Mine Owners' New Business: Sitting on Land and Collecting Rent, Earning Billions Annually

比推Xuất bản vào 2026-03-16Cập nhật gần nhất vào 2026-03-16

Tóm tắt

The article "Mine Owners' New Business: Collecting Rent, Earning Billions Annually" explores the strategic pivot of Bitcoin mining companies towards AI infrastructure and high-performance computing (HPC) as Bitcoin approaches its supply limit. By 2026, with only 1 million Bitcoin left to mine and rising operational costs squeezing profitability, major mining firms are capitalizing on their existing assets—large-scale power capacity, data centers, and cooling systems—to serve the exploding demand for AI compute. Companies like IREN, Core Scientific, Cipher Digital, and Hut 8 have secured long-term contracts worth tens of billions of dollars with tech giants (Microsoft, Amazon, Google) and AI firms (Anthropic, CoreWeave) to provide GPU cloud services and HPC hosting. Financial reports highlight a stark contrast: while Bitcoin毛利率 have plummeted post-halving, AI-related services boast margins as high as 86%. Firms are rebranding, exiting mining, and leveraging their power infrastructure advantages—deploying AI data centers in months versus years for traditional builders. However, this转型 comes with risks: high debt from infrastructure upgrades, strict contract deadlines, regulatory hurdles, and operational challenges. The shift positions these companies as key "digital power stations" in the AI era, where control over electricity and grid access becomes a critical competitive edge. The period from 2026 to 2028 will be crucial for determining which players succeed in this ...

Author: KarenZ, Foresight News

Original Title: Mine Owners, the New Landlords of the AI Era


The world's mineral deposits will eventually become difficult to mine and one day be exhausted. The gold in California, the coal in the Ruhr, the tin in Cornwall—the turning point of vein depletion always forces an industry to restructure.

Bitcoin mining is no exception.

On March 9, 2026, the 20 millionth Bitcoin was officially "mined," with less than 1 million left to be mined. As the shadow of the 2028 block reward halving looms early, mining companies are falling into a survival anxiety.

However, unexpectedly, the explosion of the AI wave has opened up a "second growth curve" for mining companies.

Since 2024, AI companies and cloud providers have begun frantically competing for the power and data centers of mining companies. From the second half of 2025 to the present, this trend has not slowed down but has instead expanded and accelerated. The GPUs, electricity, and data centers required to train large models are creating a new thirst for computing power.

Timely, those Bitcoin mining farms once labeled "electricity guzzlers" have already laid the groundwork with large-scale power access, professional cooling systems, and high-power data centers.

These once standard features for mining have now become the most sought-after scarce chips in the AI era.

AI Companies Queue to Buy Electricity, Bitcoin Miners Sign Billion-Dollar Orders

From 2025 to early 2026, those old players who once "dug for gold" with ASIC miners quietly signed AI contracts worth tens of billions of dollars in total.

First, let's look at the names of the clients in the miners' AI mega-deals: Microsoft, Amazon, Anthropic, CoreWeave, Google, Fluidstack. These are the core players in the AI computing arms race, lining up to "give money" to the mining companies.

On November 3, 2025, Microsoft, to lock in GPU cloud services from IREN (a Bitcoin miner and AI data center), not only signed a five-year contract worth $9.7 billion but also paid a 20% advance payment upfront.

To support this "computing power engine," IREN in turn purchased $5.8 billion worth of NVIDIA GB300 chips from Dell. The speed of this move is enough to put most traditional data centers to shame.

In the second half of 2025, Google, through "cloud customer endorsement + ecosystem cooperation," guaranteed $1.73 billion for the computing power hosting project of miner Cipher Digital and Fluidstack (with Cipher's total contract revenue being $3.8 billion over 10 years) and holds at least 5% of Cipher Digital's equity.

At the same time, Amazon (AWS) also locked in a 15-year, approximately $5.5 billion high-performance computing (HPC) long-term contract with Cipher Digital on November 3, 2025.

If the giants are betting on the future, then AI newcomers like Anthropic and CoreWeave are the "hard-demand buyers" for mining companies.

U.S. Bitcoin miner and high-performance computing digital infrastructure company Core Scientific, leveraging its existing power capacity, delivered 350MW of high-density hosting for CoreWeave by March 2, 2026, aiming to complete 590MW by early 2027, to provide HPC hosting services for CoreWeave's NVIDIA GPUs. This 12-year contract has an estimated total revenue of $10.2 billion.

