Bitdeer: the price assumes today's bitcoin mining profits last, so sell at $7.58

Bitdeer has grown fast in 2026. It mined 1,355 bitcoin in September, three times a year earlier, its self-mining capacity reached 82.3 exahash per second, and it signed a 16-year, $4.7 billion data center lease in Norway with an AI customer. It still lost $92.3 million in the second quarter, burned $158.5 million of operating cash and ended June with $1.84 billion of borrowings. At $7.58 the shares are priced as if September's mining income is permanent, even though the next bitcoin halving, in April 2028, will cut mining rewards in half.

At $7.58 on October 29, 2026, the shares trade at 2.8 times my Base value of $2.74, so the Base is 63.9% below the price. The Bear value of $1.02 is 86.5% below the price and the Bull value of $12.60 is 66.2% above it. I assign no probabilities to the cases. The required margin of safety of 55% gives a buy range of $0.64 to $1.83 and a conservative limit of $0.64. The verdict is Sell. The third quarter report on November 10 is the next test.

Valuation route: bitcoin miner converting into AI and data center hosting, valued as a sum of its parts: mining cash flow after rig replacement, the contracted Norway lease, the AI cloud business and the Malaysian contracts, less net debt. Economic classification: speculative and capital intensive, with the value of the mining fleet driven by bitcoin and not by management. Cross-check: book value per share. Values are present fair-value estimates, not 12-month price targets.

What Bitdeer is

Bitdeer has four activities. It mines bitcoin for its own account (82.3 exahash per second in September on about 243,000 machines, with an average efficiency of 15.8 joules per terahash). It hosts or co-mines for others (28.7 exahash per second of third-party hash rate). It builds its own mining machines under the SEALMINER brand, which it currently uses for itself and does not sell. And it is turning part of its power into AI infrastructure: a GPU cloud with 4,328 NVIDIA GPUs at about 95% utilization, new data halls in Malaysia, and a long lease in Norway for an AI customer.

Online electrical capacity was 1,752 megawatts on July 31, and the company reports about 3 gigawatts including its pipeline. Average power cost was $44 per megawatt-hour in the second quarter. Two share classes (Class A and Class V) were outstanding, 271.8 million shares in total at June 30.

The latest quarter

Revenue grew, but the gross margin was negative because depreciation rose almost fivefold as new machines came online. My own division of self-mining revenue by hash rate gives about $26.6 per PH/s per day for the quarter, well below the $39.33 average of September, which is why the second quarter looks so weak and the third may look much better. The mining business is highly sensitive to bitcoin, and the results move with it.

What the cash flow is made of

• Adjusted EBITDA is not cash. $31.1 million of adjusted EBITDA sits next to an operating cash outflow of $158.5 million and capital spending of $266.0 million. The business is funded by borrowing and by selling shares.
• Mining machines wear out quickly. Depreciation of $107.7 million a quarter, or about $431 million a year, is a good guide to what must be spent each year to stand still. I use about $8 per PH/s per day of sustaining spending in the model.
• The AI business is early. $14.0 million of revenue in the quarter is about 6% of the total. Its capital needs and financing terms are not disclosed in the release.
• Related parties are involved. Borrowings from a related party were $633 million at June 30, $491 million of it current, and the company holds a receivable of $162.2 million in digital assets from a related party. I count half of that receivable in the Base and none in the Bear.

Debt, shares and financing

The company sold $325 million of 5.00% convertible notes due March 2032 in February 2026, with a conversion price of about $9.93, and used part of the proceeds to buy back $135 million of 5.25% notes due 2029. At $7.58 the 2032 notes are below their conversion price, so I treat them as debt and not as shares. I do not know the amount and terms of the other notes. Management also says it filed a new $1 billion share sale programme; at today's price that is about 132 million shares, or 48% of the current count. Shares sold above my Base value add to value per share, but the cash is being used to build capacity whose returns are uncertain.

I use 271.8 million shares, the reported figure at June 30. I have not seen the share count after June, so any sales since then are not in the model.

