On July 23, Goldman Sachs made it clear that the current semiconductor rally is likely to continue. The backdrop for this view is that, since mid-June, hedge funds have unwound roughly 80% of their net buying in global semiconductor and semiconductor equipment stocks accumulated since the start of the year, resulting in a significant reduction of previously crowded positions. Over the past couple of days, capital has started flowing back in, and Goldman Sachs believes this demand may persist.
The most concentrated buying has focused on the two segments that saw the sharpest corrections—memory chips and semiconductor equipment. The most notable recent buying has occurred in memory stocks such as Seagate Technology (STX), Western Digital (WDC), Micron Technology (MU), and SanDisk (SNDK), as well as equipment stocks like Applied Materials (AMAT), ASML Holding (ASML), and Lam Research (LRCX).
This capital movement echoes recent market trends. After AI and chip stocks faced sustained pressure, memory stocks like Micron, SanDisk, and Seagate rebounded strongly, and the Philadelphia Semiconductor Index (SOX) also posted a robust recovery. Following a phase of "aggressive deleveraging" in the semiconductor sector, why is capital prioritizing memory and equipment for re-entry? What key variables will determine the sustainability of this rebound? This article systematically analyzes these questions using Goldman Sachs’ latest positioning data, memory price trends, and the logic behind AI capital expenditures.
Semiconductor Positioning: Extreme Crowding Has Eased, Rapid Deleveraging Mostly Complete
To understand the logic behind the current capital inflows, it’s important to clarify the root causes of the previous correction in the semiconductor sector.
In the first half of 2026, AI chip stocks surged, with capital highly concentrated in the semiconductor sector. According to Goldman Sachs, the net allocation to global semiconductor and semiconductor equipment stocks as a percentage of the global Prime Book was about 10% at the start of the year, rising to a historic high of roughly 24% in June. For US semiconductor and equipment stocks, net allocation climbed from around 7% at the start of the year to about 14% in June.
Such extreme concentration means that any marginal shift in market sentiment can trigger a concentrated wave of profit-taking. The semiconductor sector then underwent a dramatic deleveraging process—hedge funds reduced their holdings on a large scale, and previously crowded long positions were heavily unwound.
As of July 23, net allocation to global semiconductor and equipment stocks has fallen from the June peak to about 19%, but remains at the 84th percentile over one year and the 97th percentile over five years. US semiconductor and equipment stocks’ net allocation has dropped to around 11%, corresponding to the 79th percentile over one year and the 96th percentile over five years.
Goldman Sachs interprets these data as signaling that semiconductor positioning has not yet reached "cheap" levels, but rapid deleveraging is largely complete. This suggests that, as long as upcoming earnings reports and AI capital expenditure guidance do not deteriorate further, some capital will likely favor re-entering segments that were hardest hit but still have fundamental support.
From a broader perspective, Goldman Sachs previously noted that investors are underweight US tech stocks, especially the "Big Seven," and are instead favoring the semiconductor sector, which stands to benefit from AI investment. This medium- to long-term allocation logic remains unchanged despite short-term adjustments.
Memory Chips: "Rigid Demand" Driven by AI Data Center Expansion
Memory chips have become the first segment to attract capital for re-entry, supported by clear industry logic.
The expansion of AI data center computing power requires not only GPUs and other compute chips, but also matching storage capabilities. As AI models grow larger, demand for computing power drives an increase in GPU numbers, which simultaneously boosts demand for high-bandwidth memory (HBM), DRAM, and NAND flash.
This logic is translating into tangible results. The memory industry is highly cyclical, and the central debate in the market is whether AI demand will permanently alter the industry’s demand curve. Current data indicate that the supply-demand balance for memory chips remains tight. In Q2 2026, LPDDR4 and LPDDR5 prices have doubled compared to Q4 2025, with single-quarter increases of about 58% to 63%. As of July 21, DDR5 16G (2Gx8) 4800/5600 prices broke the $50 mark for the first time, reaching $50.07; DDR4 16Gb (2Gx8) 3200 prices rose to $81.20, both setting new record highs.
Supply-side constraints are also evident. SK Hynix Chairman Chey Tae-won recently stated that AI semiconductor demand next year is expected to grow by 60% to 100% year-over-year, with overall memory semiconductor demand rising by 50% to 60%. He also warned that the AI-driven supply-demand gap in semiconductors will widen further next year, putting continued upward pressure on memory prices.
The lead analyst for key components at CITIC Securities highlighted that the sharp rise in memory chip prices and the persistent AI data center demand are crowding out advanced flash and memory production capacity, creating an industry-wide supply shortage that is likely to persist until at least 2028. Morgan Stanley shares a similar view, expecting tight memory chip supply-demand conditions to last through 2028.
In terms of market performance, memory stocks have posted the strongest rebound. On July 21 (UTC+8), US memory stocks continued to climb: Micron Technology rose 10.17%, SK Hynix gained 11.18%, SanDisk jumped 11.86%, Western Digital surged 12.67%, and Seagate Technology advanced 10.42%. Goldman Sachs simultaneously raised price targets for several memory giants, boosting SanDisk’s target from $1,200 to $2,200 and Western Digital’s from $400 to $650.
Semiconductor Equipment: The "Pick-and-Shovel" Logic of Fab Expansion
Semiconductor equipment has become the second major segment for capital re-entry, backed by a clear rationale—chip manufacturing expansion drives increased fab investment and, in turn, equipment orders.
