Amazon tops 2025 AI capex race with $125B+ spend

Home

Blog

市場新聞

Big tech’s 2025 AI capex race: Amazon leads the pack with $125B+ spend

November 26, 2025

A futuristic Amazon data centre scene with glowing servers, neon cables, and AI-labelled hardware emerging from an illuminated Amazon-branded wall.

The numbers are staggering. In 2025, Amazon, Microsoft, Alphabet and Meta are collectively guiding to $360–400 billion in capital expenditures – a ~60% year-over-year increase, with the overwhelming majority directed toward AI-related infrastructure (data centres, custom silicon, GPU/Trainium clusters).

On 24 November 2025, BNP Paribas Exane initiated coverage on Amazon with an Outperform rating and a $320 price target - currently the highest among major brokers and implying ~39% upside from the 26 November close of ~$230.

2025 capex guidance - The big four

Company	2025 Capex Guidance	Primary AI Focus Areas
Amazon	>$125 bn (raised multiple times in 2025)	AWS hyperscale clusters, Trainium/Inferentia chips, sovereign & government clouds
Microsoft	$80–121 bn (FY ending Jun-26)	Azure expansion, OpenAI infrastructure, enterprise GenAI
Alphabet	~$91 bn (raised from $85 bn)	Google Cloud TPUs, catching up on historical capacity shortages
Meta	$70–72 bn	Llama models, AI-driven advertising, massive single-site data centres

Sources: Company filings, earnings calls, BNP Paribas Exane, BBC, Bloomberg, Reuters

Why BNP Paribas exane sees Amazon differently

Analysts at BNP Paribas Exane argue that concerns about Amazon under-investing or being late in AI are “overblown” in light of the company’s disclosed spending and pipeline. Amazon’s finance team has discussed a 2025 capex outlook of roughly $125B, with expectations for a higher figure in 2026, and has indicated that the vast majority is focused on AI-focused infrastructure such as data centres, networking and in-house accelerators for AWS.

The note highlights several points that differentiate Amazon in this capex cycle:

Vertical integration: By designing its own AI chips such as Trainium and Inferentia, management has indicated potential cost and efficiency benefits relative to relying solely on third-party GPUs, which could help with both pricing and capacity flexibility over time.

Multiple monetisation channels: The AI infrastructure is positioned to support not only AWS enterprise and government workloads but also improvements in advertising relevance, logistics optimisation, and consumer-facing services, giving Amazon several ways to translate infrastructure into revenue.

Long-term margin narrative: The firm’s thesis references scenarios where AWS growth re-accelerates into the mid-20% range and advertising grows at 20–25%+ annually, contributing to potential group-level operating margin expansion over a multi-year horizon, though actual outcomes will depend on execution and demand.

Key investor debates & risks

Debate / Risk	Representative “Bull” Perspective	Representative “Bear” Perspective
Scale of capex	Large-scale AI capex is seen as necessary to secure long-term demand in cloud, AI services and advertising, with the view that current spending reflects structural growth in workloads.	Some investors worry about an overbuild scenario where capacity is added faster than demand, reducing returns on invested capital and leaving assets underutilised.
Timing of returns	Supportive commentators expect utilisation and monetisation to ramp through 2026–2027 as generative AI projects move from pilots to full deployment, especially in cloud and enterprise software.	Sceptical views highlight near-term free cash flow pressure and uncertainty over how fast experimentation converts into large recurring AI spending.
Competitive positioning	Proponents see Amazon’s full-stack strategy (from chips to cloud to consumer applications) as a durable advantage relative to peers focused on individual layers of the stack.	Critics point to strong momentum at Microsoft Azure and Alphabet/Google Cloud and question whether any single company can maintain a clear lead.
Macro sensitivity	Some argue that cloud and AI spend are becoming “infrastructure-like,” remaining resilient even if consumer spending slows, particularly for mission-critical workloads.	Others worry that a broader economic slowdown could weigh on digital ad budgets and e-commerce volumes—key revenue drivers for Amazon and Meta.

Upcoming catalysts/data points

AWS re:Invent - early December 2025

Market participants will likely watch for announcements on new AI services, model offerings, and capacity expansions, as well as customer case studies that illustrate production-scale workloads.

Amazon Q4 2025 results - expected late January / early February 2026

Key metrics to watch include AWS revenue growth rates, segment operating income, and management commentary on AI-driven demand and 2026 capex plans.

Peer earnings and updated guidance - early 2026

Earnings from Microsoft, Alphabet and Meta in early 2026 are expected to provide fresh details on capex trajectories, AI product adoption, and how each company is balancing investment with free cash flow.

These events may offer more clarity on how quickly AI investments are translating into revenue and whether capex levels remain elevated, moderate, or increase further in 2026.

Amazon technical insights

At the start of writing, Amazon (AMZN) is trading near $229, recovering modestly from recent lows while holding above key supports at $218.45 and $213. A drop below these zones could trigger sell liquidations, while a push higher puts the $250.15 resistance level back in focus - an area where traders may take profits or look for renewed buying.

The RSI remains flat around 50, signalling neutral momentum and suggesting the market is still searching for direction after the recent pullback.