It is worth mentioning that CoreWeave itself is an AI cloud company "born from the mining circle." NVIDIA is its key strategic partner and important investor, having continuously invested over the years. CoreWeave was originally named Atlantic Crypto, founded in 2017, primarily mining Ethereum with GPUs. In 2019, Atlantic Crypto renamed itself CoreWeave and successfully transitioned to AI cloud services, focusing on GPU-accelerated cloud computing, particularly AI workloads, becoming a leader in the AI cloud service field.

Veteran Bitcoin miner Hut 8, in December 2025, joined forces with Anthropic and Fluidstack, committing to deliver AI data center infrastructure starting at 245MW, expandable up to 2295MW (total contract value $7 billion). TeraWulf, another U.S. operator focused on HPC hosting and Bitcoin mining, has a 25-year lease agreement with Fluidstack involving approximately $9.5 billion in contract revenue.

According to CoinShares statistics, as of the end of October 2025, mining companies had reached contract agreements with AI companies and tech giants totaling $65 billion. As time progressed to March 2026, the signing spree continues.

The Truth Hidden in Financial Reports: AI Expected Revenue is Overwhelming Mining

The latest financial reports from mining companies are revealing an irreversible turning point: the high gross margin and long-term expected revenue of the AI business are structurally overwhelming traditional mining.

Mining profit margins are being continuously squeezed. After Bitcoin's fourth halving in 2024, on one side, block rewards were halved, and on the other, the network's total computing power continued to climb. Riot Platforms' financial report shows that the average mining cost per coin has risen from $32,216 in 2024 to nearly $50,000 in 2025, a cost increase of 54%. And this is the number after excluding depreciation costs.

These pressures are ultimately reflected in the cliff-like drop in mining gross margins. Core Scientific's self-mining gross margin plummeted from 23% in 2024 to 5% in 2025; Bitdeer's self-mining gross margin in 2025 was only 10.9%.

In stark contrast is the "money printing" nature of the AI business.

AI hosting and cloud services generally have much more stable and higher gross margins than mining. IREN's AI cloud service gross margin (after deducting operating costs, excluding depreciation and amortization) is as high as 86%; Bit Digital's subsidiary WhiteFiber's cloud service gross margin (excluding depreciation and amortization) also reaches about 65%. Even Core Scientific, which uses a stricter accounting method by fully including depreciation and operating expenses in costs, still had a full-year AI hosting gross margin of 30% in 2025, climbing to 46% in the fourth quarter.

More decisive is this: the long-term expected revenue locked in by the AI contracts signed by mining companies is overwhelming mining by orders of magnitude.

In the fourth quarter of 2025, IREN's AI cloud service revenue was $17.3 million, a sequential growth rate of 137%. Although the current scale is still limited, thanks to the long-term GPU cloud service agreements signed with tech giants like Microsoft, the company expects to push its target Annual Recurring Revenue (ARR) to $3.4 billion by the end of 2026, which will significantly exceed mining revenue (mining revenue in fiscal year 2025 was $485 million). And all this consumes only about 10% of the company's locked grid capacity (>4.5GW), leaving huge room for subsequent growth.

Core Scientific similarly presents a pattern of "mining contraction, AI explosion." In 2025, its AI hosting revenue increased to $65.4 million, a surge of 168% year-on-year;同期 self-mining revenue shrank from $400 million to $230 million. The 12-year long-term contract signed with CoreWeave alone locks in total revenue of $10.2 billion, with an annualized stable revenue of about $850 million.

TeraWulf also reached a turning point. Of its total revenue of $168.5 million in 2025, mining still accounted for 90% (about $150 million), but it achieved HPC leasing revenue for the first time ($16.9 million). Although the current proportion is not high, it has signed long-term contracts for 522MW, with a total value of over $12.8 billion, basically locking in stable cash flow for the next few years.

However, the current general industry situation is: most mining companies still have a low proportion of AI revenue, but the logic of high gross margins + long-cycle locked orders has been established. Once the construction is completed and the proportion of AI revenue increases to 30%–50%, it will have a significant multiplier effect on overall profit quality. This is also the core logic behind the capital market's willingness to give high valuation premiums to transforming mining companies.

