Bernstein dismantles the congestion of AI hardware transactions: chips and network devices become concentrated capital storage stocks diverge

The Zhitong Finance App learned that as the boom in artificial intelligence (AI) infrastructure investment continues to heat up, stocks related to semiconductors, servers, and network equipment have become sought after by the market, but investors' concerns about overcrowded AI concept stock trading are also increasing. According to Bernstein's latest research report, although some AI-related stocks have recently shown signs of capital withdrawal, the overall trading congestion level of the US technology hardware, semiconductor and network equipment industries is still high. Among them, stocks such as AMD (AMD.US), Arista Networks (ANET.US), and Applied Materials (AMAT.US) are particularly crowded, while stocks such as Qualcomm (QCOM.US) and SNDK.US (SNDK.US) are relatively low.

Bernstein said in a report released on October 25 that as investors increasingly recognize AI infrastructure suppliers as direct beneficiaries of the wave of artificial intelligence development, capital continues to pour into related stocks, driving a marked increase in transaction congestion in various technology and hardware segments.

According to the agency's quantitative analysis, communications equipment is currently the most heavily traded segment in the global technology, media and telecommunications (TMT) industry, followed by semiconductors and electronic devices. It is worth noting that in history, the network equipment and traditional hardware industry was often not favored by investors, but now it is a field with a high concentration of market capital, reflecting the profound impact of AI investment themes on the industry's capital allocation pattern.

However, Bernstein stressed that although the technology sector is currently the most crowded trading sector in various industries in the world, from a historical perspective, its level of congestion has not reached an unusually extreme level. In contrast, the current level of overcrowding in the financial and industrial sectors is more prominent compared to their respective historical levels.

In the semiconductor sector, out of the 12 semiconductor and semiconductor equipment stocks covered by Bernstein, 7 are in the top 10% range with the highest trading congestion. Among them, the AI chip sector is the most crowded, mainly involving Broadcom (AVGO.US), AMD, and Nvidia (NVDA.US). The semiconductor equipment sector is followed by companies such as Applied Materials, Fanlin Group (LRCX.US), KLAC.US (KLAC.US), and PDF Solutions (PDFS.US).

In terms of individual stocks, AMD is one of the most heavily traded companies in technology hardware and semiconductor stocks covered by Bernstein, and its standardized congestion score reached 2.35. Arista Networks and Lumentum (LITE.US) reached 2.20 and 1.91, respectively, and the congestion level of Coherent (COHR.US) and applied materials was also at a high level.

In contrast, Qualcomm became the most notable exception in the semiconductor sector. According to the report, since the beginning of this year, Qualcomm has been in the 10% range with the lowest trading congestion. This situation only changed briefly from April to May due to news related to the data center business driving up stock prices.

According to Bernstein's analysis, Qualcomm is not only one of the least crowded semiconductor stocks in its coverage area, but also at a low level among the world's largest technology hardware and semiconductor companies.

There is also significant differentiation in the field of semiconductor equipment. Bernstein pointed out that since June, Kelei's trading congestion level has been significantly lower than that of peers such as Applied Materials, Fanlin Group, and PDF Solutions.

The agency believes that this is mainly related to the relative lack of attractiveness of Ke Lei for short-term investment. Due to the long product delivery cycle, Kelei's growth in 2026 is expected to be slower than that of some of its peers. At the same time, the increase in DRAM costs used in its products may put pressure on gross margins by about 100 basis points. Furthermore, Kelei's exposure to the memory chip industry chain is relatively limited, and the valuation is higher than Applied Materials and Fanlin Group, prompting some investors to switch to other semiconductor equipment companies.

In the field of technology hardware, trading congestion for stocks related to AI servers and storage devices is also at a high level. According to Bernstein data, Apple (AAPL.US), Western Digital (WDC.US), Seagate (STX.US), and Dell Technology (DELL.US) are all in the 10% range with the highest transaction congestion, while HP (HPQ.US) and ultra-micro computers (SMCI.US) are in the less crowded range.

Among them, Apple is one of the most continuously crowded stocks in Bernstein's coverage area. Since the beginning of 2024, the stock has almost remained in the top two most congested ranges.