Mining economics

Mining pays by the amount of computing power, so the key number is revenue per PH/s per day, called hashprice. It depends on the bitcoin price and the network's total hash rate, and it halves with each halving unless the bitcoin price doubles. Bitcoin closed September at $83,687 and hashprice averaged $39.33 per PH/s per day, the highest monthly average since January, while network hashrate was about 975 exahash per second.

At the September hashprice the mining business earns about $547 million of EBITDA a year by my estimate, and the whole company trades at 6.2 times that figure. That looks cheap, but it assumes September's hashprice lasts, ignores machine replacement of roughly $250 million a year and ignores the halving. In the long run hashprice tends to settle near the cost of the marginal miner, where an operator earns about its cost of capital. Bitdeer's power cost of $44 per megawatt-hour is only a little below the industry average of about $48.

The Norway lease and the AI business

The Tydal lease is the one contracted asset that is large relative to the share price. The company describes a 16-year lease of 121 megawatts of IT load, worth about $4.7 billion in the base term, with a 3% annual escalator and electricity passed through. Average revenue is about $2.4 million per IT megawatt a year. The customer is described only as a leading AI lab. Phases are due on December 31, 2026 and March 31, 2027.

• Termination right: the tenant may end the lease at the end of year 10.
• Remaining spending: about $500 million, to be funded mainly by project-level debt that has not yet closed.
• Credit support: described as anticipated, in the form of a letter of credit, and not yet in place.
• Execution: management said there are "probably a hundred things to do" before the first phase.

The AI cloud business had annual recurring revenue of about $86 million in September on 4,328 GPUs, and a contracted backlog of about $2.9 billion, up from $0.9 billion in August. Most of the increase is offtake commitments for more than 70% of the 21.7-megawatt A201 hall in Malaysia, worth more than $1.7 billion over five years, with full deployment targeted for April 2027. Two further Malaysian halls (67 and 60 megawatts) have data services agreements with delivery planned for the second quarter of 2027 and the first quarter of 2028; I could not find their contract values.

Primary valuation: sum of the parts

• Mining Base: hashprice of $34 in 2027, $29 in 2028 and $26 from 2029, 85 exahash per second, efficiency improving from 15.8 to 12.5 joules per terahash by 2031, power cost of $44 per megawatt-hour. At $26 the all-in cost is about the same, so the mining business earns its cost of capital and adds nothing beyond 2028.
• What $26 means: if network hashrate reaches 1.2 zettahash, a hashprice of $26 after the halving needs bitcoin at about $136,000. My Base therefore already assumes bitcoin well above today's price. This conversion is my own arithmetic and depends on the network estimate.
• Operator floor: I assume the company stops replacing machines in any year when mining cash flow would be negative, so mining value is never negative.
• Tydal terminal share: the lease is valued over 16 years; years 11 to 16 contribute 15% of its value.
• Not valued: the SEALMINER manufacturing business, the 47 megawatts retained at Tydal, the land bought in Texas and any new AI sites.

The table shows value per share from 2029 onward with all other Base inputs unchanged. The price of $7.58 is reached only with a hashprice of about $36 and a lease that is partly renewed. The value is mainly a bet on bitcoin and the network, which management does not control.

Secondary valuation: book value

Shareholders' equity was $1,102.8 million at June 30, or $4.06 a share. At 1.0 times book this gives $4.06. I chose 1.0 times because at the hashprice that I assume in the Base the company earns about its cost of capital on its mining machines, so the assets are worth about what they cost. This is a check on what the assets are worth, not on the income they will earn: much of the book value is mining machines bought in 2026 whose resale value depends on hashprice. The price is 1.87 times book.

Growth diagnostic

Growth is real but largely capital-funded: self-mining hash rate rose about 342% in a year and co-mining 260% in the quarter, bitcoin output tripled, and AI cloud revenue went from $1.3 million to $14.0 million. Each of these needed machines, power or GPUs that were paid for with borrowing and new shares, and so far growth has not produced net profit or positive operating cash flow. I classify the growth as structural for AI hosting, where there are signed contracts, and cyclical for mining. The share count at the end of June was about 10% above the second quarter's weighted average.

