Author | Lin Wanwan

Editor | Sleepy.txt

No one initially anticipated that the real bottleneck of AI is not capital, not large models, but electricity.

With large-scale training running at full capacity long-term and AI inference operating 24/7 non-stop, a problem arose: there isn't enough electricity available, forcing chips to sit idle. Over the past decade, the U.S. has relatively lagged in electric grid infrastructure, with new large loads connected to the grid taking 2–4 years on average, making "readily available electricity" a scarce commodity across the industry.

Generative AI has brought to the forefront a raw and cruel fact: it's not the model that's lacking, it's electricity.

As a result, the story took a turn, and crypto mining enterprises, those who earliest treated electricity as a "means of production," began moving from the periphery to the center stage of capital.

Iris Energy (IREN) is a prime example of this trajectory. This year, IREN's stock price soared nearly 600% at one point within the year, with a 52-week range from 5.12 USD to 75.73 USD. Seizing the opportunity while Bitcoin's rise remained attractive, it decisively allocated power to transform its self-built AI data center.

When giants like Microsoft came in with long-term orders totaling 9.7 billion USD, the market visually understood for the first time the real path from mining to AI: starting with electricity and land, then GPUs and clients.

However, not all mining enterprises, like IREN, chose to go all-in on AI. Amid this power-driven hashrate migration, there is a prudent force worth our attention — Bitdeer.

Bitdeer Technologies Group (NASDAQ: BTDR), founded by the crypto legend Wu Jihan and headquartered in Singapore, strategically held nearly 3GW of global electricity resources, steering clear of the shallow trap of relying on external "power supply" from the beginning. When the wave of AI surged, Bitdeer did not opt for the aggressive "All-in" approach like IREN but retained profitable Bitcoin mining as a "foundation," while steadily upgrading some mining farms to AI data centers.

This "ability to attack and defend" strategy has made it the best example to observe how global players are thinking and positioning themselves in this hashrate race.

For this reason, we interviewed Wang Wenguang, Vice President of Global Data Center Business at Bitdeer, hoping to shed light on the current global AI power shortage and their perspective on mining enterprises transitioning to AI data centers. Do they see it as a capital hype or a genuine need for AI? We engaged in an in-depth conversation on this series of questions.

Why is the US Power Shortage So Severe?

Insightful Inquiry: Let's start with a fundamental question, do you think electricity prices will continue to rise in the future?

BitDeer: I believe they will, as this is a crucial future supply-demand dynamic.

Insightful Inquiry: Regarding the power shortage in the US, there is a belief in the market that it is very difficult to obtain an "electricity license" in the US?

BitDeer: It's not that this so-called "electricity license" is hard to obtain, but rather that the physical expansion speed of the power grid is lagging behind. In the years following the relocation of heavy industry out of the US, the US power grid construction did not undergo systematic expansion. After mining companies moved into the US in 2021, much of the "already connected, already signed PPA" power was locked up by these mining companies. With the influence of ChatGPT, pure AI players entered the scene and only then did they realize the immediately available massive power in the mining farms.

This explains why major companies are willing to collaborate with mining companies. Instead of waiting 2–4 years to build 500MW from scratch, it's better to refurbish existing facilities in 12 months.

Insightful Inquiry: When did the industry truly realize that "inference is also power-intensive"?

BitDeer: Probably after the widespread adoption of GPT-4. As companies embedded models in customer service, office work, search, risk control, etc., the long-term and scenario-based inference demand did not see the expected decrease in power consumption.

This led to two types of changes.

One is engineering upgrades: from stronger air cooling to liquid cooling/hybrid heat dissipation, cabinet power, power distribution paths, fire protection, and monitoring have all been elevated to the level of AI data center.

The other is resource strategy: electricity has truly become the primary bottleneck. People no longer just discuss "buying cards" but rather advance to getting power and grid connection, long-term PPA, grid connection schedules, cross-regional capacity backup, and, when necessary, upstream power procurement like mining companies (self-generated power/direct procurement).

In fact, we have seen a similar trend in the mining industry long ago. Chips can scale infinitely (silicon comes from sand), but power expansion is limited. We have done natural gas self-generated power in Canada to ensure power supply for mining farms, following this logic. Today's AI is almost identical.

Insightful Inquiry: How does the power consumption of AI data centers compare to traditional internet data centers?