As Ben Gagnon, CEO of the listed Bitcoin mining company Bitfarms, said, after the company's Washington site was converted into a high-performance computing center, it accounted for less than 1% of the overall resources but is expected to create higher net operating income than previous Bitcoin mining.

The Most Radical Miners Are Abandoning Mining

The determination to transform is even engraved in the change of company names. Iris Energy changed its name to IREN, Marathon Digital upgraded to MARA Holdings, Applied Blockchain transformed into Applied Digital, Cipher Mining was renamed Cipher Digital, and Bitfarms plans to change its name to Keel Infrastructure and relocate to the United States.

This is never a simple name重塑, but a public and firm declaration—we are no longer the "mine owners" who focused solely on mining.

According to the degree of transformation radicalism, mining companies have diverged into the following three choices:

The radical embrace faction is represented by Cipher Digital, TeraWulf, Bit Digital, Bitfarms, and IREN. The characteristics of these companies are: transitioning fully to AI through brand renaming, exiting some mining sites, or even completely abandoning mining operations.

  • Bit Digital began gradually exiting the Bitcoin mining business from 2025, focusing all computing resources on its high-performance computing subsidiary WhiteFiber, while also focusing on Ethereum assets;

  • TeraWulf's AI transformation is progressing at a high speed, positioning HPC/AI as the "primary growth engine," while Bitcoin mining becomes "opportunistic operation."

  • Besides renaming, Cipher Digital is gradually exiting some existing mining site assets while accelerating its bet on AI hosting. Its mining assets totaling 4.4 EH/s in multiple mines in West Texas have been acquired by Canaan Technology to free up resources for a full transformation. On February 24, 2026, Cipher Digital officially announced this strategic transformation at its Q4 and full-year 2025 earnings conference.

  • Bitfarms announced in February 2026 its plan to rename the company to Keel Infrastructure and has clearly planned to gradually exit the Bitcoin mining business between 2026-2027 and convert sites into HPC/AI data center infrastructure.

  • Although IREN is still running Bitcoin mining operations, its business focus has clearly shifted towards AI.

In the hybrid model, representatives like TeraWulf, CleanSpark, and Bitdeer retain some or most mining operations as a cash flow buffer, while directing core resources to AI data centers.

  • Core Scientific currently derives most of its revenue from Bitcoin mining, but has significantly shifted from Bitcoin mining to high-density AI/HPC hosting. However, mining still dominates the income source currently.

  • CleanSpark uses the cash flow generated from mining to accelerate power and land expansion, controlling a portfolio of over 1.8GW—able to attack AI hosting and defend the mining position.

  • Bitdeer still regards Bitcoin mining as the cornerstone of its business but is evaluating the acquisition of land and the layout of high-performance computing infrastructure and hosting services by liquidating all Bitcoin and issuing notes.

Of course, some choose to stay true to their original intention and remain in the core battlefield of Bitcoin. They firmly believe in the long-term value of BTC as digital gold, refuse to chase the AI boom, and instead choose to weather the cycle through computing power expansion, position accumulation, and ecosystem cultivation. American Bitcoin continues to increase its Bitcoin mining efforts and continuously accumulates Bitcoin; BitFuFu adheres to the dual drive of cloud mining + self-mining. On the hardware side, Canaan Technology and Bitmain still firmly bet on ASIC manufacturing. This also provides stable computing power support for the Bitcoin network and may usher in excess returns when the coin price rebounds.

Regarding specific transformation paths, the "sit on land and collect rent, stable hosting" landlord model is the first choice for most mining companies. Most miners like Core Scientific, TeraWulf, Riot Platforms, Applied Digital, Hut 8, Cipher Digital, CleanSpark, etc., utilize existing power, sites, cooling systems, and operational experience to transform original mining farms into infrastructure that meets AI requirements, rent them out to AI companies, and collect stable rent. AI companies achieve "move-in ready" rapid deployment.

Different from this, mining companies like IREN, Bit Digital's subsidiary WhiteFiber, Bitdeer, and Bitfarms are not satisfied with this. They choose to go further, purchasing expensive chips themselves, building their own computing power pools, and directly conducting GPU cloud services (GPUaaS).

There is also a "speed run path"—directly acquiring mature data centers. The core is to use capital for time, seize track positions, and directly acquire companies with existing infrastructure and customer resources. The most representative is MARA, which acquired a 64% stake in Exaion, the high-performance computing subsidiary of the French electric power group EDF.

