15 Mining Stocks Slump, Bitdeer Plunges Nearly 32% in a Week: AI Compute Pivot Collides With the 'Nvidia Storm' and BIS Iron Curtain

BTC
BIS export controlsBitdeerAI compute pivotBitcoin MinersGPU computeStock DeclineNvidia
16 hours agoSource: blockweeks.com
15 Mining Stocks Slump, Bitdeer Plunges Nearly 32% in a Week: AI Compute Pivot Collides With the 'Nvidia Storm' and BIS Iron Curtain

Author: Heart of Computing Power

Over the past week (as of November 18), the stocks of Bitcoin mining companies fell almost across the board:

Of the 18 listed mining companies, 15 declined, with the median stock price falling 10% over the week.

The biggest decliner was Bitdeer (BTDR), which fell 31.9% over the week.

IREN fell 23.4%. Only a few companies rose, with the biggest gainer being Cango (CANG), which rose 32.4% over the week.

Many people's first reaction is that Bitcoin fell, so mining stocks followed, which is perfectly normal.

That's true, but it only tells half the story.

Because for the past two years, mining companies have almost all been doing the same thing: shifting from mining Bitcoin to renting out AI computing power.

And this week, the most critical link in that transformation ran into trouble.

I. The Coin Price Fell, and Mining Is No Longer Profitable

The reason everyone can see is that Bitcoin fell.

As of November 18, the price of Bitcoin was about $64,377, down 12.96% over the past 7 days.

On October 6, 2025, Bitcoin hit an all-time high of $126,198, and it has now fallen 49%, roughly half.

Looking at November alone: it was still at $80,405 at the start of the month, fell below $68,000 on November 13, and dropped another 6% in a single day on November 14.

Rate hike expectations added another layer of pressure. The latest inflation data came in higher than expected, and the market believes the Fed will definitely raise rates in December, with traders putting the probability at 100%.

Fed Governor Waller even hinted that there could be a larger hike in December.

Rate hikes mean borrowing is more expensive, and mining companies happen to be especially reliant on borrowing, needing to constantly buy equipment and build data centers, so their stock prices are particularly sensitive to rate hikes.

The most direct consequence of the coin price falling is that mining is becoming less and less profitable.

There is an industry metric called hashprice, which means how much money 1 PH/s of computing power can mine per day.

That number is now only about $33.3, and many mining machines can barely break even, or even lose money, at this level of revenue.

JPMorgan did the math:

A batch of high-cost miners has already shut down and exited, and the industry's cost to mine one Bitcoin has dropped from $90,000 to about $78,000.

But even after the drop, $78,000 is still higher than the current coin price of more than $64,000. Put simply, many mining companies now lose money on every coin they mine.

With mining this difficult, mining companies have to find another way out. The way out they found is AI.

II. Mining Companies Are All Going into AI

Why AI, of all things?

Because mining companies happen to have exactly what AI companies lack most: power, land, and data centers.

Mining Bitcoin and running AI both require large amounts of electricity and data centers that can dissipate heat.

These mining companies built all of that long ago; they just need to replace mining machines with Nvidia GPUs (that is, AI chips) to rent computing power to AI companies and collect rent monthly.

This is far more stable than mining, which depends on the coin price to make a living.

The fastest to make the shift is IREN.

In the most recent quarter, its Bitcoin mining revenue fell 58% from the previous quarter to just $28 million, while AI cloud business revenue rose 182% to $200 million, accounting for about 88% of total quarterly revenue.

In other words, IREN basically no longer relies on mining to make a living.

Companies like Bitdeer, Core Scientific, and CleanSpark are doing the same thing: converting their existing data centers and power into a business of renting out GPU computing power.

This path itself is smooth: mining companies have power and data centers, and only lack GPUs.

The problem lies precisely with GPUs: usable AI chips are basically supplied by Nvidia, and Nvidia is an American company, so who it sells to and where it sells are both subject to the U.S. government.

This week, there were developments on both the Nvidia side and the U.S. government side.

III. The Path to Buying GPUs Got Stuck

According to reports, on November 12, Nvidia announced that founder Jensen Huang had resigned from the positions of president, CEO, and director after being caught up in an extramarital affair scandal (the other party was alleged to be a suspected Chinese spy), with CFO Colette Kress temporarily taking over as CEO.

On November 12, Nvidia's stock fell more than 13% at one point during intraday trading.

Measured from its all-time high in early October, Nvidia's stock has fallen more than a third, and is down nearly 17% year to date.

However, Nvidia's business itself is very good.

After the close on November 17, Nvidia reported earnings: revenue of $115.8 billion, up 103% year over year, and earnings per share of $2.68, both better than market expectations.

