In the first half of 2026, the AI chip bull market fostered a narrative where investors grew accustomed to viewing every market pullback as a buying opportunity. However, the market performance in the first week of July put this familiar logic through a genuine stress test.
From July 1 to 2, the Philadelphia Semiconductor Index (SOX) dropped more than 11% over two trading days. Marvell Technology plunged 9.84% in a single day, hitting an intraday low of $237.20. Over two days, its cumulative decline exceeded 18%. From its historic high near $330 in June, Marvell has pulled back more than 25% in roughly three weeks. Micron Technology fell over 11%, Intel dropped 9%, and AMD gave back 7%. The VanEck Semiconductor ETF declined more than 5%. This sell-off wasn’t an isolated event—it reflected a systemic revaluation across the entire AI hardware supply chain.
Confronted with the same market conditions, two of Wall Street’s top investment banks offered sharply contrasting advice. JPMorgan strategist Mislav Matejka stated clearly that the recent weakness in semiconductor stocks should be seen as a buying opportunity. Meanwhile, Morgan Stanley’s chief US equity strategist Michael Wilson signaled to clients: reduce exposure to semiconductors and shift toward hyperscale cloud providers.
Their disagreement doesn’t stem from diverging views on the long-term prospects of the AI industry—both agree the long-term AI trend remains intact. The real difference lies in their expectations for short-term valuations, market sentiment, and the pace of the next rally. This article unpacks the core logic behind both banks’ perspectives, analyzes potential drivers for the AI chip sector, and explores the structural impact of the AI boom on the crypto market.
Drivers Behind the Pullback: Multiple Factors in Play, Not a Reversal of Logic
Before examining the banks’ divergence, it’s important to clarify the causes behind this round of market correction.
First, excessive gains earlier triggered concentrated profit-taking. The Philadelphia Semiconductor Index surged over 80% in the first half of 2026. The memory segment soared 318.49%, leading all US equity sub-sectors; computer hardware rose 165%, and semiconductor equipment and materials climbed 129%. After such massive short-term gains, any marginal negative news can spark widespread profit-taking. Trading volumes spiked during the first week of July’s sell-off, reflecting a release of pent-up market anxiety.
Second, Meta’s announcement acted as the emotional tipping point. Last week, Meta announced it would begin selling its excess computing power to external customers. The market interpreted this as a sign that even hyperscale cloud providers—with capital expenditure guidance as high as $145 billion in 2026—may face excess capacity. For the past two years, the semiconductor sector traded on the assumption of persistent shortages in GPUs and high-end memory. If Meta has idle capacity available for rent, future orders for GPUs, HBM, and NAND flash may shrink. Wilson’s report bluntly stated that this move signals "the capital expenditure growth rate for hyperscale cloud providers may be hitting a temporary inflection point."
Third, Citigroup analysts amplified negative sentiment with their warnings. Citi questioned whether large cloud platforms can convince investors that massive AI infrastructure spending will yield substantial returns, and whether such high expenditures are sustainable. This goes straight to the heart of the AI investment wave’s narrative—when the visibility of capital expenditure returns becomes blurred, the foundation of the entire valuation system weakens. Goldman Sachs data shows that Alphabet, Amazon, Microsoft, and Meta will collectively spend $725 billion on capital expenditures in 2026, up 77% from 2025; hyperscale cloud providers’ capex will reach about 100% of operating cash flow. Yet, when AI businesses will deliver matching profits remains unclear.
Fourth, valuations remain elevated with little margin of safety. Even after nearly a 10% single-day drop, Marvell’s rolling P/E as of July 2 was still about 84x, significantly higher than the semiconductor industry average (around 75.5x). Nvidia’s current forward P/E is about 22x; after a flat first half, its valuation appears "relatively cheap," but AMD and Intel, after triple-digit rallies, are now in expensive territory. Goldman Sachs notes Nvidia’s P/E has fallen to the lower-middle range of the past three years—highlighting just how high valuations were before this correction.