Bit Doe: It's not about incremental change, but about orders of magnitude change. In the past, a 20-30 MW data center in the traditional internet world was considered large, but today, AI data centers have power demands of 500MW or even 1GW. AI has transformed data centers from a "rack business" to a "power engineering" business, requiring a reevaluation of everything: power lines, substations, cooling, fire safety, redundancy, PUE... The experience gained from traditional internet data centers is still valuable, but no longer sufficient.

Observations: Why has "electricity" become the scarcest upstream resource?

Bit Doe: Chips can scale up because they come from silicon and capacity management; however, scaling up electricity is challenging because it comes from power generation and grid upgrades. In the past, the mining industry has attempted to "source energy from the upstream," including self-power generation projects in Canada. The path AI is taking is similar—whoever secures electricity first gains an advantage in deployment time.

AI New Battlefield: From "Grabbing GPUs" to "Grabbing the Grid"

Observations: When mining companies transition to AI data centers, what specific changes are needed?

Previously, everyone said, "Bitcoin mining power can be used to run AI," but mining chips (ASIC) and GPUs required for AI are not compatible. So why can mining companies now "provide AI computing power"?

Bit Doe: The global mining industry used to be divided into two parts: Bitcoin relied on mining chips ASICs, which were efficient but had a single purpose; Ethereum relied on NVIDIA GPUs, which were versatile but have exited the mining stage after transitioning to PoS.

Therefore, when the market talks about "mining farms transitioning to AI" today, it almost always refers to Bitcoin mining farms undergoing transformation. The key point is that these farms are no longer "calculating hashes" but are upgrading themselves into AI data centers.

This is an infrastructure replacement process: removing ASIC racks and installing GPU servers; upgrading the power systems from "just enough" to a professional-grade power distribution with N+1/2N redundancy; upgrading traditional air cooling to a system capable of handling high-density GPU cooling; and standardizing and making facilities auditable in terms of room sealing, dust prevention, and fire protection.

By completing these four steps, a cryptocurrency mining farm transitions from a "mining workshop" to an "AI data center."

Why can mining companies self-build faster than AI giants? Electricity.

AI is a business of "electricity and heat," and the construction period of AI data centers is 3-4 years, with time being the most significant barrier. Mining companies happen to hold these "hard assets" in their hands, allowing them to have a head start in the transformation race.

Insight: In the past few days, Microsoft and Amazon have successively signed multi-year AI contracts with cryptocurrency mining companies. Iris Energy (IREN) signed with Microsoft for a total value of 9.7 billion, spanning 5 years; another company, Cipher, signed with Amazon Web Services for 5.5 billion over 15 years. This is seen as the first batch of cases where mining farms collaborate with tech giants. What is your take on this?

BitDeer: Iris Energy is a forward-thinking Australian company that has been mining in the United States for a long time.

Iris Energy's decision to pivot towards AI is like a signal flare. At a time when the price of Bitcoin is high and peers are still expanding mining operations, it diverted some of its power to invest in a self-built AI data center. Subsequently, AI enterprises proactively came knocking.

The real trigger came from the Hyperscalers' substantial investment—such as Microsoft's commitment of around 9.7 billion U.S. dollars—the market has, for the first time, clearly seen that between mining companies and tech giants, it is not just about "technology integration," but about "exchange of power and time."

The heat of AI amplifies the infrastructure demand, opening up collaboration opportunities.

Insight: Why are leading mining companies more easily chosen by U.S. AI giants at this stage?

BitDeer: Because of "available electricity + engineering delivery speed." The site selection and grid connection from the previous cycle of mining companies have now become the upfront capital for AI data centers. Time is the biggest discount factor, directly determining who can go online within the window, get customers, and generate rolling cash flow.

Insight: Is the land selection requirement for AI data centers challenging?

BitDeer: Overall, not quite. In the United States and most countries, what is truly scarce is electricity, not land.

The reason is simple—places with abundant electricity are often energy-rich areas (natural gas fields, coal mining areas, near hydroelectric power stations), sparsely populated, and with low land costs.

For example, Bitdeer's large data centers in Norway and Bhutan are located far from population centers, where power resources are concentrated and land costs are low. Similarly in the U.S., these parks are not in city centers but in more remote locations, where land is easy to find and inexpensive. The "first principle" of site selection is power and grid connection, with land usually following the electricity and not being the main bottleneck.

Insight: AI is now being referred to as the upstream business of "steel, electricity, and land," even resembling another form of real estate. What are your thoughts?

BitDeer: With the emergence of large models, the energy consumption of AI far exceeds most people's expectations.

