Nvidia FQ3 25 – Valuation Support

Reasonable multiple prevents a crash.

• Nvidia reported excellent results as the AI boom remains on track, but it failed to raise guidance which caused some disappointment, but as the valuation remains in the realm of reality, this only triggered a small correction in the share price.
• FQ3 25 revenues / Adj-EPS were $35.1bn / $0.81 ahead of estimates of $33.2bn / $0.75 as demand remains ahead of what Nvidia is capable of delivering.
• Q4 25 will start to see commercial deliveries of the new Blackwell chip and fears of problems hampering delivery have not materialised in Nvidia’s figures, forecasts or commentary.
• This is because I suspect that the problems are mostly about the infrastructure built around these chips in the data centre, so customers not wishing to lose their place in the queue are buying anyway and fixing the problems later.
• FQ4 revenues are expected to be $37.5bn (+/- 2%) giving a range of $36.8bn – $38.3bn slightly ahead of consensus at $37.1bn but below the most optimistic of estimates.
• Gross margins will also decline slightly in FQ4 to 73.0% to 73.5% from 75.0% in FQ25 as a result of the ramp-up of Blackwell, but this will come back as the chip reaches full volume meaning that margin pressure is not a concern that I have.
• This is because Nvidia continues to sweep its competition contemptuously to one side as the combination of the unique CUDA development platform and being at least one generation ahead of everyone else leads to interest in alternatives remaining low.
• While the control point in the AI training ecosystem remains at the level of the silicon development platform, no one else is going to get a look in but there are signs of development moving to the foundation model which will reduce the power of CUDA.
• This will take time and from a competitive standpoint, Nvidia is in the clear for at least 2025 and maybe 2026 as well.
• For 2025, Nvidia is banking on the AI boom continuing with the emphasis finally moving from training to inference and the use of AI agents to analyse data and execute tasks.
• While I think this use case will work well in the enterprise and the vehicle, I am yet to be convinced that we will start controlling all of our devices with AI agents.
• As always, the devil is in the details and there is a lot of messy software plumbing that is needed to connect the AI agents to the apps and services that they will be controlling.
• Hence, even if this is the next big thing as the AI industry wants us to believe, it will take a long time before the AI agents can execute all of these tasks flawlessly.
• However, I don’t think that this matters very much for next year as I suspect that like Micron, Nvidia is already pretty much sold out till the end of 2025 meaning that it has much better visibility in its forecasts than it is letting on.
• This allows it to offer guidance that it knows it can beat by a small margin, which will help the share price hold onto the massive rally it has enjoyed over the last 18 months.
• It is here where Nvidia’s superlative execution comes into play as the growth in the top line has been more than fully reflected in the profits leaving the company on a high, but not difficult to justify valuation.
• With a consensus FY 2026 EPS of $4.25, the company is trading on 33.6x which for a company with this level of growth and profitability is not a huge stretch.
• This is why I think its failure to raise expectations has not hit the share price harder supporting my view that this is one of the least risky ways to play the AI boom directly.
• Hence, if I were forced to invest directly, it would be Nvidia that I would own but I continue to prefer the adjacencies of inference at the edge and nuclear power both of which have their best growth ahead of them.