Phase Space AI

Research

Bitdeer Technologies Group [BTDR]

Bitdeer Technologies Group [$620.253m (FY2025, 20-F, IFRS) — no interim XBRL exists] — Research

As of: 2026-07-29 · Spot: $8.92 (2026-07-29 close, Alpaca) · Screen spot: $10.27 (2026-07-28) CIK: 0001899123 · EDGAR SIC: 6199 (Finance Services) — MISCLASSIFIED · FY end: 31 December · Listing: Nasdaq Primary filings used: 20-F FY2025 0001213900-26-049770 (filed 2026-04-30); interim reporting via 6-K only — no 10-Qs, no quarterly XBRL Framework: v1.5.1 / criteria.md 2026-07-29 · Cluster context: CLUSTER_Analysis.md

This memo issues no position verdict. It scores Criteria and outputs an analysis. The book decides.

---

1. Headline verified numbers

Pure-play benchmark for the last row: CoreWeave 6.6x, WhiteFiber 10.9x on the same measure and the same date.

---

2. Mechanism — the named, evidenced driver

Bitdeer is the least converted name in this cluster and its genuine mechanism is not the one the AI narrative implies. AI cloud services were $6.769m of $620.253m of FY2025 revenue — 1.1%.

What actually changed in FY2025 is a semiconductor business:

• Sale of mining rigs and accessories: $108,328,000 in FY2025, against $585,000 in FY2024 and $2,000 in FY2023. From effectively zero to 17.5% of revenue in one year. Bitdeer designs and manufactures its own ASICs (the SEALMINER line) and is now selling them to third parties — i.e. competing with Bitmain, the supplier every other miner in this cluster depends on.
• Research and development of $153,876,000 in FY2025, up from $76,946,000 — 24.8% of revenue, and more than double the combined R&D of any other name here. This is chip development spend, not datacenter spend.
• Mining rigs on the balance sheet went from $67,324,000 to $620,667,000, and inventories from $64,888,000 to $251,999,000 — consistent with both a self-mining fleet expansion and a hardware business carrying stock.

The legacy businesses are in structural decline, which the revenue schedule shows plainly (see §3).

Recorded explicitly: the ASIC hardware business deserves a semiconductor comparator set, and this analysis has not built one. Valuing a chip designer against datacenter multiples would be the same category error this cluster exists to expose, merely in the opposite direction. It is the largest unexamined opportunity in the BTDR file.

---

3. Accounting quality — is the reported growth real?

The revenue schedule — the whole BTDR story in one table

From the 20-F's schedule of revenues, all three years:

Three things are simultaneously true and only one of them is in the AI narrative. Self-mining more than doubled. A hardware business appeared from nothing. And four legacy service lines lost 95–98% of their revenue — hash rate subscription went from $40.3m to $0.8m, electricity subscription from $27.4m to $1.3m, cloud hosting from $3.2m to $0.1m, general hosting down 64%.

AI cloud services grew from $3.450m to $6.769m — an increase of $3.3m on a $620m revenue base. At the pure-play anchor of 6.6x that business supports $45m of enterprise value, 1.4% of BTDR's actual EV of $3,207m.

Is the reported growth a bitcoin price move? Yes for the 64% of revenue that is self-mining

Self-mining is $396.046m of $620.253m — 63.9% of revenue — and it is bitcoin price × hashrate share. The history makes the point without needing a decomposition: total revenue was $368.554m in FY2023, FELL to $349.782m in FY2024 (the April-2024 halving year) and then rose to $620.253m in FY2025 (the price run). That is not a growth curve; it is a bitcoin chart with a hardware line added on top.

And BTDR gives no way to check the recent trend. It files a 20-F annually with 6-K interims and no quarterly XBRL exists. The most recent data point available anywhere in companyfacts is 2025-12-31 — seven months stale as of this analysis. Every other name in this cluster could be checked for sequential deceleration and every one of them showed it (IREN −21.6%, HIVE −23.9%, HUT −19.7%, WYFI −7.0%, SHAZ −9.6%). BTDR cannot be checked, and the base rate in its own peer group is that the trend has turned. That asymmetry is itself a finding.