Bernstein pointed out that although rising memory chip costs have raised market concerns about Apple's profitability, the level of transaction congestion has not declined significantly. The agency believes that rising DRAM and NAND costs may pose a downside risk to Apple's gross margin and earnings per share for the December quarter. If the upcoming results fall short of expectations, highly concentrated investor positions may amplify short-term fluctuations in stock prices.

However, Bernstein still maintains Apple's “outperforming the market” rating. He believes that Apple is expected to become an important entry point for consumers to use AI services, while not bearing the huge AI capital expenses faced by some large cloud computing and software companies.

In the storage industry, Western Digital has been in the 10% most crowded trading range since June 2025, and is one of the stocks that has remained highly congested for the longest time in Bernstein's coverage area.

Seagate's trading congestion also increased markedly in the second half of this year. As market acceptance of the hard disk drive (HDD) investment logic increased, the stock continued to be in the most crowded zone since July. Despite this, Bernstein listed Seagate as one of his preferred stocks and maintained an “outperforming market” rating. According to the agency, Seagate's near-line hard drive production capacity for the 2026 calendar year has been sold out, and most of the relevant production capacity as of 2028 has received order commitments.

Furthermore, Seagate's leading edge in heat-assisted magnetic recording (HAMR) technology is also an important reason why Bernstein is optimistic about its profit prospects. The agency predicts that by 2027, HAMR products will account for about 70% of Seagate's near-line hard drive shipment capacity, while Western Digital's corresponding proportion is expected to be only about 5%.

The stark contrast to these popular storage stocks is SanDisk. Bernstein pointed out that although SanDisk is a direct beneficiary of the NAND flash industry's upward cycle, and its stock price has risen sharply since this year, its trading congestion level is at a low level. The agency believes that this may reflect the market's underestimation of SanDisk's profitability and the continuation of the current storage industry boom cycle.

Bernstein pointed out that the market still tends to view SanDisk as a typical cyclical transaction target, and is concerned that the increase in the average sales price of NAND products may slow down. However, SanDisk's new long-term supply agreement with customers is expected to mitigate this risk.

According to the report, more than half of SanDisk's memory chip shipments in FY2027 and about two-thirds of FY2028 shipments have been included in long-term agreements with upper and lower price limits. The agreements involved eight data center and edge computing customers and received a total of $16.5 billion in financial guarantees.

Bernstein believes that these arrangements will help improve the predictability of SanDisk's future revenue and profits. Coupled with the company's own operating leverage effect, its profit growth potential may not yet be fully reflected in the market.

However, the agency also warned that since SanDisk's transaction history is short after re-listing, and some congestion indicators require at least two years of historical data, the reliability of SanDisk's congestion assessment is currently relatively limited.

In the field of network equipment, out of the 7 stocks covered by Bernstein, 5 are in the 10% range with the highest trading congestion. Among them, companies such as Arista Networks, Lumentum, and Coherent are all targets with concentrated capital allocation. Bernstein believes that although these companies are still expected to benefit from increased demand for AI infrastructure, if the market cools down in a new round of popular transactions, their stock prices may face greater pullback pressure.

Celestica (CLS.US) is showing a different trend. The report points out that although the overall trading congestion level of this stock is still higher than the average for the technology hardware and semiconductor industry, it has declined markedly since the summer of this year. Currently, it is a relatively low level of congestion among the network equipment stocks covered by Bernstein.

Bernstein believes that as market position congestion decreases and the valuation multiples fall somewhat, Celestica may have more favorable investment conditions before the upcoming Investor Day.

In terms of investment ratings, Bernstein maintained “outperforming the market” ratings for companies such as AMD, Nvidia, Broadcom, Applied Materials, Apple, Western Digital, Seagate, Dell, and SanDisk, while maintaining “on par with the market” ratings for companies such as Qualcomm and Huipu Pu Ultramicrocomputers.

Bernstein stressed that crowded transactions do not mean that the fundamentals of related companies have deteriorated, nor does it necessarily indicate that stock prices are about to fall. However, when a large number of investors concentrate on holding similar AI investment targets, once market expectations change, concentrated position reduction may amplify stock price fluctuations.

The agency believes that as the AI infrastructure investment theme continues to attract capital, investors need to pay more attention to the degree of congestion in individual stock positions, use it as an important reference indicator for risk management, and be wary of potential pullback risks brought about by the concentration of popular transactions while evaluating the growth potential of enterprises.