LWAYS OPEN
The artificial intelligence, semiconductor and memory markets are experiencing their most important reassessment since the current AI investment cycle began. After an extraordinary rally during the first half of 2026, many of the companies that benefited most from the global expansion of AI infrastructure have suffered sharp share-price declines.
30. Juy 2026, 09:00 AM (Irish Time) The correction has affected semiconductor designers, memory manufacturers, networking specialists, data-centre suppliers, semiconductor equipment companies and major technology platforms. In several cases, shares have fallen even after companies reported record revenue, exceptional margins and continued demand growth.
This apparent contradiction has created uncertainty among investors.
Is the AI boom ending? Are semiconductor and memory companies entering a new downturn? Have the major technology groups invested too aggressively in artificial intelligence? Or is the market simply correcting excessive valuations after an unusually strong period?
Final Resurrection Ltd. initiated a detailed review of the AI, semiconductor and memory markets after the owner of Portfolio HB005 questioned whether the market environment had changed fundamentally and required a different investment approach.
The conclusion is important.
The AI investment cycle has not yet collapsed. However, the first phase of the AI equity trade has ended.
The market is moving from enthusiasm to financial scrutiny, from broad AI speculation to company-specific selection, and from assumptions of unlimited growth to a much closer examination of valuations, financing structures and actual investment returns.
The Market Did Not Fall Continuously From May
The development since May 2026 must be understood in three separate stages.
May was not the beginning of a broad decline. It represented the final acceleration of the AI rally.
Demand for advanced semiconductors, High-Bandwidth Memory and data-centre infrastructure remained exceptionally strong. Memory prices were rising rapidly, hyperscalers continued increasing capital expenditure, and investors expected shortages in GPUs, HBM and advanced manufacturing capacity to continue for several years.
Companies linked to AI infrastructure received increasingly high valuations. Investors were prepared to pay substantial premiums for businesses involved in semiconductor manufacturing, memory, networking, optical connectivity, power management and data-centre cooling.
June marked the second stage.
The rally continued, but valuations became increasingly difficult to justify without assuming almost perfect future execution. Several companies had reached levels that already reflected years of expected growth.
The first warning sign appeared when strong corporate results no longer produced equally strong share-price reactions.
Companies continued to report favourable demand, but investors began selling positive news rather than buying it. Hedge funds started reducing semiconductor exposure, market leadership became narrower, and capital began rotating into previously neglected sectors.
July became the third and decisive stage.
The gradual shift developed into a broad reset across the entire AI infrastructure complex. Semiconductor indices entered correction or bear-market territory, highly valued technology companies declined, and the South Korean market experienced exceptional volatility because of its concentration in Samsung Electronics and SK Hynix.
The correct chronology is therefore clear.
May was the final acceleration. June was the period of overextension and peak formation. July became the valuation and positioning reset.
Why Are AI Stocks Falling When Demand Remains Strong?
The current correction is not primarily being caused by a collapse in demand.
Leading companies throughout the semiconductor supply chain continue to report strong operating conditions.
TSMC remains highly profitable and continues to experience substantial demand for advanced manufacturing capacity. ASML has maintained a strong outlook for semiconductor equipment. Micron has reported exceptionally high memory revenue and demand exceeding available supply. Microsoft, Meta, Alphabet and Amazon continue to invest enormous amounts in AI infrastructure.
The weakness in share prices is therefore being caused by a change in what investors are willing to pay for future growth.
During the first stage of the AI cycle, rising capital expenditure was interpreted almost entirely as positive.
More spending by Microsoft, Meta, Alphabet or Amazon meant more demand for Nvidia accelerators, TSMC manufacturing, Micron memory, Broadcom custom chips, Arista networking equipment and data-centre power systems.
The market is now considering the other side of the same investment cycle.
Higher expenditure also means higher depreciation, lower short-term free cash flow, greater electricity consumption, faster hardware obsolescence and pressure on operating margins.
Investors are no longer asking only how much AI infrastructure is being built.
They are asking how much revenue and profit will ultimately be generated by every additional dollar invested.
That is a fundamental change in market psychology.
The AI Market Is Experiencing a Valuation Reset
Many AI-related companies were valued on the assumption that demand would continue increasing almost without interruption.
The market expected advanced GPUs and HBM to remain permanently scarce. It assumed hyperscaler expenditure would rise every year, margins would remain exceptionally high and Chinese competitors would take many years to become relevant.
At such valuation levels, strong earnings were no longer enough.
Companies were expected to exceed already exceptional forecasts, raise guidance and provide evidence that demand would continue accelerating.