The Chips of Mining Companies

The confidence of mining companies in collective transformation stems from the hardcore chips they hold: existing power capacity, heavy asset infrastructure, and professional cooling systems. These chips ultimately translate into two core killer features:极致的时间优势 (ultimate time advantage), and flexible power elasticity that can advance or retreat.

First, let's talk about the time advantage.

Building a new AI data center requires going through the entire process from 0 to 1, often stuck in multiple rounds of approval, grid connection queues, and infrastructure construction, taking years.

Mining companies, holding approved power capacity, ready-made data center sites, and extensive experience in high-voltage electricity use and computing power operation and maintenance, directly save most of the infrastructure hassle, able to shorten the deployment cycle by up to 75% (according to Bernstein analysts' estimates).

CleanSpark CEO Matt Schultz told a story in an interview with CNBC: "The company recently won a 100MW data center contract in Cheyenne, Wyoming, competing against Microsoft."

A mining company with a market cap of less than a few billion dollars beat a tech giant worth trillions.凭什么? Matt Schultz explained, "Because we can build and power it within 6 months, while traditional AI data centers take 3 to 6 years."

Secondly, there is the rarely mentioned "power dispatch flexibility."

AI large model training has extremely high requirements. Data center contracts usually specify an absolute online rate of 99.99999%, with no room for compromise.

Mining companies adopting a "mining + AI" dual-line layout can shut down miners when the grid load is critical, feeding back into the grid. This flexibility to advance and retreat is precisely what pure AI data centers cannot do.

The Cost of Transformation: The Unseen Shoals

The AI transformation of mining companies seems smooth sailing with broad prospects, but this road is full of unseen shoals.

Infrastructure upgrades require real money investment. Original substations and fiber connections must be completely transformed to adapt to the high-density electricity demand of GPU clusters. Where does this money come from? Besides selling assets, most can only rely on aggressive borrowing.

Take Applied Digital's subsidiary APLD 2 as an example. To secure Oracle's 15-year data center lease, it issued $2.15 billion in guaranteed notes to build a data center in Polaris Forge 2. According to S&P Global estimates, its debt/EBITDA (an indicator used to measure a company's debt repayment ability) is expected to be as high as 8 times by 2028. S&P Global毫不客气地 gave APLD 2 a "B+" entity credit rating.

Similar financing pressures are普遍存在 in the industry. Cipher Digital also financed related data centers through high-yield bond issuance, with total funds raised reaching $3.73 billion. Core Scientific obtained a maximum $1 billion 364-day loan facility (initially $500 million fully drawn) from Morgan Stanley in March 2026 for data center development and energy procurement.

Risks are not limited to leverage. Contracts cannot be late, construction cannot go wrong; there is not much room for error in every link.

S&P Global pointed out that in the cooperation agreement between Applied Digital and Oracle, if the delivery is delayed by more than 150 days, Oracle has the right to terminate the lease. According to statistics from real estate services firm Jones Lang LaSalle, in 2025, 57% of data center projects were delayed by more than three months.

Finally, there is the multiple squeeze of operations and macro environment.

Externally, grid access and environmental impact approvals in places like Texas and New York are becoming increasingly stringent; internally, talent gaps are often underestimated. AI cluster scheduling, GPU operation and maintenance, and SLA service management are completely different capability systems. Coupled with the uncertainty of Bitcoin price fluctuations—mining cash flow一旦收缩, will directly drag down AI project investment and disrupt the entire transformation rhythm.

The High-Stakes Gamble Has No Retreat

In the AI era, what is truly scarce may not be chips, but power入口.

And Bitcoin mining companies—an industry that has long lingered on the边缘 of the tech industry—are step by step闯入 the core seats of this competition. They are no longer just miners, but are reshaping into "digital power stations" of the AI era. In this computing power war, GPUs can be manufactured, data centers can be built, but power capacity and grid access are difficult to replicate—and mining companies happen to sit on this entrance.

2026 to 2028 will be the key兑现窗口 for this transformation. Can billion-dollar contracts truly transform into gross profit in financial reports? Will the high-leverage debt structure expose risks due to construction delays? These issues are not only operational indicators but also matters of survival.

Five years from now, this track may give birth to new infrastructure giants, but more participants will likely only become footnotes in this migration.

