Goldman Sachs Research Report Interpretation: Cloud providers’ capex adjustments are transmitted to the equipment side, benefiting semiconductor equipment manufacturers like ASML directly.

CN
1 hour ago
The chain of AI capital expenditure is gradually strengthening, with cloud vendors, wafer fabs, and equipment manufacturers each confirming demand.

Written by: Rita

The signals conveyed by the financial reports of the four major cloud vendors in the second quarter are very consistent, indicating that capital expenditures are still on the rise. Amazon raised its cash capex guidance for 2026 from $200 billion to $220 billion, Google increased its forecast from $180 billion - $190 billion to $195 billion - $205 billion, and Meta narrowed its expected range to $130 billion - $145 billion. TSMC raised its full-year revenue guidance from "growing over 30%" to "slightly above 40%", and capital expenditures from $52 billion - $56 billion to $60 billion - $64 billion. Goldman Sachs pointed out in its research report on July 31 that the transmission chain from cloud vendors to wafer fabs and then to equipment manufacturers is strengthening, and semiconductor equipment manufacturers such as ASML are direct beneficiaries.

Goldman Sachs outlined the benefit logic for five relevant companies in Europe. ASML benefits from EUV and advanced DUV demand, ASMI benefits from advanced logic and DRAM investment-driven ALD and epitaxial demand, and BESI benefits from the proliferation of custom chips and investment in advanced packaging. Nebius is benefiting from the widening gap in supply and demand for computing power, while Technoprobe is benefiting from increased accelerator volume and heightened testing intensity. Goldman Sachs' judgment is that the scale of AI capital expenditures is expanding, the number of custom chips is increasing, advanced packaging is being upgraded, and equipment manufacturers can gain a share from multiple dimensions.

Cloud Vendors' Capex Adjusted Upwards, Custom Chip Orders Lock in Long-term Demand

Microsoft's fourth quarter Copilot paid seats surpassed 30 million, with a net increase of 10 million in a single quarter, double that of the previous quarter. Azure revenue exceeded $100 billion, representing a 41% year-on-year increase. Goldman Sachs believes that enterprise AI adoption is translating into actual revenue, which is the prerequisite for sustained expansion of infrastructure investment.

Amazon has raised its cash capex guidance for 2026 from $200 billion to $220 billion, expecting available power capacity to double by the end of 2027 compared to 2025. The Trainium chip has secured multi-year commitments worth several gigawatts from Anthropic and OpenAI. AWS backlog has reached $496 billion, a year-on-year increase of over 100%.

Google Cloud's revenue accelerated to 82%, with a backlog nearing $514 billion. Capex guidance for 2026 has been raised to $195 billion - $205 billion, with management reiterating that spending will "significantly increase" in 2027. Meta narrowed its 2026 capex guidance to $130 billion - $145 billion and is working with BlackRock to develop a 1GW data center.

Goldman Sachs specifically emphasized the trend of custom chips. Amazon's Trainium has received multi-year commitments from major customers, Google's TPU has directly contributed to backlog growth, and Microsoft is also expanding its self-developed chip deployments. Custom chips will not replace Nvidia but will create new incremental demand for advanced packaging and testing equipment.

Wafer Fabs and Memory Suppliers Confirm Tight Supply; WFE Estimates Continue to be Upgraded

TSMC has raised its revenue guidance for 2026 from "growing over 30%" to "slightly above 40%", with HPC and AI being the main driving forces. Capital expenditures have increased from $52 billion - $56 billion to $60 billion - $64 billion, and investments over the next three years will be "significantly higher" than the previous cycle. Spending is focused on N2 and subsequent nodes, advanced packaging, and the mass production of A14 in 2028.

Samsung's management expects that memory supply tightness will worsen and continue until 2028. Long-term agreements have already covered the five largest data center customers globally, with more negotiations ongoing that could ultimately cover 60% to 70% of planned capacity. Goldman Sachs believes that the increase in LTA coverage improves the visibility of future investment demand.

