Name: Big Tech Aims to Spend $650 Billion This Year in AI Race
Uploaded: 2026-02-06T15:06:14+00:00
Duration: 186 s

Speakers

Mandeep Singh — Senior Analyst, Bloomberg Intelligence

Summary

Mandeep Singh of Bloomberg Intelligence discusses the expected $650 billion in 2026 capital expenditures from four major US tech companies. He argues supply constraints justify some spending but sees wide differences in returns: Google has earned its CapEx through strong cloud growth and margins, while Amazon's growth is pressuring margins and Oracle faces OpenAI-linked pushback. He also highlights TSMC as a key bottleneck and sees limited upside for Nvidia from the CapEx wave.

Four large US tech companies plan about $650 billion in 2026 CapEx, largely for data centers and AI chips.
Supply constraints and insufficient capacity justify some CapEx increases.
Google is seen as the only hyperscaler that has earned its higher CapEx, with strong cloud and infrastructure revenue growth and 30-40% incremental margins.
Amazon's cloud growth has accelerated but at the expense of margin, and its $200 billion CapEx raises customer and monetization questions.
TSMC cannot meet hyperscaler demand, forcing a significant ramp in semiconductor capital equipment purchases.
Nvidia's upside from the CapEx increases appears limited despite the spending wave.
Oracle faces investor pushback because its AI story is heavily tied to OpenAI.
Investor tolerance for public cloud free cash flow and margins remains a key risk if CapEx continues at this pace.

Ideas

Mandeep Singh Senior Analyst, Bloomberg Intelligence 0:25

Google earns AI capex with cloud margins

Google is the only hyperscaler that has earned its CapEx increase: cloud revenue grew 48%, infrastructure revenue grew north of 50%, and the spending is being done at incremental margins of at least 30-40%, unlike Amazon where margins are declining. This supports a positive relative view on Google.

Mandeep Singh Senior Analyst, Bloomberg Intelligence 0:39

Amazon AI capex pressures margins and returns

Amazon's cloud growth has accelerated but at the expense of margin, and its $200 billion CapEx raises questions about who the customers will be and whether the spending is for revenue-generating inference or sunk-cost training, making it a less attractive hyperscaler than Google.

Mandeep Singh Senior Analyst, Bloomberg Intelligence 0:54

TSMC is key supply bottleneck for AI

There is a rush among hyperscalers to secure supply at TSMC, and larger customers have a better shot; however, TSMC's capacity cannot meet the demand implied by hyperscaler CapEx, making it a critical supply-chain bottleneck that must ramp capex.

Mandeep Singh Senior Analyst, Bloomberg Intelligence 1:16

TSMC must buy more semiconductor equipment

Hyperscaler CapEx demand exceeds TSMC's capacity, so TSMC must significantly ramp its own CapEx and semiconductor capital equipment purchases to service hyperscaler demand, supporting the semiconductor capital equipment sector.

Mandeep Singh Senior Analyst, Bloomberg Intelligence 1:29

Something limits Nvidia's AI capex upside

Although hyperscaler CapEx is increasing, something is limiting the upside NVIDIA may see from those increases, suggesting NVDA's benefit is constrained despite the AI spending wave.

Mandeep Singh Senior Analyst, Bloomberg Intelligence 2:30

Oracle's OpenAI tie draws investor skepticism

Oracle is facing investor pushback because its AI story and backlog are heavily tied to OpenAI, raising questions about revenue quality and whether the associated CapEx can be monetized.

Big Tech Aims to Spend $650 Billion This Year in AI Race

Summary

Ideas

Up Next