It’s important to put NVIDIA’s last two quarters into context. When the company reported its April quarter, it guided July revenue up 50% from the Street. Ultimately they exceeded those revenue expectations by 20%. I put that guidance and results in the top decile of tech company earnings I’ve tracked. On top of that, the company guided October revenue up 28% higher than the Street.

That guidance would have been even higher if not for production constructions, largely at TSMC. The impact of production constraints is unclear, but the company did emphasize that backlog gives them visibility into CY24. I estimate guidance would have been 10 to 20% higher if not for the bottleneck.

NVIDIA’s results are driven by a shift in compute architecture, largely with hyperscalers which include AWS, Azure, and Google Cloud. These cloud providers see the opportunity to host AI applications and need to rearchitect their tech stack to accommodate this increased computing capacity.

Underlining this shift to accelerated computing is a move from CPUs to GPUs. CPUs have been around for 40 years, and while they have gotten faster (doubling in speed about every two years) they are limited in their ability to work in parallel. GPUs are optimized for parallel computing, which is essentially required for AI training. The big picture: The shift away from CPUs to GPUs is driving NVIDIA’s success.

Investors are always looking forward, and the central question around investing in NVIDIA is the rate and duration of its top-line growth. As it stands today, and likely for the next three to five years, NVIDIA’s position as the dominant GPU provider will remain unchanged. The reason is that NVIDIA not only sells chips but also has a development platform, CUDA, which creates a moat around NVIDIA’s chip business. Developers learn CUDA and therefore develop applications that only run on NVIDIA’s GPUs. In other words, for NVIDIA to lose its pole position, another chipmaker would have to provide a return on investment that’s so compelling it motivates developers to learn a new way to build the AI models.

At Deepwater, we believe in five-plus years that there will be compelling hardware that will challenge NVIDIA’s GPU dominance. We’re investors in Rain, a company with ambitions to lower AI compute costs by a factor of 100 to 1,000x. It’s that kind of performance improvement that will be required to unseat NVIDIA. The bottom line here is that NVIDIA has a runway for the next three to five years to run the table on the shift from CPUs to accelerated GPU computing. I believe that translates to NVIDIA’s growth exceeding the 50% top-line expectation for CY24 and the 25% top-line expectation for CY25.

Taking a step back, NVIDIA’s results show us that the AI theme is accelerating. It’s easy to get lost in the details of a mega-cap tech earnings report. Investors debate about revenue pulled forward from China trade restrictions, the sustainability of growth, and the potential for competition. What powers through all of those debates is that something bigger is going on in tech.

AI has moved from a science project to a mainstream arms race spanning the largest tech companies in the world to 15,000 startups that purchased NVIDIA last quarter. ChatGPT lit the fuse, and we’re just beginning to see the build-out of the infrastructure to enable this paradigm shift. I expect in the next couple of years there will be an acceleration of applications that will influence our everyday lives, built on AI and powered by NVIDIA chips.