Six key issues in the early stage of the AI ​​​​big cycle

# Source: Wall Street Insights

By Zhang Yaqi

The AI mega-cycle is still in its infancy. HSBC's report outlines six core investment theses for 2026: cloud giants' capital expenditure is set to surge by 44%, power supply emerges as a critical bottleneck, and liquid cooling technology may become a standard configuration. As ASIC chips rise to challenge NVIDIA and AI smart glasses and other end devices see explosive growth, HSBC is bullish on the S&P 500 with a target price of 7,500 points. Keeping a close eye on the dual game of computing power and electricity supply will be the key to success.

With the kickoff of the AI mega-cycle, the global technology industry is in the early stage of structural transformation. Although there are still discussions in the market about the pace of capital expenditure and returns, the potential productivity gains brought by AI have established the theme of its long-cycle growth. 2026 will be a critical window to observe AI infrastructure constraints, the evolution of the chip competitive landscape, and the integration of end-user applications.

According to information from Zhuifeng Trading Desk, HSBC's Nicolas Cote-Colisson analysis team released a report on the 7th, stating that AI is currently in the initial phase of a mega-cycle. Despite the fact that capital expenditure expectations outpaced initial revenues in October 2025, the market remains optimistic about the outlook for 2026 considering the speed of AI development and its potential to drive the global GDP of over $110 trillion. The analysis suggests that risks, opportunities, and narrative focuses will converge on six core dimensions.

At the macro level, HSBC is optimistic about the performance of U.S. stocks in 2026, setting a target price of 7,500 points for the S&P 500. Analysts advise investors to expand the AI trade from pure infrastructure suppliers to adopters and enablers.

## Constrained Cloud Capacity and Rising Capital Expenditure

The world's three major cloud giants (Amazon, Microsoft, and Alphabet) are facing severe order backlogs, but capacity expansion is constrained by the data center construction cycle. HSBC expects the capacity crunch to persist until 2026.

Against this backdrop, tech giants' willingness to invest remains strong. Alphabet, Meta, and Amazon have all hinted at "significant/material" growth in capital expenditure for 2026. HSBC forecasts that the aggregated capital expenditure growth rate of relevant enterprises will reach 44% in 2026. Except for Oracle and CoreWeave, most giants are expected to rely on free cash flow to fund their investments. HSBC believes that the current pace of spending is more constrained by "construction capacity" rather than "willingness to invest", and there is an upside risk to the 2026 spending guidance.

## Power Supply Becomes the Core Bottleneck for AI Development

The biggest obstacle to capacity expansion is not capital, but electricity supply. According to HSBC's analysis, the order delivery cycle for heavy-duty gas turbines has extended to more than four years, while small modular reactors (SMRs) are unlikely to contribute to capacity before 2035. This means that the power gap in data centers will persist in the short term.

In contrast, the Chinese market exhibits distinct characteristics. Local Chinese brands are rapidly penetrating the large-scale engine market thanks to shorter delivery cycles. In addition, liquid cooling technology will become a hot topic in 2026; as chip power consumption increases, liquid cooling will evolve from an optional feature to a mandatory configuration.

## Substantial Growth in Capital Expenditure

Alphabet, Meta, and Amazon have all indicated "significant/meaningful" increases in capital expenditure for 2026, and HSBC expects most large tech companies to follow suit when announcing their 2026 guidance (in late January/early February). HSBC's forecast shows that total annual capital expenditure will grow by 44%. Overall, funding is expected to come from current free cash flow generation, except for Oracle and CoreWeave, which will require additional financing.

Given the aforementioned capacity constraints and strong demand, the report expects capital expenditure to exceed initial guidance in an opportunistic manner. Therefore, an upside risk to the 2026 guidance is anticipated.

## Rise of ASIC Chips Challenges GPU's Dominance

In the AI chip space, NVIDIA's GPUs remain the top choice for cloud service providers, but competition from ASIC chips is intensifying. Alphabet's Ironwood TPU and Amazon's Trainium3 are notable examples. With better cost-performance ratios and cost advantages, ASICs are attracting more in-house R&D investments from hyperscalers.

HSBC predicts that the share of ASICs in cloud service providers' capital expenditure will rise from 2% in 2023 to 13% by 2027. While external chip sales may not make a substantial contribution to financial statements until 2027, discussions about orders and technology agreements in 2026 will impact valuations. NVIDIA's recent licensing agreement with Groq is regarded as a positive move to address the trend of falling inference costs.

## Competition in Cutting-Edge Models Trending Toward Oligopoly

The competition in cutting-edge large models is entering a phase of consolidation. HSBC believes that high sunk costs will lead to market rationalization, eventually forming an oligopolistic landscape dominated by a few giants alongside small, specialized players. Currently, the intelligence level of open-source/open-weight models is rapidly catching up with closed-source models, with the gap narrowing to approximately three months.

Gemini's growth rate in intelligence is outpacing ChatGPT. The focus in 2026 will be on the premium model pricing model and the commercialization of AI advertising. HSBC forecasts that AI chat advertising will account for 2% of digital advertising spending by 2030.

## AI Integration and Form Factor Innovation in End-User Hardware

2026 is regarded as a pivotal year for the integration of AI into smartphones, as well as the first year when new hardware challenges traditional platforms. Apple is expected to significantly enhance the AI capabilities of its hardware between 2026 and 2027, including an upgraded Siri and a potential foldable iPhone or a 20th-anniversary commemorative edition.

Smart glasses are seen as an important carrier for unleashing the power of AI large models. Meta's partnership with EssilorLuxottica currently leads the pack, while Samsung, Google, Apple, and Amazon all have plans to launch related products between 2026 and 2027. These devices are expected to partially replace the functions of smartphones through voice interaction and contextual information capture.

---

## Risk Warning and Disclaimer

The market is risky, and investment requires caution. This document does not constitute personal investment advice, nor does it take into account the specific investment objectives, financial situations or needs of individual users. Users should consider whether any opinions, views or conclusions in this document are consistent with their particular circumstances. Any investment made based on this document shall be at the user’s own risk.