원문 보기
Here now, at post 9, is CBC in person, CoreWeave founder and CEO Michael Intrator. Hi, Mike, nice to see you. Perfectly. Okay, let me start with a question that I think might end the debate. It struck me as I listened to you that the GPU depreciation discussion, which is key for many, is over. Instead, Nvidia's $500 billion financial platform looks quite alive, despite two skeptics.
I would like you to comment on this, because during your conference no one asked about Nvidia. This is First Bridge. So, first of all, thank you for inviting me. Always happy to come and chat with you after the earnings report. This report is probably more pleasant than some of the previous ones. Last time I talked about how the company is going through this turning point.
And how things will change as we continue to scale the business. A turning point, clarify that. This is a turning point for margins. There are a lot of turning points here, aren't there? And we're already there , right? So, we added 500 MW of capacity, which is an astronomical figure. Our installed base is now one and a half gigawatts, right?
We are a huge provider of this resource. Corporations are starting to implement our platform. Contract margins, contribution margins, as you said, are increasing. And ...One of the discussions that accompanied this industry concerned outdated generations of GPUs. And what we've hopefully put an end to is that older generations of GPUs will have a longer lifespan than anyone expected .
They will be contracted for a longer term and at a higher price. And we are already seeing this reflected on our platform. After the close of the second quarter, in the first 5 weeks, we signed additional contracts worth over $25 billion . You know, to get a sense of scale, $25 billion is practically the size of our order book a year ago.
This gives an idea of how high the demand for our product is and how fast it is growing. Nvidia and the financial companies that have joined in creating this financial platform are, in my opinion, a confirmation of our business model and the demand signals that everyone is starting to see and that we have been talking about for quite some time.
This inspires us very much. But when we first met, you told me that these chips would remain valuable for another 5-10 years or that you would have residual value after that. It turned out that this was not entirely true. They turned out to be more valuable than at the time of purchase , and these are very old chips, released in 2020.
Now you can earn money on them until 2029. Um, I want to know: is it just the Nvidia chip, and others might not be like that ? Because it seemed like you were saying over and over again what a great partner this is. But I want to make sure people understand: this is "currency," not just any old chip.
Yes. So, we look at it as a combination of three things, right? This is an Nvidia GPU, which is the best solution on the market. It's Nvidia's software, the CUDA platform, that enables this interchangeability, which is extremely important. And the final element is delivery via CoreWeave Cloud, which has the best software solution on the market for providing this infrastructure.
When you combine these three components, you can extend the lifespan of that infrastructure. You can set a premium price. And as you said, we have contracts for the 2020 architecture all the way through 2029 at full rate. This is simply an incredible indicator of the market demand for computing power and for CoreWeave's solutions for how to provide that power.
Um, can you provide more details on this 25+ billion that you contracted in the five weeks after the end of the quarter? Customer concentration, I don't know, maybe some names or details you could give us. So, I think this contract reflects our overall order book pretty well, right? It features corporate clients .
It features fundamental laboratories. It also includes representatives from AI startups, as well as industrial enterprises. So we are very excited. Financial indicators are also reflected there. It's a really great mix of customers coming in to buy the solution that we offer. So that's pretty exciting.
You know, it's important to understand that not only are we signing contracts at a faster pace, but margins are also increasing, which will ensure the long-term sustainability of our business model. During the reporting conference call, I talked about a 5-10% increase in contribution margin and operating margin associated with these specific deals.
This will have a lasting impact on our ability to run our business profitably. technologies driving the market. Nvidia is entering a new phase of the AI race. The problem isn't just that competitors are creating faster chips. And the fact is that more and more companies are developing their own specialized chips.
And Nvidia's answer is surprisingly simple. Instead of stopping this trend, it is trying to become a platform that unites everything. One example is D-Matrix. This startup is developing a specialized inference processor designed for AI workloads. But rather than considering these chips a direct threat, Nvidia is working to make specialized processors work with its systems .
This is where NVLink Fusion becomes important. The idea is to allow dedicated XPUs and accelerators to connect to Nvidia's broader computing architecture. So customers could potentially use Nvidia for certain tasks. And accelerators from another company are for others. There have been changes in Nvidia's strategy.
It doesn't have to produce every processor inside the AI system. It can provide an infrastructure that allows different processors to work together. And the reason why this is important is the " memory barrier." AI models are becoming huge, and data transfer between the processor and memory can become as important as the computing power itself.
That's why technologies including high- bandwidth memory, advanced packaging, and even 3D- stacked DRAM are becoming increasingly important. Nvidia's future platform could be designed with this reality in mind. Future AI systems are not just boxes filled with GPUs. This will be a huge combination of processors, memory, networks, software, and data center infrastructure.
And Nvidia wants its technology to be at the center of this ecosystem. But there is another important development related to Groq. Nvidia's relationships with companies that produce AI chips have attracted the attention of regulators, including the US Department of Justice. This shows how strategically important the AI accelerator market has become.
Competition is no longer limited to chip companies. The question of who controls the infrastructure for AI is becoming increasingly acute. And as the hardware race continues , AI security is becoming another important issue. Anthropic recently released a detailed report that explores and addresses potential threats from AI.
As models become more sophisticated, the discussion shifts from what AI can do to what will happen when it can do much more than humans expect. At the same time, AI goes beyond chatbots. Physical AI is becoming a priority area of development. Robots, automated machines, industrial systems, and other applications may eventually require enormous computing power.
This brings us to the fact that Oracle has sent another important signal about the high demand for AI infrastructure. The company's cloud infrastructure revenue has skyrocketed, and its contract commitments have reached hundreds of billions of dollars. The company also announced plans for a significant share issue to support its expansion.
The main message is clear. The demand for AI is spreading throughout the entire technology chain. It's not just Nvidia .
Comments 0