HIVE Digital’s $350M AI Deal Hinges on $185M GPU Investment

HIVE Digital Technologies just landed a $350 million AI cloud deal. Big number. But the whole thing depends on whether the company can spend $185 million on GPUs fast enough to make it real.

The five-year contract runs through HIVE’s BUZZ High Performance Computing unit and was signed with a customer the company hasn’t named — only describing them as investment-grade. The target is roughly $70 million in annual revenue once the hardware is actually running. That’s the plan, anyway. Getting there is a different story, and the path is messier than the headline suggests.

Not yet confirmed: the $35 million upfront payment.

The unnamed customer agreed to pay nearly 10% of the contract’s total value — around $35 million — upfront. HIVE said so publicly. But as of now, HIVE hasn’t confirmed it actually received the money. That’s a pretty important detail to leave hanging when you’re talking about a nine-figure infrastructure build.

What the GPU Buildout Actually Looks Like

BUZZ plans to deploy 2,016 NVIDIA Blackwell Ultra GPUs, housed inside GB300 NVL72 systems, at the Bell AI Fabric facility in Merritt, British Columbia. The hardware, equipment, and service warranties together account for that $185 million cost estimate. It’s a dense, expensive setup — and the whole thing needs to be commissioned by the fourth quarter of 2026 for the revenue clock to start ticking properly.

HIVE CEO Aydin Kilic said the company’s recent debt raises would fund the GPU acquisition. HIVE pulled in $130 million from zero-percent exchangeable senior notes in June, on top of $245 million raised through similar notes the quarter before. That’s $375 million in total from those two rounds combined. Kilic pointed to those funds as the source for the GPU spend, but the company hasn’t disclosed how much of that capital is still available, or what the terms look like for any additional equipment financing they might need. So the math isn’t fully visible from the outside.

HIVE also reported $208 million in cash. But it’s unclear how much of that is actually earmarked for this project versus other operations. The company hasn’t said.

See also: Optimism Redirects 546.9 Million OP Tokens from Airdrops to Strategic Ecosystem Fund

Revenue That’s Mostly Still on Paper

Here’s where things get a bit murky. BUZZ’s total reported annual recurring revenue is now around $180 million on paper. Only $35 million of that is currently active. The remaining $145 million kicks in — supposedly — once the hardware is deployed and operational by Q4. So the bulk of HIVE’s AI revenue story is still contingent. It’s not booked, it’s not flowing, it’s waiting on a deployment timeline that hasn’t been tested yet.

Once everything is running, HIVE projects its HPC and AI operations could generate roughly $500,000 per day. That’s a meaningful daily run rate. But it requires hitting a pretty specific target: hardware delivered on time, commissioned without major delays, and a customer relationship that holds through the buildout period. Any one of those slipping could push the revenue live date and complicate the company’s end-of-year numbers.

The broader AI infrastructure market is competitive right now. Demand for high-performance GPU compute has been intense across the industry, which is probably part of why HIVE can sign deals at this scale. But that same demand makes hardware procurement harder and timelines less predictable. Lead times on NVIDIA’s latest GPU architectures have been a real issue for operators across the board — not just HIVE.

HIVE’s financial structure here is worth watching closely. Zero-percent exchangeable notes are a clever way to raise capital cheaply, but they come with conversion mechanics that can dilute shareholders down the road. The company has leaned on that instrument twice in recent quarters, which tells you something about how it’s managing the tension between capital needs and balance sheet optics. Long-term sustainability of that approach is an open question — and it’s one HIVE hasn’t addressed directly in its disclosures.

The Merritt, British Columbia facility adds a geographic dimension worth noting. Canada has become a real destination for compute infrastructure builds, partly because of power costs and partly because of available land near reliable grid access. Bell AI Fabric is a relatively new entrant in that space, and HIVE’s decision to anchor there says something about where the company sees its infrastructure future.

What HIVE is basically trying to do is use a mining company’s capital-raising muscle to fund an AI infrastructure pivot. It’s not the only crypto-adjacent company chasing that playbook right now. But HIVE is further along than most, with an actual signed contract and a named hardware spec. The execution risk is real, though. A $350 million contract with $145 million in revenue still waiting on a hardware build is a bet, not a guarantee.

The Q4 deadline is the number that matters most right now. Miss it, and the revenue story gets complicated fast. Hit it, and HIVE’s daily run rate jumps sharply. Kilic and the team have until the end of the year to prove the timeline holds — and HIVE’s $208 million cash position, whatever portion is actually available for this project, may not be enough to cover the gap without more financing.

The company has 2,016 GPUs to install and a $35 million payment it hasn’t confirmed receiving.