Bitdeer Technologies

Bitdeer has assembled a ~3.0 GW power portfolio with 1,797 MW already energized, is the only Bitcoin miner that designs and manufactures its own ASICs, and has an AI Cloud business whose ARR nearly quintupled to ~$76M at 95% utilization — the option value on converting cheap power into contracted AI compute is large relative to a $2.77B market cap.

• Operational execution is real and accelerating on the mining side: 990 BTC in June (+388% YoY), self-mining hashrate 73.0 EH/s from 16.5 a year earlier, 243,000 self-mining rigs from 114,000, and average fleet efficiency improved to 16.4 J/TH from 29.0.
• Vertical integration is a genuine structural difference. Designing SEALMINER in-house (A4 launched in Q1 2026) plus the Sparks, Nevada manufacturing facility targeted for end-2026 means Bitdeer is not a price-taker on the single largest capex line every other miner faces — and it can sell rigs to third parties.
• The AI pivot has moved from slideware to signed contracts: a 5-year GPU cloud sales contract with delivered GB300 NVL72 clusters, 3,517 of 4,248 GPUs under external subscription at 95% utilization, and a 10-year lease for 21.7 IT MW in Malaysia to host 128 GB300 NVL72 systems from Q1 2027.
• The Tydal, Norway colocation lease has been executed (subject to conditions precedent) — the specific catalyst management had flagged as 'advanced negotiations with a credit-worthy tenant' in May actually landed, converting 225 MW of pipeline toward contracted colocation.
• The conversion pipeline is concrete and dated, not aspirational: Knoxville phase 1 (37 MW) and Wenatchee (13 MW) to AI Cloud by Q4 2026, Rockdale's 563 MW in active evaluation for AI transition. Powered land with grid interconnection is the scarce asset in AI infrastructure today.
• Upgrading the auditor from MaloneBailey to Deloitte (Singapore) on 2026-08-03, with the filing stating no disagreements and no qualified opinions, is a credibility improvement for a company whose reporting quality has been a discount factor.
• Adjusted EBITDA turned positive ($14.4M in Q1 2026 versus negative $45.6M a year earlier) even as the fleet was mid-ramp, and efficiency gains plus the halving of average rig J/TH mean the marginal EH/s is far more profitable than the average deployed to date.

This is a business selling revenue below cash cost — a gross LOSS of $39.0M on $188.9M of revenue — while burning $440.6M of free cash in a single quarter, carrying ~$1.62B of net debt (of which ~$713M is related-party), holding only 150 BTC, and funding the entire buildout from convertible notes. And the figures above are already four months stale: Q2 does not report until 2026-08-10.

• Negative gross margin is the whole problem. Cost of revenue ($228.0M) exceeded revenue ($188.9M) in Q1 2026. Growing a business whose incremental revenue does not cover its direct cost destroys value with every added exahash, and the gross loss widened tenfold YoY.
• Free cash burn of $440.6M in ONE quarter against a $2.77B market cap. Operating cash flow was negative $346.9M and capex $93.7M; the gap was plugged with $568.3M of convertible notes and $206.8M of digital-asset sales. Access to capital IS the business model, and the equity is the shock absorber.
• Balance sheet: ~$1,920.1M of borrowings against $297.7M of cash. Equity fell from $848.0M to $730.1M in a single quarter and the accumulated deficit is $693.7M. Third-party data puts total debt at ~$2.03B including leases, so enterprise value (~$4.53B) is ~63% debt-funded.
• ~$713.4M — over a third of all borrowings — is owed to RELATED PARTIES, including $517.8M in current maturities. That is a related-party funding dependency at the centre of the capital structure, and it is not arm's-length capital by definition.
• The AI story is being valued off ARR, not revenue. AI Cloud ARR is ~$76M; AI Cloud REVENUE in Q1 2026 was $3.7M. At 4,248 GPUs, this is a rounding error in the P&L against a segment being credited with the re-rating, and 731 GPUs are still not externally subscribed.
• BTC held collapsed to 150 coins from 1,502 a year earlier. There is no treasury cushion left — every coin mined is sold to fund construction, so the company is fully exposed to hashprice with no buffer and no optionality on a BTC rally.
• Inventories ballooned $361.0M in one quarter to $613.0M. If SEALMINER demand or deployment timing slips, that is a write-down risk larger than the entire quarterly revenue line, and it sits ahead of the $1.9B of debt in the cash-conversion queue.
• The information is stale by construction. The most recent financials predate the quarter that just ended, and Q2 lands 2026-08-10 — any conviction formed today will be re-based in under a week. An auditor transition in the same week (2026-08-03) adds a further reporting-continuity risk into that print.
• The market has already voted: -6.1% over twelve months while revenue grew 169%, and the stock trades ~24% below its 50-day average. Growth that the market refuses to pay for is usually growth it does not believe converts to cash.