WFE (Wafer Fab Equipment Spending) estimates have also been upgraded. Lam Research has raised its 2026 WFE estimate from $140 billion to a range of $150 billion, and KLA has similarly adjusted upwards to over $150 billion, believing that $190 billion is a reasonable level for 2027. Goldman Sachs' U.S. team estimates WFE for 2026/27/28 will be $141 billion, $186 billion, and $208 billion, respectively. Goldman Sachs believes that China's WFE will level off or grow slightly in 2026, not serving as a drag.

ASML, ASMI, BESI Directly Benefit; Nebius and Technoprobe Also in the Transmission Chain

Goldman Sachs has analyzed the transmission logic for the five companies individually.

ASML: TSMC's investment upgrades and memory capacity expansion are driving EUV and advanced DUV demand. Advancing from N2 to smaller nodes requires more lithography layers with each generation. Goldman Sachs also specifically addressed market concerns about new entrants in the lithography field. A long-established competitor with decades of experience in DUV dry lithography has failed to successfully transition to immersion DUV; the difficulty of producing immersion DUV systems is much greater than the market imagines, let alone EUV.

ASMI: Advanced logic and DRAM investment support ALD and epitaxial demand. Gate-all-around (GAA) technology is expanding the application range of ALD, and memory manufacturers are increasing capacity. TSMC's long-term node roadmap and Samsung's continuous expectations for memory tightness provide visibility for growth in process intensity.

BESI: The proliferation of custom chips and investment in advanced packaging are expanding the range of equipment applications. TSMC is increasing advanced packaging capacity, and self-developed accelerators such as TPU, Trainium, and Maia are ramping up. Chiplet architectures and advanced packaging necessitate that more chips pass through processes like backside connections and hybrid bonding, and BESI's assembly product portfolio encompasses these stages.

Nebius: Cloud vendors continue to expand capacity, and the gap between supply and demand for computing power is widening. Demand is spreading from large model developers to enterprise-grade workloads, and Nebius is increasing capacity while acquiring more corporate clients.

Technoprobe: The ramp-up of foundry and memory capacity is driving accelerator volume, which in turn increases demand for probe cards. The more custom chips there are, the higher the customization demand for probe cards. Chiplet architectures and advanced packaging have intensified testing strength, making it increasingly important to identify faulty chips before packaging; probe cards are a core element of this process.

Capex for cloud vendors continues to rise, revenue and capital expenditure guidance from wafer fabs are being upgraded, memory manufacturers believe tight supply will persist until 2028, and WFE estimates are also being revised upward. Goldman Sachs' conclusion is that the transmission chain of AI capital expenditure is strengthening gradually, with each link from cloud vendors to wafer fabs to equipment manufacturers confirming demand. ASML, ASMI, and BESI are strategically positioned in the key links of lithography, deposition, and packaging testing, while Nebius and Technoprobe are also capturing incremental demand in their respective niches. The market's focus on AI chips is concentrated on Nvidia and custom chips, which may underestimate the sustainability and breadth of the benefits to upstream equipment segments.

Disclaimer: This article is a summary and interpretation of third-party brokerage research reports (Goldman Sachs, July 31, 2026) by ChaoXiang Research, combined with publicly available market information. The ratings, target prices, profit forecasts, and related judgments mentioned in the text are the perspectives of the respective brokerage's analysts, representing only their institutions' views and do not represent the opinions of ChaoXiang Research, nor do they constitute any investment advice. The market carries risks, and decisions should be made independently. This article should not serve as a basis for buying or selling any securities.

免责声明:本文章仅代表作者个人观点,不代表本平台的立场和观点。本文章仅供信息分享,不构成对任何人的任何投资建议。用户与作者之间的任何争议,与本平台无关。如网页中刊载的文章或图片涉及侵权,请提供相关的权利证明和身份证明发送邮件到support@aicoin.com,本平台相关工作人员将会进行核查。

Share To
APP

X

Telegram

Facebook

Reddit

CopyLink