When results were merely excellent rather than extraordinary, share prices declined.
This is known as valuation compression.
A company can remain profitable, increase revenue and maintain a strong competitive position while its share price falls because investors are no longer willing to apply the same valuation multiple to future earnings.
The present correction is therefore not proof that the AI business model has failed.
It is evidence that the market had priced many companies too close to perfection.
The AI Trade Had Become Too Crowded
Another major reason for the correction is the concentration of investment in a relatively small number of related companies.
Many portfolios simultaneously held positions in Nvidia, Broadcom, TSMC, Micron, Lam Research, Arista Networks, Marvell, Vertiv, Astera Labs, Samsung, SK Hynix and semiconductor exchange-traded funds.
These companies operate in different areas, but they are influenced by many of the same factors.
They depend on continuing hyperscaler investment, data-centre construction, GPU deployment, memory pricing, access to financing and confidence in future AI demand.
During a rising market, these positions may appear diversified.
During a correction, they can behave like one highly concentrated investment.
When the market began falling, the decline was intensified by stop-loss orders, margin calls, leveraged investment products, hedge-fund deleveraging and institutional risk reduction.
The result was not simply a reassessment of individual companies. It was a broad reduction in exposure to the AI investment theme itself.
The Market Is Questioning the Return on AI Investment
The central question facing the AI market is now financial rather than technological.
Few serious observers doubt that artificial intelligence will continue developing or that demand for computing power will grow.
The uncertainty concerns the economic return.
Microsoft, Meta, Alphabet and Amazon are investing extraordinary sums in data centres, processors, networks and power infrastructure. These companies possess strong balance sheets and substantial cash flows, but the investment programmes are becoming large enough to influence margins and free cash flow.
Investors now want evidence that AI revenue can grow quickly enough to justify the expenditure.
This creates a difficult situation for the market.
If hyperscalers continue increasing capital expenditure, semiconductor suppliers benefit, but investors may worry about the hyperscalers’ future profitability.
If hyperscalers reduce expenditure, their free cash flow may improve, but suppliers throughout the semiconductor and data-centre ecosystem could lose expected orders.
The market is therefore searching for a sustainable balance between investment growth and financial return.
Memory Is Strong Fundamentally but Vulnerable Cyclically
The memory market represents the clearest contradiction in the current environment.
Demand for DRAM, NAND and High-Bandwidth Memory remains extremely strong. AI servers require far more memory than traditional servers, and advanced accelerators depend heavily on HBM.
Manufacturers continue to report tight supply, rising prices and strong data-centre demand.
Nevertheless, memory shares have corrected sharply.
This is because memory equities rarely reflect only current profits. They reflect expectations for prices, capacity and margins many months into the future.
The traditional memory cycle begins with rising demand and limited supply. Prices increase, margins expand and manufacturers report exceptional earnings. Companies then increase capital expenditure, customers build inventory and production capacity begins to improve.
The equity market usually anticipates the next stage before the financial results weaken.
Memory shares can therefore begin falling while prices are still rising and profits are still reaching records.
The market may already be anticipating improved production yields, additional capacity, weaker future scarcity and a possible normalisation of HBM and conventional DRAM pricing.
This does not mean that the current memory boom has already ended.
It means that memory equities may be entering a late-cycle phase even while the operating environment remains highly profitable.
China Has Become a More Important Risk
Chinese progress in memory, semiconductor equipment, AI accelerators and lower-cost artificial intelligence models has also contributed to the correction.
China remains behind the global leaders in the most advanced HBM products, leading-edge semiconductor manufacturing and extreme ultraviolet lithography.
The immediate threat to the highest-performance products offered by SK Hynix, Samsung, Micron, TSMC and ASML therefore remains limited.
The medium-term risk is more significant.
Chinese manufacturers do not need to become immediate technology leaders to affect global valuations. They only need to satisfy a larger proportion of domestic demand, expand production in standard DRAM and NAND, or reduce the expected duration of the current supply shortage.
The greatest risk may therefore emerge first in conventional memory rather than the most advanced HBM segment.
Additional Chinese capacity could eventually place pressure on commodity memory pricing, reduce margins and force established producers to increase investment.
Financing May Become the Most Important Future Risk
The first wave of AI investment was financed mainly by highly profitable technology companies with strong operating cash flow.
The next stage may involve more debt, private credit and specialised AI infrastructure providers.
This creates a new risk.
Smaller GPU-cloud companies and data-centre operators may depend on long-term customer commitments, external financing and a small number of major clients. Some businesses are required to invest heavily before receiving the corresponding revenue.