And this industrial migration revolving around electricity and computing power, once started, is difficult to stop.


Twitter:https://twitter.com/BitpushNewsCN

比推 TG 交流群:https://t.me/BitPushCommunity

比推 TG 订阅: https://t.me/bitpush

原文链接:https://www.bitpush.news/articles/7620090

Câu hỏi Liên quan

QWhat is the main reason Bitcoin mining companies are shifting their focus to AI-related businesses?

AThe main reason is the approaching Bitcoin supply limit, with only 1 million left to mine by March 2026 and the upcoming 2028 halving, which is squeezing mining profitability. Meanwhile, the AI boom has created massive demand for GPU computing power, electricity, and data centers, areas where mining companies already have established infrastructure.

QWhich major AI and tech companies have signed significant contracts with Bitcoin mining firms, and what was the largest contract mentioned?

AMajor companies include Microsoft, Amazon, Anthropic, CoreWeave, Google, and Fluidstack. The largest contract mentioned was between Core Scientific and CoreWeave, a 12-year deal worth $10.2 billion for high-performance computing (HPC)托管 services.

QHow does the profitability of AI services compare to Bitcoin mining for these companies, according to their financial reports?

AAI services show significantly higher and more profitable margins. For example, IREN's AI cloud service毛利率 was 86% (excluding depreciation), while Core Scientific's AI hosting毛利率 reached 46% in Q4 2025. In contrast, mining毛利率 for companies like Core Scientific dropped to just 5% in 2025 due to rising costs and reduced block rewards.

QWhat are the two key advantages that give Bitcoin mining companies an edge in transitioning to AI infrastructure?

AThe two key advantages are: 1) **Time Advantage**: They possess pre-approved power capacity and ready-built data center sites, allowing them to deploy AI infrastructure up to 75% faster than building from scratch. 2) **Power调度 Elasticity**: Their hybrid mining-AI setup allows them to shut down mining operations to supply power to the grid during high demand, a flexibility pure-play AI data centers lack.

QWhat are some of the major risks and challenges these companies face in their transition to AI?

AKey risks include: high debt from financing infrastructure upgrades (e.g., Applied Digital's subsidiary issued $2.15 billion in notes), stringent contract delivery deadlines with penalties for delays, increasingly strict grid access and environmental permits, a shortage of skilled AI operations talent, and the volatility of Bitcoin prices which could impact the cash flow needed to fund the AI transition.

Nội dung Liên quan

Nhà Giao Dịch Bitcoin Cho Biết Các Đỉnh Và Đáy Chu Kỳ Khớp Chính Xác Số Ngày

Nhà giao dịch Ryan (tài khoản X @DodysDD) đã chia sẻ một lý thuyết gây chú ý, cho rằng chu kỳ giá Bitcoin lặp lại với độ chính xác đáng kinh ngạc về số ngày. Theo đó, các đợt tăng giá (từ đáy chu kỳ đến đỉnh) trong các giai đoạn 2014–2017, 2018–2021 và 2022–2025 đều kéo dài đúng 1.064 ngày. Trong khi đó, các đợt giảm giá (từ đỉnh đến đáy) trong các pha 2017–2018 và 2021–2022 đều kéo dài đúng 364 ngày. Mô hình này hấp dẫn giới giao dịch vì gợi ý một cấu trúc thời gian có thể dự đoán. Tuy nhiên, lập luận này tiềm ẩn rủi ro về việc "chọn lọc dữ liệu" (cherry-picking), vì độ chính xác phụ thuộc vào việc lựa chọn các mốc đỉnh và đáy cụ thể, bỏ qua các điểm đánh dấu chu kỳ khác có thể phá vỡ sự đối xứng. Không có bằng chứng cho thấy Bitcoin vận hành bởi một bộ đếm ngày chính xác, vì thị trường chịu ảnh hưởng bởi nhiều yếu tố phức tạp như sự kiện giảm một nửa phần thưởng (halving), điều kiện vĩ mô và tâm lý nhà đầu tư. Dù vậy, lý thuyết này vẫn thu hút sự chú ý vì các câu chuyện chu kỳ luôn có sức ảnh hưởng trong thị trường tiền mã hóa, cung cấp một khuôn khổ đơn giản để định hình kỳ vọng trong bối cảnh nhiều bất ổn hiện tại. Điều quan trọng là cần tiếp cận những tuyên bố về ngày chính xác với thái độ hoài nghi, xem chúng như một góc nhìn tham khảo về mặt tâm lý thị trường hơn là một dự báo giá đáng tin cậy.

