone. They price future cash flows, competitive advantages, interest rates, risk and the probability that expectations will be disappointed.

A company can report record revenue and still decline if investors expected even more. A business does not need to deteriorate for its valuation to fall. It only needs to perform less spectacularly than its share price previously implied.

This is especially important after an extraordinary rally.

The semiconductor sector entered July 2026 with substantial year-to-date gains before experiencing a sharp correction. Such a decline may reflect weaker fundamentals, but it can also reflect profit-taking, crowded positioning, portfolio rebalancing and valuation compression.

The crucial distinction is simple:

Artificial intelligence can transform the economy while some artificial intelligence stocks underperform.

Leverage Amplifies, but Does Not Fully Explain, the Selloff

Margin borrowing, leveraged exchange-traded products and speculative options activity can make market movements more violent.

When prices rise, collateral values increase and investors can borrow more. That additional capital pushes prices even higher. When prices reverse, margin calls and forced liquidation can create a feedback loop of selling.

Academic research describes this interaction as a liquidity spiral, where declining asset prices weaken funding conditions and tighter funding conditions force further selling (Brunnermeier & Pedersen, 2009).

However, leverage is probably an accelerant rather than the original cause.

The semiconductor sector is too large for retail liquidation alone to explain the entire decline. Institutional profit-taking, risk-control strategies, dealer hedging and portfolio concentration likely contributed.

The more important question is what caused investors to reconsider the price they were willing to pay.

Artificial Intelligence Demand Is Strong, but Profit Capture Is Uncertain

Microsoft, Meta, Oracle and other major technology companies continue to commit extraordinary amounts of capital to artificial intelligence infrastructure.

Microsoft projected approximately US$190 billion of capital expenditure during calendar year 2026 and indicated that computing capacity would remain constrained. Meta increased its 2026 capital-expenditure guidance to between US$125 billion and US$145 billion. Oracle reported hundreds of billions of dollars in remaining performance obligations, although its infrastructure programme also produced deeply negative free cash flow (Meta Platforms, 2026; Microsoft, 2026; Oracle Corporation, 2026).

The spending boom is real.

The uncertainty lies in who ultimately captures the economic value.

Model developers must compete with hyperscalers, open-weight systems and rapidly falling inference costs. Cloud platforms are designing proprietary chips. Customers are becoming more selective. Suppliers must continually reinvest in equipment that may become obsolete within only a few years.

Higher artificial intelligence usage does not guarantee high margins for every participant.

The airline industry carries enormous passenger volumes but often produces weak returns. Telecommunications networks process unprecedented quantities of data while pricing remains intensely competitive.

Artificial intelligence could follow a similar pattern. Consumption may explode while profits migrate unpredictably across model providers, chipmakers, cloud platforms, data centres, energy suppliers and software companies.

Open Models Are Changing the Economics

The emergence of increasingly capable open-weight models raises legitimate questions about the pricing power of proprietary artificial intelligence systems.

If enterprises can run capable models within their own infrastructure, they may reduce dependence on expensive external application programming interfaces. Local deployment can also improve data sovereignty, customisation, latency and cost control.

However, “open” does not automatically mean free, safe or fully reproducible. Organisations must still evaluate licensing, cybersecurity, maintenance, hardware costs, regulatory compliance and model provenance.

More importantly, cheaper models may increase total computing demand rather than reduce it.

When the cost of intelligence declines, businesses can apply it to more workflows, including customer service, coding, research, compliance, education and industrial optimisation. Efficiency may therefore stimulate consumption.

The International Energy Agency has noted that improvements in efficiency per artificial intelligence task may be offset by more complex reasoning and agentic workloads requiring greater computation (International Energy Agency, 2026).

This creates a critical distinction:

Artificial intelligence adoption determines the size of the market. Competitive structure determines who keeps the profit.

The Semiconductor Cycle Has Not Been Abolished

Artificial intelligence has strengthened demand for high-bandwidth memory, accelerators, networking and storage. It has not eliminated semiconductor cyclicality.

The traditional sequence remains relevant:

Demand rises. Prices increase. Margins expand. Manufacturers add capacity. Supply eventually catches up. Pricing power weakens.

The present cycle may last longer because artificial intelligence infrastructure requires far more computing intensity than previous consumer technology cycles. Nevertheless, suppliers still face the risk of overcapacity, inventory normalisation and customer bargaining power.

A low forward price-to-earnings ratio can also be misleading. Cyclical companies often appear cheapest near peak earnings because the denominator is unusually strong.

Investors must therefore ask how much of current profitability is structural and how much reflects temporary scarcity.

Energy Is Becoming the Real Infrastructure Constraint

Artificial intelligence is increasingly an electricity story.

The International Energy Agency estimates that global data-centre electricity consumption could approach 945 terawatt-hours by 2030, approximately twice the 2024 level (International Energy Agency, 2025).

The global share may appear manageable, but the local impact can be severe. Large facilities can place pressure on regional generation, transmission, transformers, water resources and household electricity prices.

Nuclear power may become part of the solution because it offers reliable, low-carbon electricity. Natural gas, renewables, storage, grid upgrades and behind-the-meter generation will also be required.

Governments should neither prohibit data-centre development indiscriminately nor allow private companies to transfer infrastructure costs to residents.

A credible regulatory framework should require developers to fund attributable grid upgrades, disclose water and emissions impacts, secure reliable electricity and deliver measurable community benefits.

Euphoria and Business Quality Are Different Questions

A company can be excellent while its stock becomes euphoric.

Investor attention tends to concentrate on securities experiencing unusual price movements, heavy media coverage and compelling narratives (Barber & Odean, 2008). During speculative periods, excitement can push valuations far beyond what even strong fundamentals justify.

The correct analytical framework separates:

1. Business quality

2. Financial quality

3. Valuation

A company can possess outstanding technology, rising revenue and a strong balance sheet while still being a poor investment at an excessive price.

Similarly, a correction does not automatically create value. The previous record high is not intrinsic value.

The Confusion Is the Message

The market may not be declaring that artificial intelligence is finished.

It may be recognising that the original narrative was too simple.

Artificial intelligence infrastructure spending can continue while supplier valuations decline. Open models can expand adoption while compressing proprietary pricing. Semiconductor revenue can grow while multiples fall. Data centres can create economic value while raising legitimate energy and regulatory concerns.

Investors should avoid two dangerous conclusions:

“The stock has fallen, so the thesis must be broken.”

“The technology is revolutionary, so any valuation is justified.”

Both replace analysis with emotion.

The appropriate response is disciplined questioning. What assumptions are embedded in the price? Who captures the profit? How durable are the margins? How much capital must be reinvested? Can the balance sheet survive disappointment? What evidence would invalidate the thesis?

The stock market is confusing because the artificial intelligence revolution is real, but the distribution of its profits remains unresolved.

That is not a reason to stop investing.

It is a reason to become more selective.

References

Barber, B. M., & Odean, T. (2008). All that glitters: The effect of attention and news on the buying behavior of individual and institutional investors. The Review of Financial Studies, 21(2), 785–818.

Brunnermeier, M. K., & Pedersen, L. H. (2009). Market liquidity and funding liquidity. The Review of Financial Studies, 22(6), 2201–2238.

International Energy Agency. (2025). Energy and AI.

International Energy Agency. (2026). Key questions on energy and artificial intelligence.

Meta Platforms, Inc. (2026). Meta reports first quarter 2026 results.

Microsoft Corporation. (2026). Fiscal year 2026 third quarter earnings conference call.

Oracle Corporation. (2026). Oracle announces record fourth-quarter and fiscal 2026 results driven by cloud infrastructure and cloud applications.