Valuation triangulation

• Primary (sum of the parts): $1.41, weight 50%.
• Secondary (1.0 times book value): $4.06, weight 50%.
• Divergence between the two: 97% of their midpoint, above the 30% limit, so I flag a valuation divergence.
• Weighted Base: $2.74. The unweighted range of $1.41 to $4.06 is the honest uncertainty, and $2.74 is a point inside it.

The two methods differ because the book value counts $2.1 billion of equipment at its cost, while the sum of the parts values the same equipment at what it will earn at my Base hashprice, which is about its replacement cost, and then subtracts debt. Net debt of $1.35 billion is the same in both. The weights are equal because I cannot show which is right. Both values are far below the price, so the choice does not change the verdict.

What is the market already pricing in?

A buyer at $7.58 is paying for September's mining income to persist beyond the halving, or for bitcoin to roughly double. That is a possible outcome, and the Bull case shows it, but it is a bet on a price the company does not control, financed with debt.

Bear, base and bull scenarios

• Bear: hashprice of $27, $20 and $16 from 2029 (bitcoin roughly unchanged), 80 exahash per second, power cost of $48 per megawatt-hour, so mining is switched off or worth nothing; the Tydal tenant ends the lease after year 10 with a 70% margin and the hall is re-let at half the rent; AI cloud at 1.5 times its recurring revenue; no value for the Malaysian contracts or the related-party receivable. The enterprise value of $970 million is below net debt, so the equity is worth about zero and I floor it at $0.00. Book value at 0.5 times gives $2.03.
• Base: as described above.
• Bull: hashprice of $44, $40 and $34 from 2029 (bitcoin near $178,000 at 1.2 zettahash), 95 exahash per second, efficiency of 11 joules per terahash, power cost of $42, a Tydal margin of 85% and the full 16 years plus a residual, AI cloud at 5 times recurring revenue, $640 million for Malaysian contracts and the whole receivable. Book value at 2.0 times gives $8.12.

The discount rate is the same in all three cases (12% for mining and 10% for the lease). The distribution is wide and skewed: the Bull case is more than four times the Base and the Bear is close to zero. Because debt is large relative to the assets, small changes in hashprice swing the equity violently, and that is the reason for the low Confidence grade.

How the buy price is set

Confidence is D, which has a default margin of safety of 55% and a reference range of 50% to 65%. I use the default of 55%. I add no further adjustment for debt, hashprice or method disagreement, because those are already in the Confidence score and in the scenarios, so the margin is applied once, to each method.

The entry range is $0.64 to $1.83 and the conservative limit is $0.64. The price is above both bounds by a very wide margin, so there is no case for buying. The range is a statement about the evidence I can support, not an expectation that the shares will fall to those prices.

Dividend and buybacks

Bitdeer pays no dividend and I found no buyback programme. It raised $457 million from share sales in the second quarter and has a new $1 billion programme in place. The company is a net issuer of shares, and I expect that to continue while capital spending exceeds cash flow.

Reasons to own BTDR

• Mining income is highly sensitive to bitcoin, and hashprice in September was $39.33, so one more rise would add a great deal to equity value.
• The Norway lease is a contract of about $4.7 billion with 3% annual escalators and pass-through power costs.
• AI cloud contracted backlog rose from $0.9 billion to about $2.9 billion in a month, with 95% GPU utilization.
• About 3 gigawatts of power capacity, 1,752 megawatts of it online, is hard to replicate and can be converted to AI hosting.
• Its own machines are among the most efficient: 15.8 joules per terahash on average and 9.45 for the newest model.

What could go wrong?

