On November 10, Bitdeer (NASDAQ: BTDR) released its third-quarter 2026 results.

The income statement looked ugly. The contract book looked much better. For now, the market is focused only on the former, while the latter still depends on the company delivering each engineering milestone.

Record Revenue, but Crypto Mining Barely Profitable

Third-quarter revenue hit a record $320.7 million, up about 89% year over year. Adjusted EBITDA rose 173% to $69.6 million.

But the company posted a net loss of $125.5 million.

The problem becomes clear when the numbers are broken down: crypto mining. Self-mining revenue was about $218 million, while depreciation alone came to roughly $90 million, leaving just $6.2 million in gross profit. As hashprice continued to fall, newer mining rigs brought heavier depreciation. Crypto mining has gone from a profit center to a source of cash.

It can mine cash, but not profit.

The company is also pulling back on crypto mining investment. It cut its 2026 capital expenditure budget for mining infrastructure to less than $200 million. Jihan Wu said on the earnings call that mining capex would slow further next year, partly because AI demand is tightening TSMC wafer supply. Bitdeer is strategically abandoning Zcash while continuing to invest in Scrypt. Scrypt is the crypto mining algorithm used by Litecoin and Dogecoin; it is separate from Bitcoin's SHA-256, and the mining rigs are not interchangeable. Wu's rationale is that Scrypt currently generates more revenue per megawatt than Bitcoin mining, while Bitdeer's in-house Scrypt mining rigs have an edge over competing products. The same electricity can earn more by mining a different cryptocurrency.

Bitdeer is still mining, but its money is moving elsewhere.

The Real Story Is in the Contract Book

On the AI side, third-quarter AI Cloud revenue was $20.4 million, with a gross margin of about 15%. The figure remains small, but the contracts are already substantial.

In Tydal, Norway, Bitdeer signed a 16-year lease worth about $4.7 billion with neocloud company Volta, backed by an irrevocable letter of credit worth about $1.3 billion for the first six years. AI Cloud has signed contracts worth roughly $2.9 billion, with more than 70% of A201's capacity already committed and expected to generate over $1.7 billion in revenue. Together, the contracts represent about $7.6 billion in AI revenue.

The earnings call produced two disclosures more consequential than anything in the earnings release.

First, Volta has completed its GPU financing. The CFO said this removed the last major condition before the letter of credit could be activated. The completion deadline for Tydal's first phase is December 31, and he believes the project will meet it.

Second, customer prepayments. Wu said prepayments from customers would soon be “well above 50%,” because capital markets are closing to neocloud providers and no one is willing to begin projects with zero upfront payment anymore.

That point matters. It means customers are paying for more than half of the GPUs used in Bitdeer's AI Cloud business upfront. Traditional crypto miners borrow money to buy their own machines. Bitdeer's model increasingly resembles an AI infrastructure provider funded by customer prepayments.

The market values Bitdeer as an unprofitable crypto miner, even as its cash-flow structure moves in another direction. The disconnect between those two labels has yet to close.

Four Issues That Need Scrutiny

The good news must be weighed against the ugly parts.

First, cash. Cash and restricted cash fell from $496.3 million at the end of the second quarter to $213.7 million at the end of the third, a decline of more than half in three months. Current liabilities stood at $1.13 billion against $930 million in current assets, leaving a short-term gap of roughly $200 million. The company had just 47 Bitcoin left on its balance sheet.

Next is financing. Asked whether Bitdeer would issue more shares in the fourth quarter or the first quarter of next year, the CFO said the company would not “tie its hands completely.” Its at-the-market offering raised only about $58 million in the third quarter, far below the $457 million raised in the second. Project financing and customer prepayments come first, with equity financing behind them. But the door remains open.

Third is inconsistent disclosure. In his opening remarks, the CFO put Tydal's remaining capital expenditure at about $350 million. During the Q&A, he said it was about $500 million and added that “we have said it was $500 million.” The earnings release said Fox Creek would come online in the fourth quarter of 2027, while on the call he first said between midyear and year-end, then between midyear and the third quarter. Two conflicting figures from the same CFO on the same call require an explanation from investor relations.

Finally, guidance. Based on the $2.9 billion in contracts, an analyst estimated that AI Cloud could generate roughly $580 million in annualized revenue at full capacity. The analyst asked when Bitdeer would reach that level, how the business would ramp next year, and what margins it expected at scale. The CFO deferred every question to the fourth-quarter results.

No Capacity in the US? The Answer Is Malaysia

Of everything said on the call, this comment from Jihan Wu was the most memorable:

“In the United States today, it is already very difficult to lease data center capacity that can be delivered next year.”

Wu said Bitdeer's US sales team was receiving a large volume of inquiries, all focused on capacity in Malaysia. Both US companies and Asian technology firms are looking there. He added that approvals for new data center construction in the US would slow after the midterm elections, making sites that are already powered even more valuable.

That is good news for Bitdeer, provided it has power available in Malaysia.

Bitdeer's official AI capacity list names Cyberjaya and Johor in Malaysia. The 67 MW A901 project added in October is expected to be ready in the second quarter of 2027, but the earnings release identified it only as “Malaysia – A901,” without naming the city.

We previously reported that Bitdeer had secured a site in Kuantan, Pahang, Malaysia, for an AI data center and was running the EPC tender as the project owner. Yet Kuantan was not mentioned once in the third-quarter power capacity table, management's remarks, or the entire Q&A session.

Of course, omission from official disclosures does not mean the project does not exist. Awarding an EPC contract does not mean construction has started, and approved power does not mean the site has been energized. But if US capacity cannot be leased for delivery next year and customers are all asking about Malaysia, investors should press IR on two questions: Where is A901, and when will Kuantan appear in the power capacity table?

Three Triggers to Watch

On the trading day before the results, Bitdeer closed at $6.61, up 9.9% for the day. After the release, premarket moves were mixed but remained within 3% in either direction. The shares have fallen about 40% this year and more than 70% from their one-year high of $23.30.

At this level, much of the bad news is already priced in. The net loss included a $54.6 million fair-value adjustment on digital-asset borrowings, a non-cash item. Bitdeer held only 47 Bitcoin on its balance sheet, but it also had more than $340 million in digital-asset receivables from related parties. When those receivables turn into cash is worth watching.

In the near term, I see three triggers for the share price.

The first is December 31. If Tydal's first phase is delivered on schedule and the letter of credit is activated, the $4.7 billion contract will begin turning into cash flow. It is the nearest and most concrete catalyst before year-end.

The second is another share issuance. With the stock at $6 or $7, another round of at-the-market sales would be highly dilutive. If Bitdeer can carry Tydal and A201 through with project financing and customer prepayments, it would remove the biggest overhang on the stock.

The third is the fourth-quarter report. Only when Bitdeer discloses AI Cloud's ramp-up schedule and target margins will the market be able to revalue it as an AI infrastructure company.

Conversely, if hashprice keeps falling, Tydal's delivery is delayed, or the company remains vague about its new capacity in Malaysia, crypto-mining valuations will continue to weigh on the stock. The performance of peers including IREN and CoreWeave will also drive volatility.

My view is that over the next one or two quarters, Bitdeer's share price will become less correlated with Bitcoin and more closely tied to engineering milestones.

The old business is bleeding, the new one is signing contracts, and financing must bridge the gap between them.

For crypto miners moving into AI, the test is whether they can deliver power and data center capacity on schedule, and make that progress visible to the market. If they cannot deliver, even the largest contract value means nothing.