Topic: GPU ROI, HPC strategy vs. colocation, and New Brunswick hyperscaler interest
Key points:
GPU ROI is approximately 2.5 years after direct operating costs; GPUs retain 60%–80% residual value after 3–4 years.
HIVE previously operated 38,000 NVIDIA A-series GPUs; sold 34,000 at 80%–90% of face value to fund H100/H200 upgrades.
New Brunswick site: 32 acres adjacent, 53 MW IT load, target $130/kW per month → ~$80 million ARR.
Mgmt stance: Bullish — cites proven GPU operational history, strong residual value, and active hyperscaler discussions for New Brunswick.
Q2 — Fedor Shabalin
Topic: Bitcoin mining breakeven, capital allocation, and CapEx split
Key points:
2025 focus: scaling Bitcoin mining to 25 EH/s via 300 MW in Paraguay.
2026 focus: scaling HPC revenue from $20M ARR to $225M ARR (10x+), via GPU cloud (5,000→11,000 GPUs, $20M→$140M ARR) and New Brunswick colo ($80M ARR).
GPU financing: OEM vendor lease-to-own, 36-month equal payments, single-digit interest rates, $1 buyout — described as “CapEx-light.”
Mgmt stance: Bullish — emphasizes disciplined capital deployment, attractive financing terms, and clear revenue growth roadmap.
Q3 — Mike Grondahl
Topic: GPU lease structuring and depreciation
Key points:
Longer GPU lease terms reduce periodic payments, improving cash flow; optionality to sell at gain or re-rent after term.
Depreciation: Q3 9-month figures include catch-up; incremental Q1→Q3 run rate is forward guide; all ASIC equipment in Paraguay is operational.
Mgmt stance: Neutral — provides factual guidance on lease structure and depreciation normalization.
Q4 — Gareth Gacetta
Topic: CapEx for remaining Manitoba GPUs and data center costs
Key points:
504 GPUs (63-node cluster) purchased in November, contracted last week; being delivered and commissioned with InfiniBand, expected live by end of Q1 (March 31).
Model: rent cluster → order next cluster → finance similarly → deliver → rinse and repeat.
No additional CapEx needed for Manitoba data center; deposits already covered.
Mgmt stance: Bullish — signals rapid repeatable deployment model and no incremental infrastructure cost.