Hook:
On March 28, 2026, Core Scientific shareholders voted down a $9 billion acquisition offer. The decision signaled a collective bet that the company’s pivot from Bitcoin mining to AI data center hosting would generate more long-term value than a clean exit. But the accompanying announcement of a partnership with AMD to supply Instinct GPUs raised more questions than it answered. The combination of a rejected sale and a high-profile chip deal creates a classic tension: market optimism versus structural reality.
Context:
Core Scientific is a Nasdaq-listed infrastructure firm (CORZ) that operates Bitcoin mining facilities. Since emerging from Chapter 11 in early 2024, it has pursued a dual strategy: continue mining Bitcoin while repurposing its power infrastructure for AI and HPC workloads. The AMD partnership, described as a “strategic collaboration” to deploy GPUs for AI compute, is the latest step in this transformation. The company already has a hosting contract with CoreWeave, an AI cloud provider, but the AMD deal is framed as a direct supply agreement that could expand its capacity.
Shareholders rejected the $9 billion acquisition offer—a 30% premium to the pre-announcement market cap—implying they believe the company’s standalone value exceeds that figure. The AMD partnership is the primary narrative supporting that belief. However, the announcement contained no technical details: no committed megawatts, no GPU count, no delivery timeline, no revenue share structure. The entire bullish case rests on the assumption that Core Scientific can successfully convert its mining sites into AI-ready data centers and that AMD’s Instinct GPUs will be competitive with Nvidia’s H100/B200 in production workloads.
Core:
The technical feasibility of a mining-to-AI data center conversion is not trivial. It requires liquid cooling, high-density rack layouts, InfiniBand or RoCE networking, and GPU cluster orchestration—capabilities that are fundamentally different from running ASIC miners. Core Scientific’s existing power contracts (long-term, low-cost electricity) are a genuine asset, but the engineering challenge lies in the electrical and thermal infrastructure, not just the power price. From my experience auditing smart contracts and modeling DeFi cascades, I see a pattern: the market often conflates adjacency with capability. A mining company owns power and real estate, but that does not make it an AI cloud provider.
More critically, AMD’s ROCm software ecosystem remains a generation behind Nvidia’s CUDA in terms of library maturity, framework support, and developer tooling. While AMD’s hardware has improved—the MI300X offers competitive raw performance—the software stack is still catching up. For Core Scientific, this means that deploying AMD GPUs at scale will require significant engineering investment to optimize workloads, troubleshoot compatibility issues, and maintain reliability SLAs. The partnership may include joint engineering support, but the burden of operationalizing the hardware falls on Core Scientific’s team.
Consequently, the AMD deal is a supply-chain diversification move, not a technological leap. The market treats it as the latter. The shareholder vote implicitly priced in a success scenario where Core Scientific captures a meaningful share of the AI hosting market. But the company has not disclosed any key performance indicators: current hosted MWs, utilization rates, or customer commitments beyond the existing CoreWeave contract. The lack of transparency is a red flag for anyone who has modeled infrastructure businesses. Logic is immutable; incentives are the variable. The incentive for Core Scientific’s management is to announce partnerships to justify the stock price; the incentive for AMD is to secure deployment sites for its GPUs to gain market share. Neither party has an incentive to disclose technical risks or timeline delays.
Contrarian:
The contrarian view is that the $9 billion rejection might be a structural mistake. Core Scientific’s balance sheet remains burdened by legacy debt from its 2023 bankruptcy. The AI pivot requires massive capital expenditure—upgrading facilities, buying GPUs, building out networking. If the company needs to raise equity or issue convertible bonds to fund this, existing shareholders will face dilution. The AMD partnership does not eliminate that risk; it amplifies it, because capital commitments are likely contingent on GPU procurement.
Moreover, the AI hosting market is becoming crowded. Traditional cloud providers (AWS, Azure, GCP) have infinite scale. Specialized AI cloud firms (CoreWeave, Lambda, Vultr) have already built optimized infrastructure. Core Scientific’s competitive advantage—cheap power—is real but not unique. Other miners (Riot Platforms, Marathon Digital) are pursuing similar strategies. The market may be overestimating the differentiation. Structural integrity precedes market sentiment. The rejection of the sale means shareholders are betting on execution in a market where execution metrics are opaque.
Takeaway:
Core Scientific’s story is not about innovation; it’s about capital allocation and operational discipline. The AMD partnership is a necessary but insufficient condition for value creation. The real test will come in Q3 and Q4 2026, when the company must report actual AI revenue, capacity utilization, and margin data. Until then, the stock is trading on narrative, not fundamentals. The shareholder vote was a rational bet on optionality, but optionality is not a guarantee.
