Six Key Issues in the Early Stage of the AI Mega Cycle

As the global AI supercycle gets underway, the technology sector is entering the early stages of a structural transformation. While markets continue to debate the pace and payoff of capital expenditures, AI’s potential to boost productivity has firmly established its long-term growth narrative. The year 2026 will serve as a critical observation window for infrastructure constraints, the evolution of the chip competition landscape, and the integration of AI into end-user applications.

According to ZhuiFeng Trading Desk, an HSBC research team led by Nicolas Cote-Colisson released a report on the 7th stating that AI is currently in the early phase of a long-term cycle. Although capital expenditure expectations temporarily ran ahead of early-stage revenues in October 2025, the market remains optimistic about the outlook for 2026, given the rapid pace of AI development and its potential to impact more than US$110 trillion in global GDP. HSBC believes that risks, opportunities, and narrative focus will converge around six core dimensions.

At the macro level, HSBC maintains a constructive view on U.S. equities in 2026, setting a target of 7,500 for the S&P 500. The firm advises investors to broaden AI exposure beyond pure infrastructure suppliers to include application-layer players and ecosystem enablers.

Cloud Capacity Constraints and an Upward Capex Cycle

Order backlogs at the three global cloud hyperscalers—Amazon, Microsoft, and Alphabet—remain severe, while capacity expansion is constrained by long data center construction cycles. HSBC expects these capacity bottlenecks to persist through 2026.

Against this backdrop, big tech’s willingness to invest remains strong. Alphabet, Meta, and Amazon have all signaled “significant” or “meaningful” increases in capital expenditures in 2026. HSBC forecasts aggregate capex growth of 44% across the sector. With the exception of Oracle and CoreWeave, most companies are expected to fund spending through free cash flow. HSBC argues that current spending levels are constrained more by “build capacity” than by “investment appetite,” and that 2026 capex guidance carries upside risk.

Power Supply Emerges as the Key Bottleneck for AI

HSBC identifies power availability—not capital—as the primary constraint on AI expansion. Delivery lead times for heavy-duty gas turbines have stretched beyond four years, while small modular nuclear reactors (SMRs) are unlikely to contribute meaningfully before 2035. As a result, data center power shortages are expected to persist in the near term.

China presents a contrasting dynamic. Domestic suppliers, benefiting from shorter delivery cycles, are rapidly gaining share in the large-engine market. Meanwhile, liquid cooling is set to become a major theme in 2026. As chip power density rises, liquid cooling is expected to shift from optional to essential.

A Step Change in Capital Expenditures

Alphabet, Meta, and Amazon have all hinted at “significant” or “meaningful” increases in 2026 capital spending, and HSBC expects other large technology companies to follow suit when they issue 2026 guidance in late January or early February. HSBC projects full-year capex growth of 44%.

Overall, funding is expected to come primarily from existing free cash flow generation, with the exceptions of Oracle and CoreWeave, which may require additional financing.

Given capacity constraints and robust demand, the report expects capital expenditures to exceed initial guidance in an opportunistic manner, reinforcing the view that 2026 guidance carries upside risk.

ASIC Chips Challenge GPU Dominance

In AI semiconductors, NVIDIA’s GPUs remain the preferred choice for cloud providers, but competition from ASICs is intensifying. Alphabet’s Ironwood TPU and Amazon’s Trainium3 are prominent examples. With superior cost efficiency and performance-per-dollar advantages, ASICs are attracting increasing in-house investment from hyperscalers.

HSBC estimates that ASICs’ share of cloud capex will rise from 2% in 2023 to 13% by 2027. While external chip sales are unlikely to meaningfully impact financials until 2027, discussions around orders and technical partnerships in 2026 will influence valuations. NVIDIA’s recent licensing agreement with Groq is viewed as a constructive response to the trend of declining inference costs.

Frontier Model Competition Moves Toward Oligopoly

Competition among frontier large language models is entering a consolidation phase. HSBC argues that high sunk costs will drive rationalization, ultimately resulting in an oligopolistic market dominated by a handful of major players alongside smaller, specialized participants.

Open-source and open-weight models are rapidly closing the gap with closed-source models, with the performance difference now narrowed to roughly three months. Gemini’s intelligence growth rate has surpassed that of ChatGPT. In 2026, focus will shift toward premium model pricing strategies and AI-driven advertising monetization. HSBC estimates that AI chat advertising could account for 2% of global digital ad spending by 2030.

AI Integration in End Devices and Hardware Form Innovation

HSBC views 2026 as a pivotal year for AI integration into smartphones, as well as the first year in which new hardware categories meaningfully challenge traditional platforms. Apple is expected to significantly enhance AI capabilities in its hardware lineup in 2026–2027, including an upgraded Siri and potentially a foldable device or a 20th-anniversary iPhone.

Smart glasses are seen as a key form factor for unlocking the full potential of large AI models. Meta’s partnership with EssilorLuxottica currently leads the field, while Samsung, Google, Apple, and Amazon are all expected to launch related products in 2026–2027. These devices could partially replace smartphones by enabling voice-based interaction and contextual awareness.

Article source：Wallstreetcn