Amid the NY Dow hitting a new high, AI investment logic is u…

However, behind the record highs in equity indices, the core dynamics and underlying logic driving market trades are undergoing a fundamental paradigm shift. From macroeconomic factors down to micro-level industry dynamics, the rationale underpinning AI investments is caught in a whirlwind of ‘reassessment and reversal.’

The biggest shift in second-half investment logic lies in what Morgan Stanley’s Wilson team has proposed:the ‘Fourth Reset.’The recent volatility in the semiconductor sector does not signal the end of the AI cycle but rather a ‘gear shift’ in trading following the peak in the second derivative (i.e., acceleration) of capital expenditure growth.

This would mark the first time in history that annual AI capital expenditures surpass a nation’s defense spending (estimated at roughly 2.7% of GDP).

Goldman Sachs’ latest report also confirms this historic scale-up, forecasting that global AI-related capital expenditures on computing, data centers, and power will reach $7.6 trillion between 2026 and 2031. Investments by hyperscalers alone could exceed $6 trillion by 2030.

However, this unprecedented level of spending has created a fundamental tension between robust demand across the supply chain and mounting cash flow pressure on customers.Until now, upstream semiconductors—including GPUs, memory, manufacturing equipment, and printed circuit boards—have surged due to explosive order growth, while downstream hyperscalers have struggled under the strain on free cash flow caused by heavy capital expenditures.

The market is now demanding rigorous validation of return on investment.Semiconductor valuations can no longer be justified solely by anticipated budget increases from hyperscalers; investors must now assess whether those budgets translate into actual revenue, gross profit, and solid free cash flow.。This allows them to enterThe logic underpinning semiconductor valuations has fundamentally shifted—from whether customers will keep spending, to whether they can generate profits after spending.

Within the semiconductor sector, the memory segment is most prone to serving as a ‘leading indicator’ of downward adjustments.Morgan Stanley likens memory stocks to ‘silver-related equities,’ as both have experienced parabolic surges and possess strong commodity-like characteristics.

DRAM, NAND, HBM, and eSSD all significantly benefit from AI data centers, with price increases and long-term contracts underpinning robust fundamentals. However, memory remains the most commoditized segment within semiconductors.
Stock prices are highly susceptible to amplified volatility due to factors such as ‘peaking price momentum’ and concentrated capital flows.
In particular, after South Korea’s $SK hynix (SKHY.US)$ officially launched a $28 billion ADR offering on the US Nasdaq, $Micron Technology (MU.US)$ the ‘unique scarcity’ of
in the US-listed AI memory space has eroded. The interplay of capital rotation dynamics and high volatility stemming from its commodity nature means memory is now simultaneously the sector with the strongest fundamentals and the lowest risk tolerance for trading.

Simultaneously, in this new normal ‘validation period’ where true value is being tested, the competition in computing power has evolved beyond mere semiconductor specification battles into system-level integration. Technical barriers and physical limitations now represent new market constraints.

As market focus broadens, the investment framework is shifting from ‘who wins orders’ to ‘who can successfully navigate the validation phase.’ Investors should monitor five core validation signals.

At the same time, Goldman Sachs pointed out that AI investment focus is broadening from pure-play semiconductors to real-economy sectors such as manufacturing, energy, logistics, and defense. This significantly overlaps with the constituents of the Dow Jones Industrial Average, and capital is flowing into these ‘laggard sectors’ that have tangible industrial fundamentals.

Their greatest strength lies in their ‘abundant optionality.’ They control end-demand and can reliably convert investments into cash flow through ad monetization, cloud software, enterprise AI agents, and—like META—leasing out computing capacity externally. These companies are currently at a critical inflection point, transitioning from ‘cost bearers’ to ‘provers of return on investment.’

・HBM (High Bandwidth Memory) – the golden segment:

This sector exhibits the highest ‘AI purity,’ yet appears overvalued at current levels. Its core value lies in market share and yield. The industry leader$SK hynix (SKHY.US)$ is the strongest candidate to absorb inflows from U.S. equity passive funds. Samsung must focus on regaining ground through HBM4 certification and capturing customer share.$Micron Technology (MU.US)$ benefits from solid long-term contracts and its position as the gateway to AI in the U.S., making it highly attractive following recent pullbacks.

・eSSD (Enterprise SSD) – growth phase:

With AI-driven data center demand expected to surge from 18% in 2025 to 41% by 2027, the key value driver lies not in pure NAND price increases, but in certifications, controller firmware, and the proportion of high-capacity QLC. [Company] holds a significant advantage in enterprise certifications. $SanDisk (SNDK.US)$ and $Western Digital (WDC.US)$ in focus.

・Nearline HDD (Hard Disk Drive) as a cash flow asset:

Its investment thesis hinges heavily on securing cloud customer orders and transitioning to free cash flow through yield improvements in HAMR (Heat-Assisted Magnetic Recording) and Mozaic technologies. The company possesses significant potential for aggressive share buybacks and dividends. $Seagate Technology (STX.US)$ in focus.

・Structural shortage in DDR4/DDR5 commodity memory:

Massive production capacity has been diverted to HBM, creating a supply-demand mismatch in commodity memory. It is essential to continuously validate the sustainability of contract pricing and prepare for the risk of legacy production capacity returning in the future.

Beyond the gear shift within AI’s dominant theme, market broadening is shaping the second core theme for the second half of the year.

The waning momentum in semiconductors does not imply systemic risk. As long as capital remains in the market and continues seeking new sources of profit, this rotation remains healthy. The recent decline in crude oil prices has substantially eased inflationary pressures, and softening employment data has significantly reduced market expectations for rate hikes by the U.S. Federal Reserve. Against this macro backdrop, laggard sectors previously suppressed by large-cap, AI-driven stocks now have a favorable opportunity for valuation repair.

Capital is rapidly rotating away from hardware giants with high concentration of funds toward sectors highly sensitive to interest rates and economic recovery—such as consumer goods, transportation, regional banks, and biotechnology. If these sectors can demonstrate tangible evidence of earnings improvement (e.g., lower costs in transportation or stabilized net interest margins at regional banks), they could significantly alleviate the downward pressure on equity indices caused by the semiconductor sector’s gear shift.

The Dow Jones Industrial Average surpassing 53,000 marks a key milestone, reflecting that the broader market remains solid, supported by favorable macro conditions and policy tailwinds. However, beneath the surface, a shift in direction is underway. While AI infrastructure continues to expand, the market has entered an extremely stringent phase of earnings validation. Future market performance will hinge on hyperscalers’ capital expenditure outlooks, stabilization in computing capacity lease pricing, and the strength of long-term memory chip contracts.