Aug 17, 2026 earnings call — HIVE pushes GPU cloud contracts, expands AI centers
HIVE announced $180 million of contracted GPU cloud revenue including a new 5-year deal with $70 million ARR.
Revenue $0.1B (est $0.1B) · EPS $-0.54 (est $-0.21)
Key quotes
"Our team has done a tremendous job, and we are at $180 million of contracted revenue because we just announced a new 5-year deal, which has $70 million ARR to our HPC business unit."
"last year, we substantially increased our hashing power in Paraguay, in particular, taking overall complex from 6 exahash to 25."
"The net loss included an $84.7 million noncash provision related to regulatory liabilities associated with our ongoing Swedish VAT dispute."
Analyst Q&A 7
GPU cloud deal and Paraguay update
Q (Joseph Vafi · Canaccord): Analyst asked about returns (IRR) on the 5-year GPU deal, CapEx to build the cluster, financing plans, and next steps in Paraguay power build.
A: Management said the GPU cluster costs about $185 million and will follow NVIDIA reference design for 2,016 GPUs. They expect delivery and deployment in Q4, EBITDA around 75% to 80%, paying off GPUs in about 3 years and then generating free cash flow; they put about 20% down and will provide more financing details later. For Paraguay, they are working on basis of design, expect the substation energized by year-end, and will share updates as design progresses.
"And so if you kind of do the math indicatively model, of course, it will be delivered and deployed in Q4." — Aydin Kilic (CEO)
HPC gross margins and Bitcoin view
Q (Christopher Brendler · Rosenblatt): Analyst asked target gross margin for the high-performance compute (HPC) business as newer GPU contracts ramp, and then asked for views on the Bitcoin mining business and big-picture thoughts on Bitcoin.
A: Management said margins should improve as the fleet moves from older A-Series cards to newer, more power-efficient GPUs and as scale lowers fixed costs. On Bitcoin mining, they said Paraguay operations are running new-generation gear at about 19.5 exahash with power advantages, producing strong uptime and cash flow, but macro catalysts will determine Bitcoin's next breakout.
"So our fleet right now is comprised still of -- we've got about $20 million of our $35 million active revenue coming from series GPUs, a combined total of 844 H200 and H100s." — Aydin Kilic (CEO)
Big Boden lease and demand/pricing trends
Q (Mike Grondahl · Northland): Analyst asked whether anything major remains to finalize the Big Boden definitive lease and for demand and pricing trends for GPUs and colocation.
A: Management said the Big Boden lease process is advancing with drafts exchanged and nuanced terms being refined; they expect to update the market possibly in September. On demand and pricing, they said demand is very strong across customers for both latest and prior-generation GPUs, cited examples of higher prices achieved, and said they see demand beyond their $200 million ARR target.
"So the sublease for Big Boden -- sorry, the lease for Big Boden, that process has been advancing." — Aydin Kilic (CEO)
Choosing cloud versus colocation
Q (Bill Papanastasiou · Chardan): Analyst asked how the company weighs building GPU cloud clusters versus offering colocation, given their power capacity.
A: Management explained they forecast 10,500 GPUs to deploy with Bell Canada AI Fabric and have pipeline capacity across Canadian sites for hundreds of megawatts. They compared economics per megawatt showing much higher ARR and enterprise value density for GPU cloud versus HPC colo, and said they will evaluate each site's best opportunity including long-term offtake contracts.
"So what we forecast is we've got a target of 10,500 GPUs that we plan to stand up, and that is with our partnership with Bell Canada AI Fabric." — Aydin Kilic (CEO)
GTA site timing and build choices
Q (Brett Knoblauch · Cantor): Analyst asked when they must decide to break ground at the Greater Toronto Area (GTA) site to meet a 2028 readiness, whether they must pick cloud or colo before breaking ground, and how they would fund build-out.
A: Management said design and permitting are underway and they will provide updates; the project is three phases of 80 megawatts each. They expect the site to be energized end of 2027 with compute live in early 2028, and financing choices (including corporate bonds and signing offtake agreements) will guide whether parts are cloud or colo.
"we expect the site to be energized end of '27 with compute live in early '28" — Aydin Kilic (CEO)
Second-half CapEx and funding for GPUs
Q (Michael Colonnese · H.C. Wainwright): Analyst asked for estimated CapEx in H2, timing to secure funding to pay for GPUs, and details on down payment targets and financing terms.
A: Management said the NVIDIA reference architecture for the announced cluster is about $185 million, and they plan to put roughly 20% down, which is just under $40 million. They said vendor GPU financing with blue-chip lenders typically covers the remainder with lease terms shorter than the contract, and they will share more details closer to deployment.
"So if you're putting, say, 20% down on that, it's just under $40 million." — Aydin Kilic (CEO)
$84.7 million tax liability timing and plan
Q (Stephen Glagola · KBW): Analyst asked timing for any cash payment of the $84.7 million tax liability, options to mitigate or defer it, and how management would fund it if due.
A: Management said they do not plan to pay the claimed tax liability and have already paid about $50 million of tax in the normal course. They are appealing and exploring remedies (including escalation to the European Union) and consider paying this claim not a good use of shareholder capital; details are in the press release and Darcy's section.
"But really, we do not think that, that's a good use of shareholder capital to pay down this, in my opinion, an egregious tax claim, and we've paid all the taxes in the normal course of business." — Aydin Kilic (CEO)