What happened
Nvidia reported quarterly revenue of 96.2 billion dollars for the three months ended 26 July 2026, up 18 per cent on the previous quarter and up 106 per cent on the same period a year earlier. The figures were published after the US market close on 26 August and covered what the company calls the second quarter of its 2027 financial year, because Nvidia runs its accounting calendar roughly six months ahead of the calendar year.
The engine of that growth was the data centre division, which sells the specialised computer chips that train and run artificial intelligence systems. Data centre revenue reached 89.0 billion dollars, up 18 per cent on the previous quarter and up 117 per cent year on year. That single division now accounts for more than nine tenths of everything Nvidia sells. Gaming graphics cards, once the core of the business, are now a rounding error by comparison.
Management guided to revenue of about 108 billion dollars, plus or minus 2 per cent, for the current quarter. Notably, that forecast assumes no data centre compute revenue from China at all. Export restrictions have effectively closed the largest single market outside the United States for advanced AI chips, and Nvidia has chosen to build its guidance on the assumption that nothing comes back.
The scale is difficult to hold in your head. A single quarter of 96.2 billion dollars is more than the annual revenue of most companies in the FTSE 100. Nvidia is now adding roughly the revenue of a large European bank every three months, and it is doing so while raising rather than cutting the prices it charges.
Why it matters
Nvidia has become the closest thing markets have to a single-stock economic indicator. Because almost every large technology company buys its chips, Nvidia results are read as a live measurement of how much the rest of the industry is actually spending on artificial intelligence, rather than how much it says it plans to spend. When the number beats expectations, the whole AI investment story looks funded. When it disappoints, the story looks like a promise.
That matters far beyond technology investors. Nvidia is one of the largest companies in the S&P 500 by market value, which means it carries an outsized weight in index funds. A tracker fund does not pick stocks; it buys every company in the index in proportion to its size. So a single company growing this fast mechanically becomes a larger and larger share of what millions of ordinary savers own.
There is a second-order effect on the wider economy too. The money Nvidia collects comes out of the capital expenditure budgets of a handful of very large customers. That spending shows up in construction jobs for new data centres, in electricity demand, and in orders for everything from cooling equipment to transformers. AI capital spending has become a measurable contributor to economic growth in the United States, which is why economists now watch these results alongside official statistics.
The risk cuts the other way as well. Concentrated growth means concentrated fragility. If the handful of buyers funding this boom decide they have bought enough capacity, the revenue does not slow gently. It stops.
Explained simply
Nvidia is the company selling the shovels in a gold rush, except every prospector on earth has queued up at once, paid in cash, and asked to reserve next years shovels too.
Artificial intelligence systems are built in two stages. First they are trained, which means running enormous amounts of data through a mathematical model until it learns patterns. Then they are run, which means answering real questions from real users. Both stages need a particular kind of chip called a graphics processing unit, or GPU, which is good at doing millions of simple calculations simultaneously rather than a few complicated ones in sequence.
Nvidia designs the best of these chips and, just as importantly, owns the software layer that developers use to program them. That software has been the industry standard for over a decade. A rival can build a competitive chip, but it also has to persuade the worlds AI engineers to rewrite their code, which is a far harder sell.
So when a company decides to build an AI product, the decision to buy from Nvidia is usually already made. The only question is how many chips and how soon. That is why the revenue arrives in enormous, lumpy blocks: a single hyperscale customer ordering capacity for a new data centre can move a quarterly number by billions on its own.
The limit on growth is therefore not demand. It is how fast the factories that physically manufacture the chips can produce them, and how fast customers can build the buildings and secure the electricity to house them.
What it means for you
If you hold a global equity tracker, an S&P 500 fund, or the default fund in a workplace pension, you almost certainly own Nvidia and it is probably your largest single holding. You did not choose it. The index chose it for you, because index funds weight companies by market value. It is worth logging into your pension provider and looking at the top ten holdings of your default fund, because most people are considerably less diversified than they assume.
UK-focused investors have the opposite exposure. A FTSE 100 tracker holds almost none of this, because the London index is weighted towards banks, oil, mining and consumer goods. That is a large part of why UK index funds have lagged global ones over the past three years, and why a portfolio split between a FTSE tracker and a global fund behaves very differently from one holding only the former.
For savers deciding where to put new money, the practical question is not whether Nvidia is a good company. It clearly is. The question is whether you want a further seven or eight per cent of every new pound going into one share price. Adding a global fund that deliberately equal-weights its holdings, or holding some bonds or UK equity alongside, are both straightforward ways to reduce that concentration without abandoning growth.
And if you work in construction, electrical engineering, or energy in the United States or Ireland, this spending is quietly showing up in your sector as data centre projects compete for skilled labour and grid connections.
The bigger picture
Booms in capital equipment have a long and consistent history: they run further than sceptics expect and then correct faster than optimists expect. Railway building in the 1840s, telecoms fibre in the late 1990s and shale drilling in the 2010s all followed the pattern. In each case, the underlying technology proved genuinely transformative and the companies supplying the equipment still saw revenue fall sharply once buyers were satisfied.
The specific thing to watch is not Nvidia revenue but customer capital spending plans. The largest cloud companies publish their capital expenditure guidance every quarter, and those figures lead Nvidia results by roughly two to three quarters. If those budgets flatten, the slowdown arrives later at Nvidia rather than never.
The China assumption is the other variable. Nvidia has built its forecast on receiving nothing from the market, which means any easing of export rules would be pure upside, while further tightening changes little. That is a deliberately conservative way to set expectations, and it removes one of the more obvious ways the company could disappoint.
