When a bankrupt Bitcoin miner turns down a $9 billion acquisition offer, the market collectively raises an eyebrow. But for the shareholders of Core Scientific, the decision was less about rejecting a payday and more about betting on a future where the same infrastructure that secures the Bitcoin network also powers the next generation of artificial intelligence. It’s a bet that carries both profound promise and perilous engineering complexity.
Core Scientific emerged from Chapter 11 bankruptcy in early 2024 after the crypto winter. The company has been refocusing on high-performance computing (HPC) and AI data center hosting, leveraging its existing mining facilities. The AMD partnership announced in early 2025 is a key part of this pivot. The rejected $9B bid from CoreWeave would have taken the company private. The shareholders, by voting no, implicitly said: we believe the company is worth more than $9B in the long run, but only if the AMD partnership and the AI pivot deliver. This is a high-stakes rejection of a guaranteed exit in favor of an uncertain, but potentially much larger, future.
From a technical perspective, the conversion of Bitcoin mining infrastructure into AI compute is not a simple hardware swap. In my years auditing mining operations, I’ve seen the difficulty of repurposing ASIC-optimized facilities for general-purpose compute. Bitcoin miners are designed for specific, energy-intensive SHA-256 hashing. They produce tremendous heat, require massive power, and are typically housed in low-cost, often remote locations. AI workloads, on the other hand, demand NVIDIA or AMD GPUs, liquid cooling, high-density racking, InfiniBand or RoCE networking, and a sophisticated software stack for orchestration and model training. The engineering challenge is akin to converting a cargo ship into a cruise liner—the hull and engines are there, but everything else needs to be rebuilt.
Core Scientific’s existing power purchase agreements (PPAs) are the real asset. These long-term contracts, often locked in at low rates, are the economic moat that attracted both CoreWeave and AMD. In an era where AI data centers are desperate for power, having access to cheap, reliable electricity is a strategic advantage. The company’s mining facilities already have substations, transformers, and cooling infrastructure. But the final mile—the GPU clusters, the networking, and the software—remains unproven at scale. The AMD partnership, as announced, lacks any technical details: no hash rates, no wattage commitments, no delivery timelines. This is a strategic announcement, not a technical milestone, and the market should treat it as such.
During my time working with an AI verification protocol, I witnessed the gap between the promise of decentralized compute and the reality of centralized GPU clusters. AMD’s Instinct GPUs and ROCm software are still catching up to NVIDIA’s CUDA ecosystem. For a company like Core Scientific to succeed, it must not only deploy AMD hardware but also optimize its software stack to handle the demands of AI customers. The risk of supply chain concentration is real: if AMD’s chip yields disappoint or if CUDA’s network effects lock in developers, the entire pivot could stall. The shareholders’ rejection of the $9B sale is a vote of confidence, but it also raises the pressure on management to execute flawlessly.
Now, the contrarian angle that the market is missing: the AMD partnership might be a distraction. The real value of Core Scientific is not in the GPU hardware but in the power contracts and the operational expertise of running Bitcoin mining. The AI pivot requires massive capital expenditure—new cooling systems, high-density racking, and networking gear that the company must finance through debt or equity dilution. The shareholders rejected a $9B acquisition, but they may end up with far less if the pivot fails. Moreover, the AI compute market is already crowded with hyperscalers like AWS, Azure, and Google Cloud, as well as specialized AI cloud providers like CoreWeave. Core Scientific’s geographical advantage (remote mining sites) may be a liability for AI workloads that require low-latency connections to data centers in urban centers. The company’s plan to host AI at mining sites in Texas or Kentucky could face latency and bandwidth issues that undermine its value proposition.
Another blind spot is the environmental and regulatory scrutiny. Running both Bitcoin mining and AI compute from the same facilities amplifies the carbon footprint. As AI draws increasing attention for its energy consumption, regulators may target facilities that combine both mining and AI. Core Scientific’s shareholders may have underestimated the reputational risk of being seen as a double polluter.
A deeper, more philosophical question arises: what does this acquisition rejection mean for the ethos of decentralization? Core Scientific is a publicly traded company, not a DAO. Its shareholders voted based on financial returns, not on ideological alignment with Bitcoin’s promise of censorship resistance. The company’s pivot to AI is a pragmatic move, not a values-driven one. But it highlights a tension within the Bitcoin mining industry: miners are increasingly becoming centralized infrastructure providers for AI, which itself is a centralized technology. The mining industry’s original purpose—to secure a decentralized, permissionless network—is being subsumed by the demands of AI compute. This is not necessarily bad, but it is a profound shift. The same ASICs that once resisted corporate control are now being repurposed to serve the most centralized of industries.
In my experience as an open source evangelist, I’ve seen how narratives can mask technical realities. The AMD partnership is a narrative tool—a way to signal that Core Scientific is not just a mining company but a next-generation infrastructure provider. But the technical reality is that the company must solve a dozen engineering challenges before it can deliver on that narrative. The shareholders’ rejection of the $9B sale is a bet that the company can overcome these challenges. It is also a bet that the market will continue to value AI infrastructure at a premium, even as the hype cycle matures.
The takeaway is this: The true test of Core Scientific’s thesis will not be in the next quarterly earnings, but in whether they can deliver AI compute at scale without sacrificing the existential promise of decentralization. The rejection of the $9B acquisition is not a victory; it is a mandate. The shareholders have demanded that management create more value than $9B—not through financial engineering, but through real, operational excellence. The AMD partnership is a tool, but the real work is in the details: the cooling, the networking, the software, and the trust. In a world where AI compute is becoming increasingly centralized, the mining industry’s infrastructure is a last bastion of distributed resource ownership. But only if the engineering puzzles are solved. The shareholders have placed their bet. Now, the engineers must deliver.
