DeepSeek caused a stir a few weeks ago because it claimed to have been able to drastically reduce the cost and effort of training and operating a large language model (LLM) of similar power to GPT-4, Llama, Gemini or Claude.
Although this Chinese company has brought a number of improvements – as is always to be expected with technology (cheaper, faster, better, more functions) – the information that startled everyone and caused Nvidia’s share price to plummet on the stock markets turned out not to be true at all.
No more talk of DeepSeek’s V3 language model costing less than $6 million. On closer inspection, the Chinese company incurred similar costs to OpenAI, Meta, Anthropic of Google to train their language model, they just “forgot” to state that their model naturally also ran on tens of thousands of Nvidia H100 chips from their parent company (wherever they got them from) and that they also incurred the other costs that everyone else did. In the same way, the supposed Chinese chips, which are less powerful than those from Nvidia and which DeepSeek was supposed to use due to the American technology boycott, were not actually used at all.
To cut a long story short: there is still no way around Nvidia.
Upcoming investments in AI
If you look at how much money the big American tech companies put into developing and expanding infrastructure for AI last year and how much they plan to spend on it in 2025, and how much of the Stargate conglomerate (OpenAI, Softbank, Oracle & Co) is to be invested ($100 billion immediately, $500 billion in total over the next 5 years), then you don’t have to look far to see where the money will go.
Nvidia will be a key technology partner for the Stargate project, and more than 75% of the systems on the most powerful supercomputers use Nvidia technology.
But in which direction will Nvidia continue to move? All we have to do is take a closer look at the latest Nvidia figures, because they already contain the clues to the future.
Nvidia quarterly and fiscal year figures
Figures for the fourth quarter of fiscal year 2025:
| Q4/FY2025 | Increase compared to the previous third quarter | Increase compared to the same quarter of the previous year | |
| Revenue | $39.3 billion | +12% | +78% |
| Profit | $22 billion | +14% |
For the full fiscal year 2025, data center revenue was $35.6 billion, up 16% from Q3 fiscal year 2025 and up 93% from the year-ago quarter. Revenue from Nvidia’s former core gaming business fell to $2.5 billion, down 22%.
| Fiscal year 2025 | Increase compared to previous year | |
| Revenue | $130.5 billion | +114% |
| Profit | $72.8 billion | +145% |
Outlook for the first quarter of fiscal year 2026:
| Q1/2026 (expected) | Increase compared to the same quarter of the previous year | Increase compared to the same quarter of the previous year | |
| Revenue | $43 billion | +12% | |
| Profit | $22 billion | +14% |
Emerging markets for Nvidia
In Nvidia’s current balance sheet, one item almost goes under the radar: sales from the automotive and robotics industry. This rose by 27% to $570 million. That may still be peanuts at the moment, but anyone who follows these two sectors knows that a revolution is imminent. At 27%, growth there was twice as high as in the other sectors.
Nvidia is focusing on these two areas, as the in-house automotive department alone employs around 3,500 people out of a workforce of just over 30,000. At CES 2025, Nvidia also presented the GR00T software, which will serve as the basic model and technology for robots. Both cars (whether autonomous or still manually controlled) and robots require Nvidia’s GPUs, and usually several per car or robot. Even if these GPUs are priced far below the H100 and Blackwell chips, which range from $10,000 to $40,000, it is the sheer mass of chips that will make the difference, which will be in the hundreds of millions or even billions per year.
We are at the beginning of strong growth phases for these two technologies:
- Autonomous cars are now scaling in the form of robotaxi fleets such as Waymo, Zoox or Tesla and will also find their way into more and more cars from other manufacturers;
- Robots: development is currently taking giant leaps in this area.
You only have to look at the latest videos from companies like Figure.AI, a robotics startup from Sunnyvale, to see where the growth for Nvidia will really take off.
A few million AI chips (the top models) for the data centers are lucrative and nice, but the bulk and the money will come from the hundreds of millions of cars and robots in which the chips will be installed in the future. And I’m not even talking about irons, desk lamps and other household appliances that will also contain GPUs in the future.
In other words, the path for Nvidia is mapped out and the new markets to be developed will make the company really interesting, both on the stock market and technologically.