Overall, this pullback is more of a cyclical adjustment than a fundamental shift in the AI industry’s logic. Goldman Sachs forecasts global AI-related capital expenditures for computing, data centers, and power will reach about $7.6 trillion from 2026 to 2031. Worldwide data center capacity has grown from 30 GW in 2019 to 57 GW in 2024, with another 65 GW expected by 2030. AI demand is growing faster than infrastructure development. These macro data points show AI infrastructure expansion is still in its early stages.
The Real Disagreement: Two Views on Timing, Same Direction
The core of the banks’ disagreement can be summed up simply: Both are bullish on AI, but differ on whether "now" is the right time to buy.
JPMorgan: Pullbacks Are Opportunities, the Upcycle Is Far from Over
JPMorgan strategist Matejka outlined three key arguments in his July 6 report.
First, the semiconductor upcycle hasn’t peaked. Matejka wrote, "Meaningful new supply is unlikely before 2028." High-bandwidth memory (HBM) supply from Micron, SK Hynix, and Samsung is sold out through 2026, and new wafer capacity likely won’t come online until after 2028. AI data centers are expected to consume about 70% of global memory chip output this year—analysts describe this as structural undersupply, giving producers sustained pricing power.
Second, recent pullbacks in the Philadelphia Semiconductor Index and Korean equities should be viewed as buying opportunities. JPMorgan’s tech sector allocation prioritizes "semiconductors over hyperscale cloud providers, and cloud providers over high-risk AI concept stocks." Its mid-year outlook named Broadcom as a "strong buy" for the remainder of 2026, emphasizing that the AI-driven chip cycle is far from over. JPMorgan analyst Harlan Sur noted that AI chips have a backlog of orders far exceeding current capacity, with revenue visibility extending well into the future.
Third, the macro environment is improving. Matejka believes that as stagflation fears recede, market participation will broaden in the second half of 2026. JPMorgan expects global equities to hit new highs, supported by strong earnings prospects, easing inflation pressures, and relatively light investor positioning.
It’s worth noting that JPMorgan isn’t indiscriminately bullish on all tech stocks. The bank is cautious on the "Magnificent Seven," suggesting that while earnings and valuation tailwinds persist, this group "may face ongoing downward valuation pressure due to questions about monetization prospects." For software, business services, and media—sectors vulnerable to AI’s cannibalization effects—JPMorgan maintains a "fundamentally bearish" outlook.
Morgan Stanley: Momentum Weakening, Wait for a Better Entry Point
Morgan Stanley’s view also rests on confidence in AI’s long-term trend, but it diverges on short-term timing.
Wilson’s latest report notes that semiconductor momentum has faded, with the Philadelphia Semiconductor Index down nearly 12% from its peak. The high-beta momentum stock basket (memory and chip stocks) saw its biggest two-day drop since the pandemic. Wilson believes this pullback "may have further room to run."
Morgan Stanley’s core logic is based on a "rotation trade" framework. As early as November 2025, Wilson’s annual outlook proposed a "market diffusion trade": after the US economy completed a rolling recession in April 2025 and entered a new expansion cycle, earnings growth would beat expectations, and leadership would shift from AI capex beneficiaries to broader sectors. This view was interrupted by the Iran war in February 2026, but with oil prices falling and inflation expectations stabilizing, Wilson thinks conditions are ripe again.
Morgan Stanley draws a specific analogy: semiconductor performance closely mirrors silver—both experienced parabolic price surges and are tightly linked to commodity markets. Morgan Stanley first introduced this analogy in early June and now sees it playing out. Wilson further notes that this correction will be led by the memory segment—because memory is the "most commodity-like" part of the semiconductor complex, with high price elasticity and rapid reversals.
In terms of portfolio strategy, Wilson recommends "sell chips, buy cloud"—reduce semiconductor holdings and shift toward hyperscale cloud providers. This isn’t a bearish call on AI, but a rotation. There have been three similar adjustments during the AI investment cycle, and Wilson sees this as the fourth. He favors Microsoft, Amazon, and Meta, believing their core businesses can buffer AI-related volatility. Wilson maintains a year-end S&P 500 target of 8,000.
