This is my current investment thesis, not financial advice. I own NVIDIA in my main portfolio and the position can change.
The reason I continue to own NVIDIA is not simply that AI is a large theme. It is that NVIDIA has repeatedly shown an ability to turn a technological lead into an operating machine: products that customers want, supply that has to be coordinated at extraordinary scale, and earnings power that follows when the company executes.
That is the heart of my bull case. The next few years will not be decided by whether AI matters. They will be decided by which businesses can keep delivering the computing, networking and systems customers need as the buildout becomes more demanding.
Execution at a scale that is hard to ignore
NVIDIA’s latest quarterly filing illustrates the scale of the current opportunity. For the quarter ended 26 April 2026, the company reported revenue of $81.6 billion, up 85.2% year on year. Net income was $58.3 billion, compared with $18.8 billion in the comparable prior-year quarter.
Numbers at that level can make a thesis look obvious after the event. They are not. The reason they matter to me is that they reflect execution across a difficult chain: silicon design, manufacturing capacity, memory, networking, systems integration, customer deployment and software support. It is one thing to make a fast chip. It is another to help customers build and operate AI infrastructure at enormous scale.
More than a GPU supplier
I do not see NVIDIA purely as a seller of individual GPUs. The company’s position is strengthened by the wider stack around accelerated computing: systems, high-speed networking and the CUDA software ecosystem that many developers and organisations already use.
That does not mean customers cannot change suppliers. They can, and they will keep trying to improve economics. But moving a serious AI workload is not always as simple as comparing the headline price of two chips. Software tools, developer familiarity, model performance, networking and the operational cost of changing a production environment all matter.
The bull case is not that NVIDIA wins every AI workload forever. It is that its scale and execution keep it central to the most important workloads long enough for the earnings power to compound.
The architecture cycle is part of the thesis
One of the reasons I find NVIDIA interesting is the cadence of the platform cycle. Customers are not just making a one-off purchase. They are trying to build capacity for models, inference, agents and applications that are changing quickly. If NVIDIA can keep making each generation valuable enough to justify upgrades, the opportunity is larger than a single hardware refresh.
The company’s supply commitments show how much planning this requires. As of 26 April 2026, NVIDIA disclosed $119 billion of supplier commitments, with most expected to be paid during fiscal 2027. That is not proof of future demand, and it increases the importance of execution. It does show the physical scale behind the AI buildout.
What could challenge the thesis
- Demand digestion: customers may pause after a period of exceptionally heavy infrastructure spending.
- Competition and custom silicon: hyperscalers, AMD and other suppliers have strong incentives to reduce dependency and improve their own economics.
- Export controls and China: policy can affect which products can be sold and where future growth comes from.
- Infrastructure constraints: data-centre capacity, power and customer financing all matter. Demand for chips alone is not enough if the wider buildout stalls.
- Expectations: a company producing exceptional results can still be a poor investment if the valuation assumes too much perfection.
My current view
I own NVDA because I think its scale, product execution and ability to turn an AI infrastructure cycle into earnings are still unusual. The growth is already visible. The investment question is whether the company can maintain its central role as customers become more sophisticated, competitors improve and the market starts to separate durable demand from short-term enthusiasm.
I will keep watching revenue quality, gross margins, customer spending behaviour, platform transitions and the willingness of customers to build around NVIDIA’s broader stack. The thesis is strong, but it is not a reason to ignore the price paid or the risk that this cycle eventually slows.
For the wider context, see my current portfolio. The financial figures cited above come from NVIDIA’s Q1 fiscal 2027 Form 10-Q.