Nvidia to invest up to 100 billion in OpenAI

On 22 September 2025, Nvidia announced it would invest up to 100 billion dollars in OpenAI to build data centres. The figure, by its size, illustrates how far the race to lead in artificial intelligence is mobilising capital. And it was not an isolated case: investment figures in the sector kept growing, at a time when each announcement seemed to top the last.

Faced with figures of this magnitude, the natural reaction swings between enthusiasm and unease. Both have their reasons: the technology promises to transform many sectors, but economic history shows that periods of heavy investment tend to mix projects with a future and excesses. What is useful is to keep a cool head and ask concrete questions about who pays, who buys and when the money will be recovered. That discipline is useful both for an investor and for a business owner considering supplying this sector.

An unprecedented figure

An investment of up to 100 billion dollars is a long-term bet, whose return depends on demand for artificial intelligence growing at the pace that is being priced in. When several players commit sums of that size at once, the whole sector becomes tied to a single expectation.

Data centres are facilities that house thousands of servers and consume large amounts of electricity and water for cooling. Building them requires land, permits, power grids and specialised equipment, and all of it takes years. An investment announcement is therefore a statement of intent that will materialise, or not, over a long period. In the meantime, projects can be modified, delayed or cancelled depending on demand and financing conditions.

Circular announcements

Circular announcements between suppliers and customers were starting to raise doubts about a possible bubble. When a company invests in its customer and the customer uses that money to buy its products, demand can look greater than it is, and that makes it hard to tell real growth from growth driven by the financing itself.

There is nothing illegal or necessarily wrong with such agreements, which can make industrial sense. But they call for a closer look at the numbers: where the money comes from, how much comes from end customers and how much from financing between partners. Anyone analysing a company in the sector would do well to ask what part of its revenue depends on that kind of circuit. Transparency in these relationships is a good sign, and its absence a reason for caution.

Real electricity demand

The electricity demand of data centres, by contrast, was very real. Whatever happens to valuations, the infrastructure and the energy needed to power it will remain a factor for power-sector companies, large consumers and those who plan their energy costs.

Even if valuations were to correct, servers already running consume energy, and projects under way will need grid connections. This affects electricity prices, infrastructure planning and competition for land and available capacity. Companies with high consumption can anticipate by reviewing their contracts and supply alternatives. It is also worth asking how much of the electrical capacity available in each area will go to these uses and how it will affect prices for other consumers.

Big figures catch the eye; small questions are what matter.

This article is general information. It is not legal, tax or investment advice, and it does not take anyone’s particular circumstances into account.