A correction in credit markets could therefore have a direct effect on AI infrastructure orders.
If a heavily financed data-centre operator became unable to meet its obligations, the problem could spread to equipment suppliers, chip manufacturers, lenders and other infrastructure providers.
There is not yet sufficient evidence of a systemic financing crisis.
However, this is the mechanism most capable of transforming the current stock-market reset into a genuine reduction in AI investment.
Is This Another Dotcom Bubble?
The current AI cycle shares several characteristics with the technology boom of the late 1990s.
Market leadership has become highly concentrated. Valuations reached exceptional levels. Investors accepted optimistic long-term assumptions, and enormous sums were committed to infrastructure.
There are also concerns about circular commercial relationships and the possibility that some demand has been supported by financing connected to suppliers or strategic investors.
However, there are important differences.
The leading AI companies today generate substantial revenue, profits and operating cash flow. They possess established customers, valuable intellectual property and technologies that are difficult to reproduce.
The present situation therefore does not closely resemble the weakest parts of the dotcom boom, where many companies had limited revenue and no credible path to profitability.
A more realistic comparison is an infrastructure and valuation cycle in which genuine long-term winners can still lose between 30 and 60 percent of their market value before the next sustainable expansion begins.
A strong long-term business does not protect a share from a severe short-term correction.
What Would Confirm a Fundamental AI Reset?
The current correction remains primarily a valuation and positioning event.
A genuine fundamental reset would require more serious evidence.
The most important warning would be a reduction in AI capital-expenditure guidance by a major hyperscaler.
Other critical signals would include falling HBM contract prices, rapidly increasing memory inventories, postponed data-centre projects, delayed GPU orders or financial difficulties among major AI infrastructure providers.
A reduction in revenue or margin guidance from TSMC would also carry considerable significance because the company has direct visibility across much of the advanced semiconductor supply chain.
The same would apply if Micron reported weaker orders and rising inventory, or if Nvidia and Broadcom began reporting material order cancellations.
Until several of these signals appear, it would be premature to conclude that the underlying AI investment cycle has ended.
The Most Likely Market Scenario
The most probable outcome is an extended period of valuation normalisation.
AI and semiconductor equities may remain volatile for several months. Strong rebounds are likely, but they may initially be followed by further selling.
During this period, company fundamentals may remain strong while valuations continue to adjust.
The market will become increasingly selective.
Companies with strong balance sheets, durable competitive advantages and visible customer demand are likely to stabilise earlier than highly valued or highly leveraged infrastructure businesses.
A rapid return to the conditions of early 2026 appears less likely than a longer process of consolidation and differentiation.
A renewed sustainable rally would require improving market breadth, stable capital-expenditure guidance, continued strength in HBM pricing and constructive share-price reactions to corporate earnings.
What the Current Reset Means for Investors
The most important lesson is that a falling share price does not automatically create an attractive investment opportunity.
A company that has already declined by 25 percent can decline by another 25 percent during a valuation reset even if its business continues growing.
Investors should also avoid treating every short-term rebound as proof that the previous bull market has resumed.
After severe corrections, strong technical recoveries are normal. A sustainable trend change usually requires a higher low, a higher high, improving market breadth and positive reactions to earnings announcements.
The distinction between business quality and stock-market risk is essential.
TSMC, Microsoft, Alphabet, ASML and other major companies may possess exceptional long-term businesses, but their shares can still decline when valuations, interest rates or capital-expenditure expectations change.
Smaller AI infrastructure companies, highly cyclical memory shares and leveraged investment products face even greater risk during an unstable market.
Cash should therefore not be viewed as an abandonment of the AI investment thesis.
It can provide protection, flexibility and the ability to act after a genuine market bottom has formed.
Final Assessment
The AI trade is undergoing a reset, but the AI investment cycle is not yet over.
The current market is correcting excessive valuations, crowded positioning, leveraged speculation and unrealistic assumptions about permanent shortages and continuously rising capital expenditure.
The fundamental demand for advanced computing, semiconductor manufacturing, memory and data-centre infrastructure remains strong.
However, the market is entering a more mature phase.
Future performance will depend less on whether a company is associated with artificial intelligence and more on whether it can demonstrate sustainable revenue, credible returns on investment, controlled leverage, strong free cash flow and a defensible technological advantage.
The period in which rising AI expenditure automatically produced rising equity prices has ended, at least for now.
The next stage of the market will be defined by greater selection, stricter financial analysis and a much lower tolerance for excessive valuations.
The AI supercycle may continue.
The easy AI trade has already changed.