Kristina Ayanian
From Nasdaq’s MarketSite, I’m your host Kristina Ayanian, and joining me today is Ido Susan, co-founder and CEO, and Hillel Kobrinsky, co-founder and Chief Strategy Officer at DriveNets. Gentlemen, thank you so much for joining and welcome.
Good morning, thank you.
Series D $410M funding round
Kristina Ayanian
Now we have a lot to celebrate today. Congratulations on your $410 million Series D, that’s absolutely incredible. Talk to me about the funding and why it’s so important for the company.
Ido Susan
Thank you so much. Basically, this is not a regular funding. We was able to secure a significant business, over $1 billion of business that we need to deliver to our customers. In order to support the demand, we did this round. We’re going to build the inventory and the supply chain that we need and committed to our customers. DriveNets is already cash flow positive. So we don’t need to raise money to grow the company. This is what’s specific for our serving our business.
Hillel Kobrinsky
This is not kind of fundraising round. This is kind of delivery round. So it will help us, as Ido mentioned, to give our customer the best support, to build the inventory and provide them the fabric in time.
Kristina Ayanian
Absolutely. And you mentioned AMD as a strategic investor. What does this signal both about DriveNets as a company, but also the broader market shift towards open multi-vendor AI infrastructure?
Ido Susan
Today, the customers are looking for diversity and the AI infrastructure is growing so fast, so they must do multiplayers. It’s not any more question that AMD is going to be, and already, one of the main leaders in this infrastructure. So supporting together with them, this is, I think, the opportunity not only for us, but also for the end customer that’s looking for an alternative to NVIDIA and other players.
The most expensive idle asset in the world today is a GPU waiting on the network
Kristina Ayanian
You’ve mentioned before that the most expensive idle asset in the world today is a GPU waiting on the network. Talk to me more about this and what do you mean by that?
Ido Susan
Infrastructure today building from tens of thousands of GPUs that the networking basically connected all of them to be like one infrastructure. The networking is built from the physical layers and the logical layers that it’s the software stack. Today, many of the GPUs are in idle waiting for the networking, not because the physical is not connected, because the software stack is congested. DriveNets solution, basically, this is what we are solving. We are solving those bottlenecks. And we are optimizing and increasing the performance dramatically across all the GPU infrastructure.
DriveNets journey
Kristina Ayanian
You were an early mover in the space. Going back to when you started the company, did you ever think that it would lead to where it is now?
Ido Susan
No, nobody can think that the AI will be so mature in our life. So yes, we are somehow lucky because, like Hillel said, our technology that we designed for, built from scratch for the internet service provider supporting all those massive capacity, all the increasing reliability, security, everything that we are working and deployed in thousands of locations. Now we replicated to their infrastructure that need basically the same capability, and we coming with the muscles of the operation experience and serving those customers. But nobody was expecting that this is will be the next generation for DriveNets.
Hillel Kobrinsky
The dramatic shift, you can take it from the day we started we couldn’t raise money, so we brought our own money from, from home. And now we raised the $410 million. So it’s been a great journey for us.
Kristina Ayanian
It’s amazing. It’s hard work and a little bit of luck.
Hillel Kobrinsky
You need both of them.
What is heterogeneous AI
Kristina Ayanian
What is heterogeneous AI and why is that the next shift?
Ido Susan
I think this is — and the industry believe this is going to be the next generation for the AI infrastructure, basically. The workload of AI, it depends on the use case. And you don’t have specific chip that optimize to all the use case. When I’m doing chat with ChatGPT, I have a human that talking with machines. So I have some different performance and different chip that the time to first token, it’s very important. When we’re doing agentic, that you have agent, less important the time to first token, more important the parallelization. And in the future, we’ll have inference for physics, for robotics. Each of them require a different chip, different resource, different capability. Heterogeneous, basically, it’s take the AI infrastructure and to design it and to optimize that you will do the workload of the AI in the most optimized chip and capability. So DriveNets plays very critical element of the heterogeneous architecture. We already working with many very leading, you know, chip companies. I can mention AMD and Broadcom and others. And definitely we see between 3 times to 10 times better performance when you’re doing it heterogeneous compared to the infrastructure that you have today.
Hillel Kobrinsky
The first wave was about give me as much compute as you can, and it was like a brute force of compute. Now we are coming into the next wave when you want efficiency, you want cost per token per watt. So this is why heterogeneous is right now and why we are a good fit to it.
Kristina Ayanian
That’s amazing. Well, gentlemen, congratulations again. Looking forward to having you back and following your growth.
Thank you very much. Have a great day.
