Bit Digital Inc (NASDAQ:BTBT) earlier this week announced the expansion of its contract with an existing high-performance computing (HPC) customer, increasing the contract's total value to $275 million. CEO Sam Tabar highlighted that the client has doubled the fleet size of GPUs, now contributing approximately $92 million per year, locked in for three years.
The company provides specialized computational power, crucial for clients developing large language models (LLMs) and artificial intelligence (AI) applications. In this interview, CEO Tabar explains that the expansion includes deploying Nvidia H100 chips and servers from Dell and Supermicro.
Proactive: Sam, it’s good to speak to you. You've announced the contract expansion with an existing high performance computing (HPC) customer, taking the value to about $275 million. The market clearly likes the news, sending your shares significantly higher last week. Tell us more about what you'll be supplying.
Sam Tabar: We're renting basically HPC, or very specialized computational power, for our clients. We started that process in January. We've deployed about 2,048 GPUs in January. That symbolized about $50 million run rate annually. The client has been very happy with our services and has doubled the size of the fleet, which now represents about $92 million per year, locked in for three years.
That makes it about $275 million irrevocable, locked in. We draw that money every month.
So tell us what you’ll actually be supplying to the customer.
It's basically a specialized computational power for the customer who's building a large language model. People who are building LLMs need specific, very high-performance computing power in order to build their large language model. You can't just get that from a regular computational power. So we're renting that to them specifically. This particular client had needs to have a certain kind of data center geographically and a certain network, and we were able to do that.
We offer a very customized approach, in order to win clients who are building their LLM’s, clients who are building their AI applications, who need this kind of computational power that we rent out. We rent that out using Nvidia, and in this case, in the second tranche was Dell, but the first tranche was Supermicro. But the core of it are these Nvidia chips called H100s. They’re basically the most powerful computers on Earth today. We have special access to our procuring relationships with these equipment providers, and we have special access to certain data centers in which only these kinds of data centers can host these kinds of machines.
And, of course, we have the clients, so it's the Holy Trinity: Client, Data Center, and the specialized computer machines.
Of course, you've already placed an order with Dell for those servers. They're going to be delivered to the data center in Iceland. So you're planning to get moving on this pretty quickly?
Yes, that's right. When we first announced a contract last year, the first tranche of the contract was announced in October, we were basically fully deployed by January. We've announced the finalization of the terms for the second tranche of the contract, and now we intend to be deployed in August.
The deal takes you closer to reaching your targets of $100 million in annualized revenue by the end of 2024. Are you seeing more opportunities in the HPC AI vertical?
Absolutely. We have a very pregnant pipeline of clients who are knocking on our door and we're finalizing terms with some of these clients. It's all been reverse enquiry – we have not had any sales team. On that note, we do plan on hiring a head of revenue just to keep the process organized. It's a champagne problem to have a lot of pent-up demand knocking on the door for our services, but at the same time, we do need a less chaotic internal process to win this business. I am excited to be announcing the other clients in the future.
Quotes have been edited for style and clarity