Strip the marks — the FY2025 "profit" is a derivative gain caused by the share price FALLING

And the composition of that decisive line:

BTDR's $65.597m FY2025 profit is entirely a $444.861m non-cash gain on derivative liabilities — the conversion options embedded in its convertible notes, which gain value for the issuer when the share price FALLS.

Strip it and FY2025 is a loss of approximately $379.3m.

This is the mirror image of Core Scientific in the same cluster. CORZ reported a $1,155.3m quarterly loss because its share price rose (its emergence warrants are a liability). BTDR reported a $65.6m annual profit because its share price fell. Two companies, opposite signs, both non-cash, and neither number is earnings. Any screen ranking this cluster on reported net income, EBIT, F-score or accruals is ranking on the share price it is trying to forecast. The pair is the strongest single demonstration of that in the whole cluster.

Note what else the table shows: gross margin fell from 21.1% to 9.8% across the three years, and R&D more than doubled to 24.8% of revenue. Before any mark, BTDR's underlying operations deteriorated in FY2025.

The balance sheet — a current-debt wall

Total borrowings of $1,001.091m, of which $753.792m is CURRENT — and $275.0m of that is owed to a related party — against $149.352m of unrestricted cash. Cash fell $326.9m (69%) during FY2025.

Two further items deserve naming. "Cryptocurrencies — receivables" of $135,558,000 appeared from nothing — crypto owed to BTDR by a counterparty rather than held; the counterparty is not identified in the extract reviewed. And prepayments and other assets of $698,291,000 (up $406.4m) is the second-largest asset on the balance sheet, consistent with ASIC supply deposits but unexamined here.

Dilution

+110% in two years. Treasury equity also grew from $160.9m to $325.6m — BTDR is buying back stock while issuing it, and while carrying $753.8m of current borrowings.

A tagging hazard that would corrupt any per-share series: WeightedAverageShares is reported as 12,662,126,000 for FY2020–FY2022 in the FY2022 20-F and as 108,681,000 for the same periods in the FY2023 20-F — a pre/post-reorganisation restatement. The FY2025 6-K reports 192,095 for a period where the annual figure is 137,426,000, i.e. thousands versus units in the same tag family. BTDR's share tagging is internally inconsistent across filings and must be normalised manually.

---

4. Power cost and energy contracts

Not established. BTDR's power contracts, indexation and curtailment arrangements were not determined from the FY2025 20-F material reviewed in this pass.

This is a genuine gap and it is recorded as one rather than filled with inference. What can be stated from the figures:

A 9.8% gross margin on a business that is 63.9% self-mining implies power is a very large and largely unhedged share of cost, consistent with every other miner in this cluster — CORZ discloses power at 82.9% of self-mining revenue, and HIVE's gross mining margin fell from 49% to 24% on power and depreciation against a falling bitcoin price. But BTDR does not disclose it in the material reviewed and this memo does not assert it.

BTDR operates in Norway, Bhutan, Ethiopia and the United States among others (geographic disaggregation is tagged in the 20-F but was not extracted). Establishing the power contracts is a prerequisite to any position — it is the dominant variable cost of 64% of revenue and it is unknown.

---

5. Criteria scoring

BTDR is the cluster's clearest D1 case. The screen returned status: INDETERMINATE with the note "revenue MISSING" and dropped a $2.1bn company from the universe entirely. Its FY2025 revenue of $620,253,000 is fully tagged in the 20-F filed 2026-04-30 under ifrs-full:Revenue. The absence of a US-GAAP quarterly tag was read as the absence of revenue.

---

6. Downside Criteria (MEASURED — logged, blocks nothing)

Named cause: the current-debt wall, against a deteriorating gross margin and no visible interim data.

This is the most concrete near-term solvency structure in the cluster:

$753.8m of borrowings fall due within twelve months against $149.4m of cash, and cash fell 69% during FY2025. Meeting them requires monetising inventories ($252.0m of ASIC stock, in a market where BTDR is itself the seller), the crypto position ($218.6m, of which $135.6m is a receivable from an unidentified counterparty), or refinancing — including $275.0m owed to a related party, whose willingness to roll is a governance question rather than a market one.

