The biggest takeaway from earnings was the CY27 revenue guide. Management expects revenue to grow about 70%, compared to the Street consensus at 45%. After adjusting for some of the growth that is price-increase-related to offset higher memory costs, the guide suggests the “brain” of AI will grow by at least 60% next year vs. previous expectations of 45% growth. The correlation between Nvidia’s revenue growth and the growth of the “AI Brain” (i.e., data centers) is nearly 1:1. So we can use Nvidia’s guidance as a proxy for expected changes in data center capacity, which equates to changes in the size of the AI Brain.

In other words, as Nvidia grows, so does the rest of the data center world.

Management also said customer forecasts point to growth doubling next year, but supply limits the outlook to about 70%. That implies that growth next year, excluding supply chain headwinds, would be well above 70%, potentially around 88%, which is the same growth rate expected in CY26. Keep in mind, CY26 growth would have been higher too if not for the tight supply chain. Those similar growth rates off of numbers that are getting much bigger are impressive. Growing $400B by 88% next year is much harder than growing CY25’s $216B by 90% in CY26. They are bucking the law of large numbers.

That matters because investor concern around Nvidia has been the slope of the growth deceleration, with revenue growth peaking in CY23 at 126% y/y, then coming down to 114% in CY24 and 65% in CY25, before being expected to accelerate to 88% in CY26.

As a comparison, Google is roughly the same size in revenue and grows around 20% annually. The last time it saw large revenue growth was 53% in CY19 and 41% in CY21, still well below the growth rates that Nvidia should post over the next two years.

The challenge with Nvidia’s growth rates is every monster year just increases investors’ concern about a growth drop-off in the out year.

The reason why the stock was only up about 8% after guiding to revenue next year 20% above the Street was in large part due to concern over gross margin. For the October quarter, the company guided gross margin to 74%, compared to the Street at 75%, and pointed to gross margin moving to 71-72% in the January quarter, below the Street at 74%. For next year, they guided to 72-73%, while investors had been thinking closer to 74%.

The reason was, of course, spiking memory costs, which I believe could almost double next year. All management said was that component costs have risen significantly, memory pricing conditions are extreme, and the magnitude of memory price increases has exceeded prior expectations. They also said their own price increases take effect in the Apr’27 quarter, which is the reason why margins will rebound to the 72-73% range next year. While we don’t know the magnitude of the upcoming price changes, my guess is the weighted price change will be around 15%.

Overall, despite the margin wobble, they’re still in a great place. As a point of reference, margins were 59% in CY22 (pre-AI boom) and 72% in CY23. The few-percentage-point dip is explicitly temporary, driven by the short-term spike in memory costs that the company is addressing through price increases.

A day after earnings, Nvidia announced it would pay $12.9B to acquire Hugging Face. This is the clearest sign yet of how seriously Jensen takes the open-source model opportunity. This started about a year ago with Nemotron, Nvidia’s open-weight model family, and by mid-2026 Nemotron 3 Ultra had reportedly become the top-performing open-weight model out of the U.S. Hugging Face is the distribution layer for open models, which gives Nvidia a stronger position just above the chip.

That fits with last week’s reported $6B Poolside licensing deal, which brought over more than 100 engineers to work on Nemotron.

The reason Nvidia is building out the open model product line is the inverse of the reason why the hyperscalers are building custom silicon. OpenAI, Google, and Amazon are building their own silicon to reduce dependence on Nvidia. Open-source model builders are unlikely to build their own chips. By becoming a larger force in open models, Nvidia gives open-source developers more reasons to stay anchored to its hardware. Effectively, Nvidia is building a backup plan to both ride the growth of open source and hedge itself if its biggest customers get aggressive about building their own chips.

My take is this is a smart defensive move. Nvidia is protecting its position with its biggest customers while building a wider open-source ecosystem that should keep demand tied to its platform.

Disclaimer