花旗解读：设备牛市看到2500亿美元，真正考验在2027年

TL;DR

• According to market reports, Citigroup projects that global WFE under a bull case scenario could rise to $250 billion by 2028.
• TSMC, Samsung, and Intel are expected to account for approximately 55% of global WFE spending in 2025, with their financial guidance determining the potential for upward revisions.
• The benefits for equipment stocks still depend on the sustainability of AI demand, the realization of Samsung's investments, and Intel's foundry progress.

TSMC, Intel, and Samsung are set to release their second-quarter earnings results in mid-to-late July, presenting a key test of capital expenditure expectations for semiconductor equipment stocks. According to market reports, Citigroup remains bullish on wafer fab equipment spending ahead of earnings season, believing that AI/HPC demand is driving up investments in advanced nodes, memory, and foundry. For investors, WFE (Wafer Fab Equipment) spending covers procurement of key equipment such as lithography, etching, deposition, and testing, directly impacting orders and revenue for equipment companies like Applied Materials, Lam Research, and Teradyne.

Equipment Cycle Focus Shifts from 2026 to the Following Two Years

The previous rally in semiconductor equipment stocks was primarily driven by investment expectations related to AI servers, advanced packaging, HBM, and advanced logic nodes. The market is now questioning whether capital expenditure can be revised upward from 2026 and extend into continued volume growth in 2027 and 2028.

According to market reports, Citigroup's bull case for WFE projects approximately $145 billion in 2026, $200 billion in 2027, and $250 billion in 2028. Brokerage report estimates also indicate that TSMC, Samsung, and Intel collectively account for about 55% of global WFE spending in 2025. If these three companies maintain or increase their medium-to-long-term capital expenditure, there is room for the equipment cycle to continue rising.

The upcoming earnings calls will provide more direct clues. TSMC is scheduled to report earnings on July 16, and Intel will disclose its results after market close on July 23. Samsung released its Q2 earnings guidance on July 7 and will hold its earnings conference call on July 30 at 10:00 KST. The market will not only watch quarterly revenue and profit but will also focus on capital expenditure guidance, advanced node demand, memory investment pace, and management's statements regarding AI demand over the next three years.

The transmission chain for equipment companies is relatively clear. When wafer fabs increase capital expenditure, equipment companies benefit first from orders and shipments. If demand continues to be tight, equipment makers also have opportunities to support gross margins through improved product mix and higher capacity utilization. Equipment-related companies mentioned in the Citigroup report include Applied Materials (AMAT), Lam Research (LRCX), Teradyne (TER), and AEIS, but the elasticity of these individual stocks still depends on the procurement pace of their customers.

TSMC is the Strongest Anchor, with 2027 Assumptions Significantly Above Consensus

TSMC remains the most critical entity in this AI capital expenditure cycle. During its April earnings call, TSMC confirmed its 2026 capital expenditure guidance of $52 billion to $56 billion and indicated spending trends towards the higher end of the range. The market expects the company to likely maintain the 2026 guidance in its upcoming earnings report and continue to emphasize demand for advanced nodes and advanced packaging.

The bigger focus is on the following two years. Citigroup's model shows TSMC's capital expenditure at $75 billion in 2027 and $80 billion in 2028, corresponding to year-over-year growth rates of 36% and 7%, respectively. These assumptions are higher than market consensus, with the gap being more pronounced, especially for 2027.

The core support for this judgment is the continued demand from AI/HPC driving expansion in advanced nodes. TSMC handles AI chip demand from Nvidia, AMD, Broadcom, and others, and also benefits from advanced packaging, CoWoS, and migration to more advanced nodes. As long as AI chip orders remain strong, TSMC has the incentive to continue purchasing more front-end and back-end equipment.

However, high capital expenditure does not mean the equipment cycle is locked in. Whether the optimistic model for 2027 and 2028 can be achieved still depends on the sustainability of AI orders, the pace of customer chip development, the easing of advanced packaging bottlenecks, and whether equipment delivery cycles can keep up.

Samsung and Intel Bring Incremental Gains, Along with Uncertainty

If TSMC provides the base for the equipment cycle, Samsung and Intel determine the upside potential.

