Crypto miner Hive CEO anticipates $100M annual revenue from GPU computing
Hive CEO says the company is in talks with four companies and four marketplaces as it bolsters its machines that formerly mined ETH
Diyana Dimitrova/Shutterstock modified by Blockworks
Hive Digital Technologies has ambitions to exponentially grow revenues derived from its graphics processing unit (GPU) card capabilities to $100 million annually.
The firm is in talks with companies and marketplaces as it moves to capitalize on — and strengthen — the capabilities of its machines formerly used to mine ether (ETH), Hive CEO Aydin Kilic told Blockworks.
Hive’s name change — from Hive Blockchain Technologies to Hive Digital Technologies — took effect Wednesday. The idea: to reflect a newfound focus on the growth potential of artificial intelligence (AI).
The Vancouver-based firm, previously one of the largest publicly traded miners of ether, mined 7,675 ETH during 2022’s second quarter — just before Ethereum’s switch to an underlying proof-of-stake consensus model.
The company has a fleet of 38,000 Nvidia graphics processing unit (GPU) cards, according to a Wednesday filing, presenting the units as a new revenue source.
“Now that Ethereum has gone away, we have a new lever to experiment — or play — with,” Kilic said.
The company in February made plans to launch its Hive Performance Cloud business in the second quarter, adding its proof-of-concept to utilize its fleet of GPUs, which produced annual revenue on a run-rate basis over $1 million.
Ultimately, Kilic said, it could bring in $100 million annually for Hive.
“We’ve now proven this concept…and so we’re looking to scale this business,” he said. “I’m hoping we could hit something like $15 million a year revenue this year alone, but it could be bigger quicker.”
The road to scaling
Hive’s GPUs support Tensor Core and Ray Tracing core architecture, enabling the execution of AI and high-performance computing tasks.
“The next thing we need to do is populate them in more robust servers that have better [central processing unit] clock speeds, more threads, more storage,” Kilic said.
The company’s purchase of $6 million of SuperMicro servers, holding 10 GPUs and featuring 7.6 terabytes (TB) of storage, are set to arrive in the next month.
They will initially run out of Hive’s datacenter in Sweden — with discussions underway around a North American expansion — and they’ve been positioned as offering additional computing power and storage for Hive.
Hive hopes the servers will help it reach $15 million in annualized revenue this year.
“That’s like a four-month [return on investment], which is way better than crypto mining right now,” he said. “We’re getting clusters and clusters of these servers. When they’re clustered together, you have massive computing resources.”
One way to realize revenue from the Nvidia GPUs is to list them for rent on a marketplace.
The other route — via Hive Cloud — is working with companies that might need weeks or months of processing power for rendering or other tasks.
Hive is currently talking with four marketplaces and four companies, Kilic said, declining to specify them.
Use cases to support
Hive has cited demand for ChatGPT, medical research, machine learning and rendering as reasons for its bullishness on the growth of its computing business.
ChatGPT — a natural language processing tool introduced by OpenAI in November — reached an estimated 100 million users months after its launch.
Another emerging technology is zero-knowledge (ZK) proof, a cryptographic protocol allowing a party to prove the validity of a statement without disclosing any additional information.
A significant obstacle in using ZK proof for online applications is the performance overhead of its proof generation, according to the Association for Computing Machinery.
Kilic said this area is not yet something that impacts day-to-day crypto users, but added that Hive has begun researching it.
“ZK proofs, I think, will be an integral aspect of the next generation of scaling layer-1s, but you need the confluence of multiple areas,” the CEO said. “You need the computing power, you need the algorithms that drive those computing powers and then you need the application layer where there are projects where there are lots of users.”
Don’t miss the next big story – join our free daily newsletter.