But there is a detail in the earnings report worth noting:

When forecasting future performance, Nvidia assumed "zero data center computing revenue from China," and it has already stopped shipping licensed versions of Hopper chips to China.

Combined with the Jensen Huang matter, the market believes that the U.S. is watching "AI computing power flowing to China" more closely than before.

What really hit mining companies directly was the second matter.

On November 16, BIS, the export control department under the U.S. Department of Commerce, issued an interim rule: from now on, advanced U.S. chips exported to Brazil, Indonesia, Malaysia, Singapore, Thailand, or re-exported from these countries must first obtain a U.S. government license.

The reason given was that chips in these countries have "a higher risk of being diverted than normal."

The rule also specifically closed a loophole. Previously, some people did not buy chips directly but instead remotely rented U.S. AI computing power from data centers in third countries; now this kind of "remote access" is also regulated.

Image

BIS's enforcement guidance also gave an example: "a Chinese parent company purchasing Nvidia H200 servers through a Singapore subsidiary" was directly listed as a high-risk situation.

Simply put, the path of "setting up a company in Southeast Asia, buying American chips, and doing AI computing business" is now more tightly blocked.

But some mining companies with Chinese backgrounds are taking exactly this path in their AI transformation, and Bitdeer is the most typical example.

IV. Why Bitdeer Fell the Most

The sector median fell 10%, but Bitdeer fell 31.9%, more than three times the median. The reason is that several things hit it at the same time.

First, it happens to be in the places that were named.

Bitdeer is headquartered in Singapore, its founder is Wu Jihan, co-founder of Bitmain, and its data centers are distributed in the United States, Norway, Bhutan, and Malaysia.

Both Singapore and Malaysia are on BIS's new list this time.

Bitmain's Chinese background is also exactly the direction the United States is most focused on right now.

In the future, whether it is buying Nvidia GPUs or renting computing power to customers, it may be affected by the new rules.

Second, the report card it just delivered was not very good.

On November 10, Bitdeer released its third-quarter report: revenue was $320.7 million, up 89% year over year, a new high, and this point is indeed good.

But its loss also hit a new high: net loss expanded from $30.1 million in the same period last year to $125.5 million, more than four times last year's figure.

Gross margin fell from 23.4% in the same period last year to 3.4%; for every 100 yuan of goods sold, only 3.4 yuan remained after deducting direct costs.

There was also a $54.6 million book loss on a digital asset loan in the quarter.

After the earnings report was released, analysts cut their target prices one after another: Barclays from $17 to $14, Northland from $23 to $14, and Needham from $20 to $18.

Third, the money it has on hand is not ample.

Transforming into AI is very cash-burning; it requires buying large numbers of GPUs and also renovating computer rooms. To this end, Bitdeer has already issued more than $1.5 billion in convertible bonds and has also raised funds through multiple stock offerings.

With a lot of debt and diluted equity, once buying GPUs is restricted and borrowing becomes more expensive, the room it has to withstand pressure is very small.

Several things collided at once: on November 12, Bitdeer fell 12.8% in a single day; on November 13 it fell another 8.65%, and the stock price at one point approached its 52-week low of $5.03. Compared with its 52-week high of $20.72, it has already fallen more than 60%.

Not long ago, what it was telling was still a growth story of "$2.9 billion in AI cloud orders and a 3.1GW power plan."

Comparing it with the other two companies will make it clearer.

IREN also fell 23.4% this week, because Microsoft adjusted its infrastructure arrangements, and IREN's originally planned $4 billion annualized revenue target was postponed to early 2027.

But IREN is in North America, and its major customer is Microsoft. Its problem is that "progress has slowed," not that "the road has been blocked."

Cango, by contrast, moved in the opposite direction. It began to contract earlier than others: selling Bitcoin to repay debt and pushing down mining costs.

So when the sector plunged this time, funds instead hid over there, and it rose 32.4% in a week.

Bitdeer's situation is this: its location and background happen to run into the new rules, it does not have the support of major customer orders like IREN, nor has it reduced debt in advance like Cango.

▌▌▌▌▌▌▌▌▌▌

Whether mining company stocks can stop falling next depends mainly on three things:

First, whether Bitcoin can hold $64,000. If it falls below $62,000, it may return to the November 15 low;

Second, whether Nvidia's better-than-expected earnings, together with an upwardly revised fiscal 2028 growth forecast, can restore the market's confidence in AI computing power;

Third, when the new BIS rules are actually implemented, whether licenses are approved loosely or strictly.

There is one more point worth noting.

The AI computing power business is still growing rapidly, but how much of this growth can be distributed to companies with Chinese backgrounds is still hard to say.

In the past, when the market looked at mining companies transforming into AI, it mainly looked at how much power they had on hand.

After this week, everyone may also need to ask one more question: to whom can your computing power be legally sold?