Samsung stated on its April call that AI demand will drive significant year-over-year growth in capital expenditure. Citigroup's model shows Samsung's semiconductor capital expenditure maintaining relatively high growth rates from 2026 to 2028. This involves two tracks: HBM and high-end DRAM demand driving memory investment, while advanced logic and foundry business determine whether Samsung can continue to catch up with TSMC in more advanced nodes.

Samsung's long-term investment plans also amplify the imagination of equipment demand. There are discrepancies in public sources regarding specific figures, with Samsung's press releases and media reports covering different scopes such as the group's total domestic investment, Samsung Electronics' future business plans, and semiconductor cluster investments. A more prudent statement is that Samsung's semiconductor-related investment scale in South Korea over the next ten-plus years is estimated to be around 2,000 trillion won or more. This long-term plan spans many years; how much translates into specific equipment procurement in the short term depends on the pace of specific fab construction, equipment installation, and capacity ramp-up.

Intel's situation is more complex. During its Q1 earnings call, the company adjusted its 2026 capital expenditure outlook from "flat to down" to "flat" and indicated that tool and equipment-related spending would grow approximately 25% year-over-year. In Citigroup's model, Intel's capital expenditure assumptions for 2027 and 2028 still have upside, with greater elasticity in 2028.

Whether Intel can realize this incremental growth depends critically on its foundry business. The verification of the 18A process, customer decisions for 14A, and potential collaborations with major customers will all affect the intensity of subsequent investments. If progress with advanced node customers falls short of expectations, it will be difficult for capital expenditure to materialize according to the optimistic scenario. If substantial progress is made in the foundry transformation, Intel could become a significant source of incremental growth for global WFE.

Micron Validates Memory Demand, but Cannot Replace the Guidance of the Big Three

Capital expenditure by memory manufacturers is also providing validation for the equipment cycle. Micron has raised its FY2026 capital expenditure guidance to approximately $27 billion. The company also indicated that its quarterly capital expenditure level for FY2027 is expected to be higher than the approximately $10 billion level seen in FY4Q26. If this quarterly level continues, full-year FY2027 capital expenditure could exceed $40 billion.

This demonstrates that the HBM, high-end DRAM, and memory demand driven by AI servers is not just a story for the logic node side. Memory expansion will also drive up equipment procurement, particularly benefiting segments related to deposition, etching, testing, and packaging. According to reports, Micron's long-term US investment plan has also been raised to over $250 billion, with the timeline extending to around 2035.

However, Micron serves more as corroborating evidence for memory demand and cannot replace the guidance of TSMC, Samsung, and Intel. These three companies collectively account for approximately 55% of global WFE spending in 2025. What truly determines the height of the equipment cycle is their statements regarding capital expenditure for 2026 to 2028 in the coming quarters.

The $250 Billion Assumption Hinges on Post-2027

The biggest divergence in this earnings season preview is that Citigroup's assumptions for 2027 and 2028 are significantly more optimistic than the market. The upward revision for 2026 is relatively easier to understand, as AI demand is already reflected in orders and capacity expansion. For the period after 2027, a continued significant increase in capital expenditure requires more conditions to materialize simultaneously.

AI/HPC demand needs to remain strong and not just represent short-term concentrated procurement by cloud vendors. TSMC's expansion of advanced nodes and advanced packaging needs sustained support from customer orders. Samsung's massive long-term investment plan must translate into specific equipment spending, rather than just remaining at the stage of fab construction and long-term planning. Intel's foundry business must also prove that its 18A and 14A nodes have sufficient customers and production prospects.

Equipment delivery cycles, the macroeconomic environment, and semiconductor cycle fluctuations will also affect actual spending. Capital expenditure plans can be revised upward, but they can also be delayed due to changes in customer demand, declining capacity utilization, or financing pressures.

The story of equipment stocks still revolves around the three major wafer fabs. If earnings continue to release strong signals regarding capital expenditure, the global WFE bull case scenario will gain more support. If management turns cautious in their statements regarding 2027 and 2028, market expectations for $250 billion in equipment spending will need to be discounted. The current point of contention is not whether AI capital expenditure exists, but whether this expansion cycle can extend beyond 2026.