Bitdeer Technologies Group

Bitdeer Technologies Group (Nasdaq: BTDR) is a Singapore-headquartered, vertically integrated Bitcoin-mining and digital-infrastructure operator founded in 2018 by Jihan Wu, co-founder of Bitmain. Its FY2024/FY2025 Form 20-F describes a business that spans five layers: datacenter development and operation, proprietary self-mining, cloud hashrate and hosting services, and — uniquely among miners — its own mining-ASIC silicon. The SEALMINER roadmap (the 4nm SEAL01 chip rated ~18.1 J/TH, with SEAL02/SEAL03 generations following) is the company's bid to become the first fully integrated miner, controlling everything from chip to power to hashrate.

The asset base is a global fleet of owned electrical capacity. Per the June 18, 2026 Form 6-K (May 2026 operations update), Bitdeer controls 3,003.5 MW of total global electrical capacity across the US (Rockdale TX, Massillon/Clarington/Niles OH, Knoxville TN, Wenatchee WA), Bhutan (Jigmeling 500 MW, Gedu 100 MW), Norway (Tydal, Molde), Ethiopia (Oromia), Canada (Fox Creek AB) and Malaysia (Cyberjaya). Roughly 1.47 GW is energized and operating today — almost entirely Bitcoin mining — with the balance in a contracted/pipeline tier earmarked for crypto-to-AI conversion.

The pivot underway is from hashrate to AI infrastructure via two routes: colocation (leasing converted, energized halls to AI tenants) and a still-nascent owned GPU cloud. As of May 2026 the AI Cloud business runs 4,248 GPUs (H100/H200/B200/GB200/GB300) at ~90% utilization for roughly $69M of annualized run-rate revenue — real, but a rounding error against a ~$750M-annualized self-mining business. The company self-mined 921 BTC in May 2026 (+370% Y/Y) on 70.2 EH/s of self-mining hashrate (83.1 EH/s total under management).

The power thesis is straightforward: Bitdeer already owns and energizes the scarcest input in AI infrastructure — interconnected, cheap megawatts — and bought most of it years ago at crypto economics. Converting an energized 225 MW hall in Tydal or a 563 MW campus in Rockdale into AI colocation is faster and cheaper than greenfielding a substation queue. If even a fraction of the 1.35 GW contracted tier signs colo leases at hyperscaler-grade rates, the revenue and margin profile re-rates entirely away from Bitcoin's price beta. The vertical integration — owning the land, the power contract, the datacenter, and increasingly its own silicon — is the structural edge.

The bear case is that, as of mid-2026, almost none of this has happened yet. The online 1.47 GW is ~entirely crypto; the lead colocation deal (Tydal, Norway) is still only 'in advanced negotiations' with no signed lease; and the largest contracted block — Clarington's 570 MW (~42% of the contracted tier) — sits on leased Monroe County Port Authority ground and is under active litigation, with American Heavy Plate Solutions (MHR-backed) seeking a permanent injunction to bar construction. Niles' 300 MW is grid-secured but not targeted to energize until Q4'28. Crypto-to-AI conversions also derate MW (electrical capacity is not IT/critical AI load), so the headline gigawatts overstate deliverable AI compute.

Underneath sits the real pressure: ~$1.9B of borrowings against $297.7M of cash and a Q1'26 net loss of $159.5M. Bitdeer is funding an AI buildout, a SEALMINER capex cycle, and self-mining expansion simultaneously off a balance sheet leaning on convertibles and related-party debt. The bull needs signed colo leases and energized AI load before the cash and the conversion window close; the bear sees a levered miner with a great map of megawatts and very little contracted AI revenue to show for it yet.

Owns the full stack — land, power contracts, datacenters, ASIC silicon, and a small GPU cloud — but today the AI monetization is mostly a future colocation-lessor play plus a tiny (~$69M ARR) neocloud. Online capacity is ~entirely crypto. Leans toward owning power and converting it, not yet a pure GPU operator.