Interview: Post-Earnings Insight With Nvidia CFO Colette Kress

After Wall Street closed the markets for the day and Nvidia reported its financial results for the second quarter of fiscal 2025, we had the opportunity to chat with Colette Kress, chief financial officer of the accelerated computing giant.

We wanted to get a better handle on how the delays with the “Blackwell” GPUs would affect Nvidia’s financials, how its software business is different from that of other platform providers, how Nvidia manages for exponential growth, and how the generative AI revolution might expand the IT budgets of the world as well as cannibalize some of them.

Timothy Prickett Morgan: I have limited time with you, so I want to dive right in. How much did the Blackwell redesign push out the Blackwell ramp, and how much revenue shifted, if at all, from fiscal 2025 to fiscal 2026? My guess is it doesn’t matter at some level because Hopper can be made at a higher profit margin and people have to buy two of them for every one of Blackwell, and they will buy what they can get today.

Colette Kress: The reality is, it really didn’t change and impact anything because the demand is still there for systems and GPUs to continue the work that they are doing in building out for generative AI and building up for accelerated computing.

We have not reached all the enterprises that we need to. If you actually continue to go to your top cloud service providers, there’s nothing to provide you. They have sold it all. Everybody else is using it. And so this transition from Hopper to Blackwell is an important one, because Blackwell is amazing, but they’re still buying because if they don’t, they’re behind. People still need to start this journey, and that’s why Hopper is continuing.

TPM: One other clarification. When you speak of “several billion” in sales driven by Blackwell for fiscal Q4 in the CFO commentary, I am going to guess that several means more than $2 billion for sure and at least $3 billion. . . .

Colette Kress: I have this rule. When I say single, it is one. And a couple is two, like you got married or you have two cookies. Several is more, and that you’ve got it correct.

TPM: In talking to Jensen a few years back, he reminded me, or admonished me, that Nvidia is not really a hardware company. It’s a software company, and 75 percent of the employees at Nvidia are working on software. There are some people working in management, of course, but the rest of the people are working on hardware.

I bugged him about this then, and I will bug you about it now because this would be your job if you did it.

Can you extract the software value out of the platforms and more accurately reflect the inputs and the outputs for your software efforts? I know you are growing a software business separate from that, but you know, in the relational database world, for instance, the hardware costs X and the database and middleware license is 10X. Nvidia is doing the opposite.

You gave the software away for a long time as part of the hardware, and now you’re building this incremental revenue stream. Will Nvidia transform into a software company that also does hardware at some point, does it will ever go 50-50 or something like that?

Colette Kress: That’s a really good summary. And let me add a little bit more color to the situation.

Moving to accelerated computing and moving to AI is not a situation where you turn the hardware on and you just load your stuff on and it goes. The software that we built was to create the path for companies to do accelerated computing versus their initial general purpose computing. You need a different path. Most of our software, when we say we gave it away, was to fuel that transition to accelerated computing.

The reason why we have so much software engineering is because they have to redo workloads for them to move to accelerated computing. That’s why we have so many that are stitching that together for customers. Which is why competitively it is so difficult. We will have a competitor that says, “Here’s my chip,” and a company will go, “What am I going to do with that? I don’t have anything surrounding it. How do I even boot this?”

Our software is so important because you have to transition how the workflow changes. If it was going directly to the CPU, you have to reroute to use the GPU.

So our process now says we continue to build to help every industry – top workloads, top applications – be able to move to move to accelerated computing. Now we have a different thing where customers are working on generative AI, and they have a model and they need help advancing that model, making it work perfectly and provide services to make sure it has the right security, to make sure it has the right level of approval, the right level of overall operation.

Software will be an important part of Nvidia, whether we sell it, or whether or not we offer free in order to help them move. Enterprises don’t necessarily write their own software. There are a few top large ones that do, but most of them are buying from others. In order for them to put these GPUs and AI on mission critical applications, they have to ensure that they have someone managing that software. That’s us. We have to manage it. We have to keep it secure. We have to keep it continually upgraded.

TPM: Exponential growth can’t go on forever. We know that, but it can go on for a long time. As I look at Nvidia’s growth in recent years because of accelerated computing in general and generative AI specifically, it is easy to be enthusiastic about exponential curves, but there is also a desire to be practical about how long it can go on.

As CFO, how do you plan in this kind of climate? Do you plan for the exponential at this point and hope for the best? Because it is very hard to know how this will all play out. This is a unique situation.

Colette Kress: The biggest part of planning for this size of company is planning our capital to design more products and build more products for all different parts of our business.

You are correct, I am always looking forward to make sure our planning is assuming the best that I think could be there without going the other direction – and being ready to move fast enough. Do we have all of our pieces in order? We do not waste a day, we don’t waste a dime. And those two things are really important about us, our agility to move fast, but our ability to not waste things along the way.

We are in the early stages of this journey of accelerating computing. We are not going to get this all done in two years. These transitions are two decades, three decades of work to change to accelerating computing and to infuse AI in everything that we do.

Do we know what every day is going to be? Is it all going to grow at this same lovely speed? No. But we know it is going to be with us for decades. And if you keep with that understanding and vision and just continue to plan to make sure that you can do it at speed and meet the expectations on how fast things need to move, that’s all we’re going to do on the planning. And we do plan a lot.

TPM: This is a related question. When I first started out in this IT racket nearly four decades ago, IT spending was 1 percent or 2 percent of revenue for most companies. It wavered a little bit based on the industry and the size of the company – bigger companies spent more, different industries like computer services and financial services spent more. During and after the Dot Com Boom, the average for IT spending rose to between 4 percent and 6 percent of corporate revenues.

As I think about this GenAI wave, I am wondering does this new workload force companies to go to 8 percent or 10 percent of revenues for the IT budget, and therefore it expands the amount of aggregate IT spending faster than we might expect, or does it just cannibalize the current spending?

Colette Kress: So there’s a little bit of both – making the investment now, knowing that it’s not an instantaneous solution.

There is work that goes in to complete the generative AI application, move it to where it is being used. Every day, more data will come and they will work on that. So you will be making sure that you are moving faster, because if not, you’re going to be left behind. That is one piece.

But the question is: Do I just do more? Yes, but again, you will look to find this is your most important work to do and decrease in the other areas that you are investing. I can invest in that, but that is not going to give me any of the productivity improvement of what I know AI can do for me.

For example, general purpose X86 CPU computing doesn’t necessarily have a great return, okay? And so do I need to refresh that, or just keep them plugged in and they’re going to continue to be fine for probably some years, and move my allocation of where I spend my time and my money on focusing what is coming in the next two decades, not what is already in the past.

So there will be two things:  More, because it will be more productive and it will be more efficient for them, and then equally, you will stop doing things that are not providing productivity.