• Bitcoin falls or the network grows faster than bitcoin rises. The April 2028 halving cuts rewards in half.
• The Norway financing and letter of credit are not closed, and the first phase is due on December 31, 2026.
• $491 million of related-party borrowings is current. Refinancing or repayment needs cash or more share sales.
• Dilution. The June 30 share count was about 10% above the quarter's average, and a $1 billion programme is in place.
• The AI business is small ($14.0 million of quarterly revenue) and its GPU capital needs are not disclosed.
• Related-party exposure: borrowings of $633 million and a digital asset receivable of $162.2 million.
• Management said regulatory criteria for its Texas site are still unclear.

Management execution

• Delivery on mining capacity has been strong: self-mining hash rate more than quadrupled in a year and bitcoin output tripled.
• Management signed the Norway lease and the Malaysian offtake commitments within a few months and lifted contracted backlog from $0.9 billion to $2.9 billion in a month.
• Financing has relied on convertible notes, related-party borrowing and a share sale programme, and the operating cash outflow was $158.5 million in the quarter.
• The guidance I found covers only 2026 capital spending on mining infrastructure ($200 million to $280 million, excluding machines, GPUs and AI hosting), not earnings or cash flow.

Stock Analyza scorecard

The valuation score is 10 of 100, in the highly overvalued band of 0 to 24.

Business quality and valuation are shown separately and are not averaged. A fast-growing, loss-making business at a price far above its central value is a sell.

Confidence score

The grade measures how well the model is identified, not the chance of a loss. The long-term hashprice, the debt leverage and the unknown financing of the AI projects are the reasons for the low grade. The decision not to buy at $7.58 holds unless one assumes a hashprice near $36 after the halving; whether to sell an existing holding depends on how much of the Bull case a holder is willing to carry.

What I would watch from here

The three most important thesis breakers

1. Bitcoin and network growth leave hashprice near or below the all-in break-even of about $30 after the halving.
2. The Norway lease fails to close its financing or credit support, or is cut short.
3. Refinancing the related-party and convertible debt requires large share sales at low prices.

Stress test and independent check

I recomputed the values with a separate script and solved the reverse valuation for the long-term hashprice. That checks the arithmetic, not the forecasts. Many inputs are my estimates, as the table shows.

The weakest point of the model is the value of the AI cloud and Malaysian contracts, because their capital cost and financing are not disclosed; I use modest values that cannot be defended precisely. The strongest bull argument is that bitcoin will rise far enough to keep hashprice near $36 after the halving and that the AI contracts are worth more than my Base, which together would justify the price. The strongest bear argument is that mining is a cost-of-capital business at best, the equity is thin behind $1.8 billion of debt and share sales will continue.

Final verdict: sell at $7.58

Bitdeer has real assets, a real lease and a fast-growing AI business. At $7.58 the market pays for mining income at September levels to last through the halving, and it leaves little for the debt and the dilution that fund the growth.

Verdict: SELL at $7.58. Weighted Base $2.74, 63.9% below the price, cross-check $4.06, Confidence D (7/25). Required margin of safety 55%, buy range $0.64 to $1.83, conservative buy limit $0.64.

For an existing holder, the original cost of the shares is irrelevant. A holder who wants exposure to bitcoin and AI hosting should size the position so that a fall to a few dollars, or to nothing in the Bear case, is acceptable, and I would use strength to reduce it. A review is warranted on November 10, when the third quarter results and any update on the Norway financing are due. New money waits for a price at or below $1.83.

Sources

Primary sources are Bitdeer's second quarter 2026 results release, its September 2026 production and operations update and the February 2026 convertible notes pricing. Management commentary on the Norway lease, capital spending and financing is from an automated transcript of the second quarter call, which contains transcription errors, so I used it only for figures that are consistent with the release. Hashprice, bitcoin price and network hashrate for September are from Luxor's September 2026 review. The third quarter date is from the company's announcement.

The reference market price is $7.58, the October 29, 2026 regular session close, from MarketScreener. A broker screenshot on October 30 shows $7.13, which is a pre-market figure 5.9% lower and is not a regular session close. Balance sheet figures are as of June 30, 2026.