bitcoinist2 giờ trước

Nhà Giao Dịch Bitcoin Cho Biết Các Đỉnh Và Đáy Chu Kỳ Khớp Chính Xác Số Ngày

bitcoinist2 giờ trước

94 tỷ, khoản đầu tư lớn nhất của người máy năm nay đã xuất hiện

Ngành robot hình người vừa chứng kiến khoản đầu tư lớn nhất trong năm khi Neura, công ty robot hình người có trụ sở tại Munich, Đức, hoàn thành vòng gọi vốn Series C với 1.4 tỷ USD (khoảng 94.9 tỷ NDT). Điều đáng chú ý là sự tham gia của các nhà đầu tư chiến lược từ ngành công nghiệp như Schaeffler và Bosch - những tập đoàn linh kiện công nghiệp lâu đời của Đức. Sự tham gia này cho thấy sự chuyển dịch trong logic của lĩnh vực này: từ những màn trình diễn công nghệ sang triển khai thực tế trong nhà máy, và từ câu chuyện vốn đầu tư sang hệ thống thương mại thực sự. Sau vòng gọi vốn, định giá của Neura đạt khoảng 7 tỷ USD, đưa công ty vào nhóm dẫn đầu toàn cầu. Khác với các công ty như Figure AI tập trung vào robot hình người đa năng với câu chuyện về AI thể hiện (embodied AI) được hậu thuẫn bởi OpenAI hay Microsoft, Neura theo đuổi con đường ứng dụng theo ngành dọc trong công nghiệp. Công ty đã có khách hàng thực tế là BMW và sản phẩm của họ đã được kiểm chứng trên dây chuyền sản xuất. Có hai lý do chính cho làn sóng đầu tư mạnh mẽ này. Thứ nhất là sự tiến bộ vượt bậc của các mô hình lớn (AI), phá vỡ giới hạn về khả năng nhận thức và ra quyết định của robot. Thứ hai là áp lực từ phía nhu cầu: tình trạng thiếu hụt lao động và chi phí nhân công ngày càng tăng trên toàn cầu, đặc biệt ở các nền công nghiệp như Nhật Bản, Đức, buộc các nhà sản xuất phải tìm giải pháp thay thế. Mặt trận chính của robot hình người giờ đây không còn là các buổi ra mắt sản phẩm mà là mặt bằng nhà máy. Hai lĩnh vực được kỳ vọng sẽ triển khai quy mô sớm nhất là sản xuất công nghiệp (vì môi trường có cấu trúc, nhiệm vụ lặp lại) và các môi trường làm việc nguy hiểm (hóa chất, hạt nhân). Tuy nhiên, thách thức lớn nhất cho việc triển khai hàng loạt không còn là công nghệ lõi mà là các vấn đề kỹ thuật và thương mại như chi phí thích ứng với từng dây chuyền cụ thể và xây dựng hệ thống bảo trì, dịch vụ địa phương đáng tin cậy. Việc các gã khổng lồ công nghiệp lâu đời bắt đầu "bỏ phiếu" bằng tiền thật cho thấy ngành công nghiệp này đã chuyển từ câu hỏi "Liệu có làm được không?" sang "Làm thế nào để làm tốt hơn, nhanh hơn và ổn định hơn". Đây mới là tín hiệu quan trọng nhất từ khoản đầu tư kỷ lục này.

marsbit7 giờ trước

94 tỷ, khoản đầu tư lớn nhất của người máy năm nay đã xuất hiện

marsbit7 giờ trước

Thị Trường Trước Niêm Yết của Anthropic Sụt Giảm Sau Lệnh Hoa Kỳ Buộc Ngừng Hoạt Động Mô Hình

