What the report says

Power supply for AI chips has become a bottleneck, and four overseas giants have spent more than $1.8 billion to snap up cutting-edge technology in that area, according to BigGo Finance S1.

The report, dated 17 November 2026, gives no names. It does not identify the four buyers, split the $1.8 billion between them, or say whether the money went into acquisitions, licences, joint ventures or supply agreements S1. Those gaps matter, because the shape of the deals decides who ends up owning the constraint.

The jargon, in plain English

An AI chip is a processor built to run many calculations at once, which is what training and running large AI models requires. A power supply is the part of the system that takes electricity and converts it into the steady, low voltage the chip needs.

That conversion sounds like a side issue. It is not. If power delivery cannot keep pace, chips cannot run at full speed, and fewer of them can be packed into a single data centre hall. The chip gets the headline. The power hardware decides how much of the chip's capacity is actually usable.

The money angle

Taken at face value, more than $1.8 billion split four ways works out to roughly $450 million per buyer, if the amounts were even S1. The report gives no breakdown, so treat that as arithmetic, not a finding.

The more useful signal is what the spending implies: buyers are paying for access rather than waiting for the market to supply it. That is a defensive move. It suggests the purchasers see scarce capacity in this layer of the AI supply chain and want it locked up before rivals do.

For investors, this is a reminder that the AI trade is not one trade. Chips, memory, networking, cooling and power delivery are separate markets with separate suppliers, margins and bottlenecks. A constraint in a small component can shape the economics of a much larger system.

What the reporting leaves open

Because the buyers are unnamed, it is not possible to say from this report alone whether the spending is concentrated among a few large players or spread across different regions S1. Nor is it clear whether the technology bought is protected by patents, which would matter for anyone trying to compete later.

What to watch

First, whether the four companies are identified. Second, whether the deals are structured as outright purchases or as agreements to secure supply. Third, whether the cost of this scarce hardware shows up in the price of building and running data centres. Fourth, whether other buyers follow with their own spending, which would confirm that power delivery, not compute alone, is where the queue is forming.