Is the AI Bull Market Entering Its Second Phase?
The banks’ disagreement essentially points to a single question: Is the AI investment cycle shifting from "concept-driven" to "performance validation"?
Looking back at the first phase of the AI bull market—from 2023 through the first half of 2026—the market mostly traded on AI concepts and future expectations. In this phase, any stock with an AI angle saw broad gains, and the market tolerated sky-high valuations. The memory segment’s 318% gain in the first half, and Marvell’s more than 220% rally this year, are classic examples.
Now, the market is entering the second phase. This stage is characterized by greater focus on realized earnings, with clear differentiation among AI chip, cloud computing, and data center infrastructure companies. Real revenue and order growth are becoming the key drivers of stock performance. Global AI trading is "changing tracks," as capital shifts from "cash-burning" cloud providers to "profit-making" hardware suppliers. Market pricing logic is gradually moving toward validating earnings and operating cash flow.
This view is backed by solid data. On July 7, Samsung Electronics released preliminary Q2 results showing operating profit up 1,810.2% year-over-year, reaching KRW 89.4 trillion (about $58 billion), marking a third consecutive record quarter. Memory chip giant Jangbolong expects first-half net profit attributable to shareholders to jump 62,204% to 74,394% year-over-year. These figures demonstrate that performance validation is happening in the AI hardware segment.
At the same time, Goldman Sachs predicts hyperscale cloud providers’ capex will reach about 100% of operating cash flow. When capex growth far outpaces revenue growth, the sustainability of this model comes under scrutiny. The upcoming Q2 earnings reports will be a key observation window—for stocks with elevated valuations, sustained growth in operating cash flow and its alignment with consensus expectations will be the main focus.
Goldman Sachs notes investors are underweighting the "Magnificent Seven" and favoring semiconductors and other hardware segments that directly benefit from capex. Fund flow data confirms this trend: According to Wind, during the week from June 29 to July 3, the top three sectors for net inflows into equity ETFs were semiconductor chips, communications, and securities. JPMorgan also points out that AI trades are becoming more differentiated, with chips and hardware stocks attracting capital, while capital-intensive tech companies are being sold off—a pattern reminiscent of the lead-up to the 1999 dot-com bubble burst.
How Is the AI Boom Impacting the Crypto Market?
Fluctuations in AI chip prices are affecting not only traditional capital markets but also the crypto ecosystem through multiple channels.
First, AI infrastructure expansion is driving decentralized compute networks (DePIN). As global data center supply grew from 30 GW in 2019 to 57 GW in 2024, and another 65 GW is expected by 2030, decentralized physical infrastructure networks are emerging as alternatives to centralized computing power. As of March 2026, DePIN’s total market cap reached about $9–10 billion. The enormous demand for compute power from AI training and inference provides real use cases and revenue sources for DePIN projects. When centralized compute supply is constrained by power bottlenecks or capex cycles, decentralized networks’ relative value may become more pronounced. Goldman Sachs notes that in some core markets, the queue for connecting data centers to the grid is as long as 8–12 years—much longer than the GPU upgrade cycle—creating structural bottlenecks and differentiated opportunities for DePIN.
Second, AI Agents are increasingly integrating with crypto sectors. In Q1 2026, "AI Agent tokens" underwent a steep 80–90% correction. But this drop was selective—tokens with "AI" in their name but no real utility collapsed, while projects with genuine use cases stabilized and rebounded. This differentiation mirrors the shift in AI stocks from "concept-driven" to "performance validation." As the cost of AI chip infrastructure falls, deploying AI Agents becomes easier, which will benefit AI application projects within the crypto ecosystem.
Third, there are capital and sentiment linkages between AI stocks and AI crypto assets. On July 7 (Beijing time), AI-related US stocks such as AAOI, MRVL, AVGO, and ASML gained between 1.63% and 3.73%. During the same period, Bitcoin’s price hovered around $64,000. The semiconductor sell-off in early July briefly dragged Bitcoin down to about $62,000—highlighting that the health of AI trading has become a relevant upstream indicator for digital asset markets.