Công ty trí tuệ nhân tạo Anthropic thông báo đã nhận chỉ thị từ chính phủ Mỹ vào ngày 12/6, yêu cầu ngừng cung cấp quyền truy cập hai mô hình Claude Fable 5 và Claude Mythos 5 cho người nước ngoài, kể cả nhân viên nước ngoài trong công ty. Để tuân thủ, Anthropic đã vô hiệu hóa cả hai mô hình trên toàn cầu. Lệnh này được mô tả là một biện pháp kiểm soát xuất khẩu khẩn cấp liên quan đến an ninh quốc gia. Các mô hình khác như Claude Opus 4.8 không bị ảnh hưởng. Anthropic phản đối quyết định này, cho biết chính phủ chỉ cung cấp bằng chứng bằng lời nói về một lỗ hổng "jailbreak" hẹp và không phổ biến, liên quan đến việc yêu cầu mô hình xem xét một mã nguồn cụ thể. Công ty lập luận lỗ hổng này nhỏ, đã biết trước và có thể được tìm thấy bởi các mô hình công khai khác, không cần thiết phải đóng cửa toàn bộ mô hình thương mại. Họ cảnh báo tiêu chuẩn này nếu áp dụng rộng rãi có thể đình chỉ mọi triển khai mô hình mới của các nhà cung cấp AI tiên phong. Thị trường tiền điện tử đang theo dõi sự việc do các hợp đồng phái sinh liên kết pre-IPO của Anthropic, cho phép giao dịch phản ánh tâm lý về lĩnh vực AI. Ngay sau chỉ thị, hợp đồng vĩnh viễn Anthropic trên Hyperliquid đã giảm 3.7%. Sự kiện này cho thấy quy định AI đang trở thành yếu tố có thể giao dịch được, và cơ sở hạ tầng AI đang hòa vào bản đồ thị trường đầu cơ cùng với crypto. Tuy nhiên, rủi ro là các thị trường này có thể biến động mạnh dựa trên thông tin không đầy đủ, trong khi báo cáo kỹ thuật của chính phủ chưa được công khai.

bitcoinist13 giờ trước

Thị Trường Trước Niêm Yết của Anthropic Sụt Giảm Sau Lệnh Hoa Kỳ Buộc Ngừng Hoạt Động Mô Hình

bitcoinist13 giờ trước

Ví Khai Thác Chuyển Đổi Token Bị Đánh Cắp Thành 18,510 ETH Và 1,548 BNB

Ví tiền liên quan đến một vụ khai thác lỗ hổng bảo mật đã chuyển đổi tài sản bị đánh cắp thành 18,510 ETH (khoảng 30,83 triệu USD) và 1.548 BNB (khoảng 924.000 USD), theo cảnh báo theo dõi trên chuỗi được WuBlockchain chia sẻ, trích dẫn dữ liệu từ Lookonchain. Việc chuyển đổi này đáng chú ý vì sau khi khai thác, các ví thường chuyển từ các token kém thanh khoản hoặc dễ bị truy vết sang các tài sản có tính thanh khoản cao hơn như ETH và BNB trước khi cố gắng rút tiền. Kẻ tấn công được cho là liên quan đến token "H" bị xâm phạm và vẫn đang nắm giữ số token trị giá khoảng 14 triệu USD. Các giao dịch hoán đổi lớn sau khai thác quan trọng vì chúng có thể gây áp lực bán lên tài sản, hé lộ bước di chuyển tiếp theo của kẻ tấn công và cung cấp manh mối cho các nhà điều tra. Trong khi theo dõi trên chuỗi (on-chain) giúp hiển thị các chuyển động này, việc xác định danh tính thực tế của người kiểm soát ví vẫn là thách thức. Các ví có thể nhanh chóng chia nhỏ hoặc chuyển tài sản xuyên chuỗi, làm phức tạp công tác truy vết. Báo cáo nhấn mạnh tầm quan trọng của việc theo dõi dữ liệu để hiểu cách quỹ bị đánh cắp được hợp nhất, đồng thời lưu ý rằng thông tin từ các nguồn như Lookonchain và WuBlockchain cung cấp cái nhìn nhanh chóng, nhưng không thay thế cho báo cáo điều tra chính thức. Việc chuyển đổi sang các tài sản có tính thanh khoản cao như ETH và BNB thường là giai đoạn phổ biến, làm phức tạp thêm các lựa chọn thu hồi tài sản sau đó.

bitcoinist15 giờ trước

Ví Khai Thác Chuyển Đổi Token Bị Đánh Cắp Thành 18,510 ETH Và 1,548 BNB

bitcoinist15 giờ trước

Giao dịch

Giao ngay
Hợp đồng Tương lai
活动图片