Initially, everyone thought that "training consumes power, while inference is lightweight," but in reality, after inference became more widespread, it also maintained a long-term high power consumption. As ChatGPT and DeepSeek become more prevalent, with increasing terminal access, the baseline power consumption of inference continues to rise.

From an engineering perspective, AI is fundamentally a resource-intensive industry:

· Chip Side: During training, the accelerator card is basically running at 100% load, naturally resulting in high power consumption;

· Data Center Side: The heat density is much higher than that of traditional servers, the PUE is significantly elevated, and heat dissipation itself also consumes a large amount of power;

· Scale Side: The electricity demand of AI data centers has skyrocketed from the traditional 20–30MW of internet data centers to the 500MW or even 1GW level, which was almost unimaginable in the era of traditional internet data centers.

Therefore, likening it to "real estate" is only half correct. Indeed, it requires land, buildings, and a long cycle (construction periods often lasting 3–4 years), but its life or death is determined by electricity and heat—whether it can timely access high-capacity grids, implement N+1/2N redundancy, and efficient heat dissipation. In this respect, its strong dependence is very similar to that of steel, electricity, and land.

What are the characteristics of AI data centers?

Insight: What are the unique characteristics of data center construction in the United States?

BitDeer: Due to power constraints and historical reasons, in the United States, hyperscalers often need to get directly involved and collaborate with mining companies to secure available electricity.

Insight: Is it possible for foreign companies to build AI data centers in the United States?

BitDeer: In simple terms, AI data centers are a highly regional business. When it comes to facilities that easily reach hundreds of megawatts and thousands of cards, it is still predominantly domestic giants in the U.S. who are leading. We are only discussing AI data centers here, without involving traditional internet data centers.

Insight: Will AI Data Centers evolve into geopolitical tools? Will this affect your decision-making?

BitDeer: I agree with this assessment.

At the core of AI lies data, which is inherently subject to sovereignty and security constraints. To prevent data leaks and security risks, regions around the world are tightening their related policies: even though the U.S. allows foreign investment to build data centers, as AI gains more and more data, most countries will likely move towards "on-site deployment, local compliance, and data localization."

In simple terms, American AI stays in America, Middle Eastern AI stays in the Middle East, European AI stays in Europe, regionalization will be a long-term trend.

Industry Landscape and Potential

Insight: Besides IREN and BitDeer, who in the mining industry has more potential to transform into AI data centers?

BitDeer: To see who has a shot, first look at whether they have access to large-scale power, then see if they can quickly transform their mining farms into GPU data centers. The most likely candidates to receive AI orders are those with grid connection + land + power transformation capabilities, able to achieve N+1/2N redundancy, liquid cooling/high density.

For the other category, with pure hosting/light assets, who do not control electricity supply and industrial parks, transitioning to AI data centers will be more passive.

In the U.S., companies like Riot, CleanSpark, Core Scientific, TeraWulf, and Cipher, which have resources in hand and reliable expansion plans, are more likely to attract attention from major players.

So the conclusion is straightforward: electricity is the ticket, transformation capability is the speed; only when both are in place can you lead the pack.

Overall, it depends on who controls "high-quality, sustainable large-load available power." Companies with more self-owned grid-connected resources are more promising; those primarily relying on hosting, lacking their own energy and industrial park resources, are not advantageous in this round of structural transformation.

What is BitDeer Thinking?

Insight: What is BitDeer's strategy and path for "mining to AI" transformation?

BitDeer: Mr. Wu Jihan's strategy has always been to cover the entire industry chain, BitDeer holds about 3GW of power and park resources, which is our biggest underlying advantage.

When we first entered the AI field, we did not anticipate that "electricity" would become a core bottleneck, so we initially took a self-built, self-operated approach: we partnered with NVIDIA to become an NVIDIA PCSP, deployed a small-scale H100 cluster in Singapore, launched our own AI Cloud, and opened up our training services to external clients. This project has been successful.

Subsequently, we also set up a second data center in Malaysia. As the Hyperscalers entered this arena and began collaborating with mining companies, we simultaneously worked on upgrading high-load data centers to AI data centers: we have announced the complete transformation of a 180MW site in Norway into an AI DC and the conversion of a 13MW site in Washington state, USA.

Ultimately, the essence of AI is very similar to Crypto mining—both are businesses of "electricity + infrastructure"; we have end-to-end capabilities in power supply, data centers, and computing operations, making the transition to AI relatively smooth.

Insight: What is the key difference between Bitdeer and other mining companies like IREN?

