These three things are the same story: a chip family, a buzzword, and a balance sheet. NVIDIA’s FY2024 10-K names “Hopper and NVIDIA Grace Hopper processors optimized for generative AI, LLMs and other AI workloads,” says the company “introduced NVIDIA DGX Cloud,” and reports total revenue of $60.9 billion, up from $27.0 billion a year earlier. Connect the dots and the dots are very large.
Follow both the money and the technology. A year ago this site read the FY2023 filing as the H100 shipping into a market that wanted it. The FY2024 document is what that market wanting it looks like in dollars: revenue more than doubled in a single year, and the filing explicitly attributes its product roadmap to generative AI and large language models.
The DGX Cloud line is its own signal. By offering its systems as a cloud service, NVIDIA is telling investors it wants to capture AI demand not only by selling chips but by renting access to them. That is a company reading its own moment and trying to widen how it monetizes the same silicon.
For a general reader: this is the filing where “generative AI” stops being a theme and becomes the dominant explanation for a hardware company’s results. When the annual report ties the roadmap to LLMs and the revenue more than doubles in the same year, the technology narrative and the financial narrative have fully merged.
As a contemporaneous marker, early 2024: the generative-AI build-out is now legible as a number, $60.9 billion, with the product language to match. Whether this growth rate is a peak or a step is the only open question, and it is one the next filing will settle. Filing data and the evidence index via EdgarBeast.