The surge in AI chip demand is propelling global fabs into a new expansion cycle. Whether for logic chips, memory chips, or power semiconductors, all capacity expansions rely on core manufacturing equipment such as lithography machines, etchers, and thin-film deposition tools. ASML, as the world’s sole supplier of extreme ultraviolet (EUV) lithography systems, holds an irreplaceable position in this expansion cycle.
Goldman Sachs’ trading desk has observed notable buying in equipment stocks—Applied Materials, ASML, and Lam Research all made the list. These three companies cover different key segments of semiconductor manufacturing: Applied Materials specializes in materials engineering and deposition equipment, Lam Research focuses on etching and deposition, and ASML dominates the advanced lithography market.
Fundamentally, equipment stocks remain supported by strong performance. ASML’s net income in Q2 2026 reached €2.92 billion, reflecting robust profitability. Applied Materials’ share price has soared 187.59% over the past 52 weeks. Analysts rate Applied Materials a "strong buy," with an average target price of $623.06, offering about 12.5% upside from current levels.
Goldman Sachs previously raised Applied Materials’ target price from $520 to $645. This upward revision reflects the sell-side’s optimistic outlook for semiconductor equipment demand—the fab expansion cycle is not yet over, and equipment order growth is expected to continue.
Can the Rebound Continue? Three Key Variables to Watch
Goldman Sachs’ recent commentary highlights three critical factors that will determine whether the semiconductor rally can be sustained.
First, will cloud providers continue to raise AI capital expenditure guidance? The capex plans of Alphabet, Meta, Microsoft, and Amazon serve as the "ceiling" indicator for semiconductor demand. Latest data show Alphabet has raised its 2026 capex guidance from $180–190 billion to $195–205 billion, with the midpoint now at $200 billion, and expects significant growth in 2027. Combined, the four cloud giants may spend over $725 billion in capex in 2026. As long as this upward trend persists, downstream demand for semiconductors remains fundamentally secure.
Second, can memory prices and orders support earnings expectations? The gains in memory stocks like Micron and SanDisk ultimately need to translate into performance. While memory chip prices are still rising, the market needs to see sustained orders and price stability. If memory prices soften or order growth slows, valuations for memory stocks will face renewed scrutiny.
Third, can equipment stock orders confirm that AI data center expansion is ongoing? Equipment stock performance is the most direct "evidence" of fab expansion. If companies like ASML and Applied Materials continue to deliver order data and guidance that exceed expectations, it will confirm that AI data center expansion is still progressing, providing fundamental support for the entire semiconductor sector.
Conclusion
The semiconductor sector has undergone a dramatic adjustment from extreme crowding to rapid deleveraging, with hedge funds unwinding about 80% of net buying since the start of the year. While positioning has not yet returned to "cheap" levels, the extreme crowding has clearly eased. Against this backdrop, capital is starting to flow back into the two segments that saw the sharpest corrections—memory chips and semiconductor equipment.
Memory chips benefit from the rigid demand for HBM, DRAM, and NAND driven by AI data centers, combined with supply-side capacity constraints, keeping prices in an upward channel. Semiconductor equipment is supported by the ongoing fab expansion cycle, with a clear trend of rising equipment orders.
However, the sustainability of the rebound remains uncertain. Whether cloud providers continue to raise AI capex, whether memory prices and orders can support earnings expectations, and whether equipment stock orders confirm ongoing AI data center expansion—these three variables will determine if this semiconductor rally is merely a short-term technical bounce or a fundamental revaluation.
As of July 23 (UTC+8), in the crypto market, Bitcoin was quoted at $66,188.99, down 0.32% over 24 hours; Ethereum was at $1,947.90, up 1.31%. The total global cryptocurrency market cap stood at approximately $2.33 trillion. While the semiconductor rally and crypto market volatility belong to different asset classes, both reflect a repricing process for global risk assets amid macro uncertainty. For investors, understanding the logic behind capital flows across sectors offers greater long-term value than chasing short-term price swings.
FAQ
Q: Why have semiconductor stocks rebounded recently?
Previously, sustained gains in AI chip stocks led to highly concentrated semiconductor positioning, followed by profit-taking and adjustments as institutional capital unwound positions. With hedge funds giving back about 80% of net buying, extreme crowding has eased. Goldman Sachs believes rapid deleveraging is mostly complete, and some capital is starting to re-enter segments with fundamental support.
Q: Why are investors prioritizing memory chip stocks?
AI data center expansion requires not only GPUs but also matching memory chips like HBM, DRAM, and NAND. Memory chip prices remain in an upward channel, with DDR5 and DDR4 prices hitting record highs. The supply-demand gap is expected to persist at least until 2028.
Q: What is the investment logic for semiconductor equipment stocks?
Chip manufacturing expansion drives increased fab investment, which in turn boosts equipment orders. Industry leaders like ASML, Applied Materials, and Lam Research are core beneficiaries of this expansion cycle. As long as AI data center expansion continues, demand fundamentals for equipment stocks remain solid.
Q: Can the semiconductor rally continue?
Goldman Sachs points to three key variables: whether cloud providers (Alphabet, Meta, Microsoft, Amazon) continue to raise AI capex; whether memory prices and orders can support earnings expectations; and whether equipment stock orders confirm ongoing AI data center expansion. All three are essential.
Q: What is the supply-demand situation for memory chips?
Memory chip supply remains tight, with AI data center demand continually crowding out advanced flash and memory production capacity. SK Hynix expects AI semiconductor demand to grow 60% to 100% year-over-year next year. TrendForce forecasts DRAM prices to rise 13% to 18% quarter-over-quarter in Q3, and NAND to increase 10% to 15%.