Bitdeer: Three points. Firstly, we will not transition fully into an AI company; based on our calculations, the current stage of Crypto Mining is still more profitable than AI data centers, and the mining industry has stable cash flow and better returns.

Our second advantage is our international engineering organizational capabilities. The Bitdeer team's engineering organization and execution capabilities are unparalleled worldwide. For the same AI data center, which typically takes two years in the US, we can usually do it in a year and a half. This is achieved through parallel progression and supply chain coordination, aligning key aspects such as construction, mechanical and electrical work, power distribution, and cooling simultaneously to compress the regular 24-month period to about 18 months, thus forming available capacity more quickly.

Thirdly, the company's strategy remains prudent: the AI industry is very young, even younger than Crypto, so we are not going "all-in" and instead aiming for a more sustainable pace of development.

Insight: Where is Bitdeer's current power infrastructure mainly located?

Bitdeer: Bitdeer is currently strategically positioned with around 3 GW of power and related infrastructure globally, covering five countries: the US, Canada, Norway, Ethiopia, and Bhutan, to support the construction and operation of mining facilities and AI data centers.

Costs and Financing

Insight: I saw a Goldman Sachs report mentioning that an AI data center might cost around 12 billion USD. Is it really that cash-burning?

BitDeer: Indeed, it's huge, on the order of "tens of times." Here's a "layman's number" for you to easily compare:

Bitcoin Mining Farm (USA): Building 1 MW costs around 350,000–400,000 USD, but building 1 MW of an AI data center costs around 11 million USD. This is because the investment in an AI data center is a complex of "heavy machinery + high standards," along with grid connection queues, environmental and energy assessments, regional compliance, usually spanning 18–36 months.

You will find that the essence of an AI data center is not "buying a few more cards," but rather connecting a piece of land into an "electricity city" capable of consuming 500MW–1GW, ensuring proper power connection, heat dissipation, redundancy, compliance, all of which are very costly.

Insight: Where does the money come from? Does it require financing?

BitDeer: To be honest, all require financing.

Let me share, here are a few common financing methods in the industry:

1. Project Financing/Infrastructure Loans: Using the park + equipment as collateral, relying on long-term leases or hashpower offtake (customer commitment to buy your hashpower for many years) to reassure banks.

2. Equipment Leasing/Sale-Leaseback: Lease the GPUs and some electromechanical components, spread over a long period, avoiding the need to cash out a large sum at once.

3. Long-term PPA: Lock in electricity prices and available capacity first, making bondholders willing to offer low-interest rates.

4. Tie-up with major companies: Large customers/companies offer minimum consumption, prepayment, guarantees, or even joint ventures (JVs), allowing you to access cheaper funds.

In partnerships like IREN, CoreWeave, Google/Microsoft, these terms can be observed.

Insight: Does BitDeer also need financing? Will it soon announce collaborations with major companies?

BitDeer: It's not something that can be publicly discussed at the moment.

Conclusion

Not long after the interview, BitDeer revealed its next move in the capital market.

On November 13, BitDeer announced that it would raise $400 million through the issuance of convertible preferred notes. The initial purchasers would be granted an option to subscribe for an additional $60 million of notes within 13 days, with the fundraising cap potentially reaching $460 million. The new funds will be used for data center expansion, ASIC miner development, AI and HPC cloud business expansion, and general corporate purposes.

In a time when power has become the scarcest upstream resource in the AI industry, the allocation of this $460 million, and how many megawatts of new load it will connect to, will significantly determine BitDeer's position in the next round of computing power competition.

For BitDeer, this money is more like encapsulating the judgment from interviews into the balance sheet: on one end, connecting to the cash flow bedrock of the mining sector, and on the other, linking to the long and thick snow-covered business line of AI data centers. It may not immediately reflect in the revenue and profit of the next quarter's financial report, but it will slowly reshape the power structure of the computing power business over the next few years—who is qualified to sit at the negotiation table, and who can only wait in line on the grid connection list for electricity.

Looking forward from the results, this round of AI infrastructure story is not complicated: power has truly become the upstream, time has become the new currency, and the industrial parks and grid connection quotas in the hands of mining companies have turned into "old assets" that others can't buy even with money.

As the buzz about models and applications gradually fades, the market will likely flip the books again: whose narrative sounds louder will no longer be important. The companies that can securely connect every megawatt of electricity in a power shortage world and keep running smoothly will be eligible to stay at the next stage of the table.

Welcome to join the official BlockBeats community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Official Twitter Account: https://twitter.com/BlockBeatsAsia