From a long-term perspective, the relationship between AI stocks and AI crypto assets isn’t simply correlated or inverse. When AI chip stocks correct due to high valuations, some capital may seek AI crypto assets as alternative exposure. Conversely, when AI infrastructure investment expands, the underlying value of DePIN and decentralized compute sectors is reinforced. There are complex capital flows and value transmission mechanisms between the two, warranting ongoing attention.
Conclusion
The early July 2026 pullback in AI chip stocks was essentially a concentrated correction of excessive gains from the first half of the year. The disagreement between JPMorgan and Morgan Stanley isn’t about the long-term AI trend, but rather about short-term timing and valuation levels. JPMorgan sees structural supply shortages and a continuing upcycle; Morgan Stanley sees fading momentum and a window for rotation trades.
For investors, the key question is whether the AI investment cycle is shifting from "concept-driven" to "performance validation." If so, the market narrative will move from "who has an AI story" to "who has AI revenue." Earnings reports from hardware giants like Samsung and Micron have demonstrated the possibility of performance validation, but the visibility of capital expenditure returns for hyperscale cloud providers remains the biggest uncertainty.
AI’s impact on the crypto market is also moving from a conceptual phase to a substantive one. The growth of DePIN’s market cap, differentiation in the AI Agent sector, and capital linkages between AI stocks and crypto assets all indicate that AI-crypto integration is becoming a structural trend that cannot be ignored. Regardless of short-term fluctuations in AI chip stocks, the long-term expansion of AI infrastructure remains unchanged—and its profound impact on the crypto ecosystem is only beginning to emerge.
FAQ
Q1: What is the fundamental disagreement between JPMorgan and Morgan Stanley regarding AI chip stocks?
Both banks are bullish on AI’s long-term prospects, but differ on short-term tactics. JPMorgan sees the current pullback as a buying opportunity, expecting the semiconductor upcycle to last at least until 2028. Morgan Stanley believes chip stock momentum is fading, recommends taking profits, and shifting toward hyperscale cloud providers. The core difference lies in their assessment of "whether current valuations are reasonable" and "which sector will lead the next rally."
Q2: What are the main reasons for the current pullback in the AI chip sector?
There are four key factors: excessive gains in the first half (SOX up over 80%), concentrated profit-taking; Meta’s announcement of selling excess compute, raising concerns about overbuilding AI infrastructure; Citi and others questioning the visibility of cloud providers’ capex returns; and elevated valuations with little margin of safety. These factors are mostly about sentiment and valuation corrections, not fundamental changes in the AI industry.
Q3: Is the AI investment cycle entering its second phase?
Yes. The first phase (2023 through the first half of 2026) was driven by AI concepts and future expectations, with broad gains for AI-themed stocks. The current market is entering the second phase, characterized by greater focus on realized earnings, clear differentiation among AI-related companies, and real revenue and order growth becoming the main drivers of stock performance. Samsung’s Q2 operating profit up 1,810% year-over-year and similar earnings reports confirm performance validation in hardware.
Q4: How will fluctuations in AI chip prices impact the crypto market?
The impact is transmitted through three main channels: AI infrastructure expansion provides real use cases for DePIN (decentralized compute networks), with DePIN’s market cap reaching about $9–10 billion; AI Agent and other crypto sectors are experiencing differentiation similar to AI stocks, with projects that have real utility proving more resilient; and there are capital and sentiment linkages between AI stocks and crypto assets, with semiconductor sell-offs dragging down Bitcoin.
Q5: Is now a good time to invest in AI chip stocks?
It depends on your investment horizon and risk tolerance. JPMorgan advises long-term investors to use the pullback as an entry point, believing the semiconductor upcycle is far from over. Morgan Stanley recommends waiting for a better entry, as the correction may have further room to run. Investors should watch upcoming earnings reports from Nvidia, Micron, and others to assess whether AI demand can sustain the current cycle.