Compounding it: - Ex the $444.9m derivative mark, FY2025 was a ≈$379.3m loss. - Gross margin fell from 21.1% to 9.8% while R&D rose to $153.9m (24.8% of revenue). - Four legacy revenue lines lost 95–98% of their revenue, so the declining businesses are declining fast. - 63.9% of revenue is bitcoin self-mining, exposed to price and to a rising global hashrate. - No interim XBRL exists. Every checkable peer in this cluster decelerated 7–24% sequentially in the most recent quarter. BTDR's position is unobservable and the base rate is adverse.

Realistic permanent-loss scenario: bitcoin sustained lower reduces self-mining revenue and simultaneously reduces third-party ASIC demand — the two largest revenue lines are correlated, not diversifying. Inventories become harder to monetise precisely when cash is needed. The $753.8m of current borrowings is refinanced dilutively (shares already +110% in two years) or renegotiated with the related-party lender.

Probability 35%. Magnitude −75%.

Going-concern: YES — a going-concern case exists and is argued explicitly, as criteria.md requires. Current borrowings are 5.0x unrestricted cash, total current liabilities are $1,520.7m against $1,386.6m of current assets — a negative working-capital position — and the largest current asset after prepayments is ASIC inventory in a market BTDR competes in.

Mitigants, stated fairly. Net assets are positive at $867.8m. Mining rigs ($620.7m) and PP&E ($441.8m) are real, saleable assets. The ASIC business is a genuine differentiator, not a narrative — $108.3m of third-party revenue is evidence of demand. And the related-party lender is more likely than a syndicate to extend. But the data is seven months old, and that is the honest limit of this assessment.

---

7. Screen defects found on this name

See CLUSTER_Analysis.md §9 for the full cross-name defect log. Specific to $620.253m (FY2025, 20-F, IFRS) — no interim XBRL exists:

BTDR's screen defect is different in kind from every other name in this cluster: there was no record to be wrong. The screen returned status: INDETERMINATE with a single note, "revenue MISSING", and no revenue, no share count, no market cap, no net cash, no EV and no multiples. A $2.1bn company was dropped from the universe entirely.

1. The revenue was not missing. BTDR's FY2025 revenue of $620,253,000 is fully and correctly tagged in the 20-F filed 2026-04-30, under ifrs-full:Revenue. So are three years of comparatives and a nine-line revenue disaggregation. The absence of a US-GAAP quarterly tag was read as the absence of revenue — calibration item D1 ("existence is not validity") in its purest form, and the most consequential instance in this cluster because it is total rather than partial.
2. The mechanical cause is filer status, and it is a whole class of names, not one name. BTDR is a 20-F filer reporting under IFRS with 6-K interims and no quarterly XBRL. A resolver that requires us-gaap:RevenueFromContractWithCustomerExcludingAssessedTax and/or four trailing quarterly durations returns nothing. CLSK failed the same way ("revenue STALE (666d)"), and HIVE — a 40-F/6-K filer until FY2026 — was given a stale quarter for the same underlying reason. Every foreign private issuer in this universe is exposed to this.
3. The screen still assigned it SIC 6199 (FINANCIALS). Bitdeer designs and manufactures semiconductors and mines bitcoin. Had the revenue resolved, is_ebit_valuable() == false would have set valuation = "NOT_APPLICABLE" anyway — so BTDR was excluded twice over, on two independent defects.
4. revenue_cagr_demonstrated could not be computed, and would have been misleading if it had been. Total revenue went $368.6m (FY2023) → $349.8m (FY2024, the halving year) → $620.3m (FY2025, the price run). Any CAGR across that window is a bitcoin price print.

Recommended fix: the revenue resolver must fall back to ifrs-full:Revenue and to annual durations, and status: INDETERMINATE must distinguish "no data found" from "data found and unusable." Reporting absence of data as absence of revenue is what removed a $2.1bn issuer from consideration.

---

8. Requirements not met, stated rather than omitted

• Transcript mention-frequency over time is NOT DELIVERED. mention-frequency.md requires it on every single-name analysis. No transcript source was reachable inside the time box for a multi-quarter series. The substitute actually gathered — the ratio of management's claimed AI capacity/ARR to reported AI revenue — is in §3 and measures the same disclosure-drift signal on audited numbers.
• No options chain was pulled. Per Liquidity Criteria, no options structure may be proposed without the actual chain, open interest and quoted size for specific strikes and expiries. None is proposed.
• Consensus Criteria is INDETERMINATE (Alpha Vantage quota shared and not attempted). Blocks nothing.