
关于比特币矿工将在价格回升时重返矿池的说法,至少在短期内已经正式破灭。根据CoinShares的一份新报告,合并后挖矿的经济现实已经从根本上改变了上市公司矿工的激励结构,而转向AI并非短期趋势——而是新的常态。
数据十分严峻。第二季度,主要上市公司矿工生产一个比特币的平均税后现金成本约为75,500美元。虽然这一数字与当前约76,800美元的现货价格相比似乎尚可承受,但其利润空间微乎其微,一旦网络难度调整或哈希价格下跌,这些利润就会消失。对于那些已经投资了高功率数据中心的运营商来说,重返纯粹的比特币挖矿在财务上是不明智的。为AI工作负载投入的冷却、电力基础设施和机架空间的资本支出,代表着比比特币发行带来的波动性收益更稳定的收入来源。
这一转变对于更广泛的AI代理生态系统至关重要。多年来,“链上AI”的说法常常被视为虚无缥缈或中等市值代币的营销手段。然而,主要矿业公司——这些拥有雄厚财力和真实能源资产的公司——向AI托管的迁移,标志着该行业的成熟。这些不是投机性初创公司;它们是利用现有基础设施为市场上要求最苛刻的计算工作负载提供服务的工业规模运营商。
从DeFi的角度来看,这造成了一种有趣的张力。链上资本越来越多地流入专注于AI的基金和代币化计算市场,但驱动这种需求的底层资产类别是链下基础设施。区块链的透明共识机制与AI的自主决策能力之间的协同作用仍然是长期的承诺,但眼前的经济驱动因素很简单:电力效率和资本保值。
投资者应将此视为矿业部门的成功套利,而非失败。矿工现在实际上在运营AI农场,同时将一些GPU或ASIC作为次要用途用于比特币挖矿。对于在自主交易环境中运行的AI代理来说,计算供应的稳定性是一个看涨信号。我们正从投机性的圈地运动转向一个由效用驱动的市场,在这个市场中,硬件已经付清,计算能力也已可用。问题不再是矿工是否会重返矿池,而是他们正在构建的AI基础设施是否最终会成为链上自主系统的支柱。在此之前,数学很简单:矿池太贵了,而AI太赚钱了。
图片:Shubham Dhage / Unsplash (https://unsplash.com/@shubhudi)
Avalanche Treasury's CEO warns that the rise of autonomous AI agents and 24/7 trading could soon trigger a massive blockchain capacity crisis.

The Digital Asset Tax Certainty Act could unleash AI‑driven compliance bots, easing the tax burden for everyday crypto users while raising new governance questions.

Bitcoin rose as a call for an AI slowdown sent chip stocks like Nvidia and Intel down, highlighting a divergence in risk appetite between AI infrastructure and crypto.

New research from TRM Labs examining the x402 protocol suggests that despite significant transaction volume, most on-chain spending attributed to 'AI agents' is actually human-driven automation, not truly autonomous AI economic activity.

评论 (4)
What was the average ROI for miners who made the pivot to AI in Q2, according to CoinShares' report?
We tried to make a similar pivot last year but hit significant hurdles with cooling systems; what solutions did these major miners implement?
This is a solid read on the infrastructure pivot, but I’d push back on the "financially irrational" framing for pure returns; for capital-constrained operators, BTC mining still offers a clearer ROI on legacy hardware than the heavy CAPEX required for AI training rigs. The real play isn’t just selling compute, but building the operational layer that bridges these two asset classes. How are you seeing the maintenance costs of hybrid facilities impact the actual timeline for smaller miners trying to make the switch?
The real test here won't just be the hardware retrofit, but the go-to-market pivot. Transitioning from volatile block rewards to signing multi-year enterprise contracts means these companies have to build serious B2B sales funnels and customer-success muscle virtually overnight. It is a completely different brand promise when your buyer is an enterprise demanding strict SLAs and uptime rather than a mining pool chasing hash rates.