Cryptocurrency mining service providers are transitioning and expanding into AI computing power, forming an emerging trend
TechFlow Selected TechFlow Selected
Cryptocurrency mining service providers are transitioning and expanding into AI computing power, forming an emerging trend
Cryptocurrency asset service providers are gradually becoming a significant force in AI computing power supply.
Navigating Web3 tides with focused insightsAuthors: Song Jiaji, Sun Shuang
Bitdeer Technologies Group (BTDR.O), a Singapore-based cryptocurrency mining services provider, announced its inclusion as a Preferred Cloud Service Provider in NVIDIA's Partner Network and plans to launch AI GPU cloud computing services targeting the Asian market. The service will be built on NVIDIA DGX Superpod and DGX H100 infrastructure.
According to public data, leveraging NVLink technology, an NVIDIA DGX Superpod can connect up to 256 H100 GPUs, achieving a total computational capacity of 1,013 petaFLOPS—calculated based on the FP16 precision performance of a single H100 (NVLink version) at 3,958 teraFLOPS.
AIGC Boom Drives Soaring Demand for Computing Power
According to OpenAI estimates, from 2012 to 2018, the computing power required to train AI models doubled approximately every three to four months, increasing by a total of 300,000 times during that period—compared to only a sevenfold increase under Moore’s Law over the same timeframe. Leading-edge AI training models have seen annual compute requirements grow tenfold, reflecting an exponential trend. On the supply side, NVIDIA dominates the market; according to Barron’s, demand for NVIDIA chips remains extremely strong, with current H100 orders already booked through 2024.
On November 9, OpenAI CEO Sam Altman noted on social media that ChatGPT experienced outages due to user traffic far exceeding expectations. Such incidents highlight the growing bottleneck in AI computing capacity.
Cryptocurrency Mining Firms Transitioning to AI Compute Services
Bitdeer is not alone. According to Sina Finance, stablecoin issuer Tether recently acquired 10,000 H100 GPUs. Other crypto miners such as Hive, Crusoe, and Hut 8 are also repurposing idle GPUs, renting them out by the hour to AI startups urgently needing chip access. Cryptocurrency mining firms are increasingly becoming a significant force in the AI computing supply chain.
Advantages of Crypto Mining Firms Entering the AI Compute Market
First, low-cost electricity procurement—a core competitive advantage for crypto mining firms—is equally critical for AI computing. For example, Bitdeer reported an average electricity cost of approximately $38 per megawatt in the first half of 2023, significantly lower than commercial electricity prices in the U.S. during the same period.
Second, these firms possess extensive operational experience managing high-power equipment, including heat dissipation and troubleshooting hardware failures—skills highly transferable to AI computing operations.
Additionally, some crypto mining firms already offer cloud computing services. Bitdeer, for instance, has cloud hashing as one of its core business lines.
Both Crypto and AI Cloud Computing Are Moving Toward Centralization and Cloud-Based Models
1. Centralization: We observe that multiple listed companies are entering the AIDC (AI Data Center) sector, similar to Bitcoin mining farms, centralizing GPU management, securing optimal power sources, monitoring GPU failures, performing maintenance, and handling workload scheduling and cooling.
2. Cloudification: Amid surging demand for AI computing power, GPUs used for large model training face severe supply shortages, driving the rise of the "cloud computing" model. This allows AI model developers to rent computing power on-demand from cloud platforms without purchasing expensive hardware like NVIDIA A100 GPUs, enabling startups and non-top-tier players to enter the AIGC space.
We believe that in the upcoming era of supercomputing internet, beyond simply having or lacking computing resources, efficiently utilizing existing capacity—particularly optimizing computing resource allocation and scheduling—will become equally crucial.
Investment Recommendations
We recommend focusing on cryptocurrency mining service providers and AI computing resource scheduling and optimization service providers:
1. Crypto Mining Services: Bitdeer, Canaan Inc., Zhidu Co. (an investor in Bitdeer);
2. Optical Communications: Innolight Technology, New H3C, Tianjin TF Communication, TFC, Tengjing Technology, DEKRA, Lumentum, Huagong Tech, YYD Tech, CIG, MPO,;
3. Communications Equipment: Scisco, Feilinkeyisi, ZTE Corporation, Tsinghua Unigroup, Ruijie Networks, Foxconn Industrial Internet, Cambricon, Zhenyou Technology;
4. Computing Resource Scheduling & Optimization: Stargate, Sinocrown, Orient Materials, Hengrun, Borui Data, QingCloud, Zhongbei Telecom, Sugon, China Mobile, China Telecom, China Unicom;
5. Data Visualization: Hengwei Tech, Haohan Depth, Sinovac;
6. BOSS Systems: AsiaInfo Technologies, Tianyuan Dikco, Oriental Gowan.
Risk Warnings: AI technological development may fall short of expectations; AI computing demand may be weaker than anticipated; intensified market competition.
This article is excerpted from a report titled “Great Momentum: Crypto Mining Firms March Into AI,” published by Guosheng Securities Research Institute on November 11, 2023. For full details, please refer to the original report.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
