How Nvidia makes its money
Nvidia develops graphics processors, or GPUs for short. Originally, these chips sat mainly in gaming PCs. Today they form the backbone of data centers for artificial intelligence. A GPU carries out a very large number of simple calculations at the same time. AI models need exactly this ability, both in training and in day-to-day operation.
The figures show how far the focus has shifted. In fiscal year 2026, which ended at the end of January 2026, Nvidia reported revenue of $215.9 billion, according to its official announcement. Of that, $193.7 billion came from the data center business. That is around 90%. Gaming contributed $16.0 billion. Professional visualization and automotive remain small additions.
Nvidia does not manufacture its chips itself. Contract manufacturers such as TSMC handle production. Nvidia concentrates on chip design, software and complete systems. A central building block is CUDA. Developers have been programming Nvidia GPUs through this software platform for almost two decades. Anyone who has built their AI applications on CUDA cannot switch to another supplier without effort. This lock-in is an important part of the business model.
Why I took a closer look at the company
Hardly any portfolio gets around Nvidia. At the beginning of August 2026, its market value was more than $5 trillion, according to Stock Analysis. That makes Nvidia one of the most valuable companies in the world. If you own an ETF on the MSCI World or the S&P 500, you automatically hold Nvidia as one of the largest positions.
This has a consequence that many people underestimate. The performance of this one stock moves entire indices. So the question of how Nvidia is doing is not just a matter for individual shareholders. It affects almost everyone who invests broadly. That is why a close look at the business is worthwhile, even if you never buy the stock on its own.
I keep the share price separate from the company. A rising price does not prove healthy development. A falling price does not prove a problem. What matters is whether demand, margin and competitive position hold up. And how much of the future is already contained in the current price.
AI data centers drive the business
Cloud providers and AI companies are building data centers for artificial intelligence around the world. Nvidia’s accelerators are the standard for them. The company is currently shipping the Blackwell generation and is already introducing the next platform, Vera Rubin.
According to Nvidia, fiscal year 2026 brought revenue of $215.9 billion, an increase of 65%. GAAP net income reached $120.1 billion. The first quarter of fiscal year 2027 ended on April 26, 2026. For that quarter, Nvidia reported revenue of $81.6 billion, up 85% on the prior year. The data center segment grew by 92% to $75.2 billion. The gross margin was 74.9%. For the second quarter, the company projected revenue of around $91 billion. The next figures are due at the end of August 2026.
Margins like these are exceptional in the semiconductor industry. They show that Nvidia currently sells with almost no pricing pressure. That does not have to last. It is also striking who is buying. In the third quarter of fiscal year 2026, four direct customers accounted for 61% of revenue. Behind these customers are mainly the large cloud groups and leading AI labs.
Three questions before a possible purchase
The first question concerns the business. Is revenue growing because more systems are being shipped, and is the margin holding up at the same time? A single record announcement is not enough of an answer. The trend over several quarters says more. This also includes the question of how customers finance their huge orders. If the billions come from ongoing profits, that is more stable than financing through debt and equity stake arrangements.
The second question concerns the trend. Even an excellent company can be a weak stock for a long time if expectations have run ahead. That is why I check whether the price is trading above its long-term moving averages and which direction these lines are pointing. This does not replace a valuation. But it shows whether a decision would go with the market or against it.
The third question concerns momentum. After strong moves, a stock is often overstretched in the short term. The Relative Strength Index helps to put this state into context. On its own, the value is not a signal. Together with business figures and trend, it reveals whether nervous trading is at play or whether the situation has calmed down.
The key risks
A few customers carry revenue. When four direct customers account for well over half of revenue, every single budget decision carries a lot of weight. If one large buyer postpones its orders, Nvidia feels it immediately.
Export controls cut off the China business. The US restricts the sale of advanced AI chips to China. In its outlook for the second quarter of fiscal 2027, Nvidia expects no data center revenue at all from China. Since June 2026, according to Al Jazeera, the US rules also apply to foreign subsidiaries of Chinese groups. Politicians can tighten or loosen these rules at any time. None of this can be planned for.
The largest customers are building their own chips. Google, Amazon, Microsoft and Meta are developing their own AI accelerators, often together with Broadcom or Marvell. AMD has signed large supply agreements with OpenAI and Meta. Nvidia still clearly dominates. But every chip a large customer builds itself potentially replaces an Nvidia order.
AI investment runs in cycles. Data centers are built in waves. Once operators have filled their capacity, or if the income from AI services disappoints, orders can pause. Semiconductor history has seen many such turning points.
The valuation assumes further growth. A market value of more than $5 trillion contains the assumption that the AI build-out will continue for many years. If only part of that fails to materialize, the stock can fall sharply even though the company keeps growing.
How I put the high expectations into context
Nvidia has been reporting record figures for years. The bar rises with every quarter. Even a strong result can disappoint on the stock market if the market had priced in even more. That is why I always read headlines about Nvidia stock in two steps. First I check what has changed in the business. Then I check what has only changed in expectations.
I leave leveraged products out of this assessment. Leverage can force a position to close before a long-term thesis even has a chance to prove itself. With a stock that reacts as strongly to news as Nvidia, that is a risk of its own and not a tool for analysis.
My conclusion on Nvidia stock
Nvidia sits at the narrowest point of AI infrastructure. The combination of leading chips, the CUDA ecosystem and complete systems has given the company a rare market position. The published figures show a level of demand the industry has never seen before.
At the same time, the stock bundles several risks. A small number of customers, political intervention, emerging competition from those very customers and a valuation that assumes sustained growth. The simple story of the unstoppable AI winner therefore falls short.
For my own assessment, I check three levels in a fixed order: business development, long-term trend, short-term momentum. This does not result in a recommendation and does not make a decision for you. But it ensures that Nvidia is judged by the same standards as any other stock.