The news broke on Tuesday: Core Scientific shareholders formally rejected a $9 billion acquisition offer. The board instead pointed to a newly announced partnership with AMD as the path to greater value. The market reacted with a modest 2% bump in CORZ shares. But as a 7x24 market surveillance analyst who has spent the better part of a decade auditing the gap between crypto press releases and on-chain reality, I find this narrative thread dangerously thin.
Context: Why Now? Core Scientific is a publicly traded Bitcoin mining operator (NASDAQ: CORZ) that has been pivoting toward AI/HPC data center hosting since its 2023 bankruptcy restructuring. The company operates large-scale mining facilities across the United States, with access to cheap power locked in through long-term purchase agreements. The strategic pivot is straightforward: repurpose existing power infrastructure—originally built for ASIC miners—to house GPU clusters for AI workloads. The AMD partnership, announced alongside the rejection of the $9 billion offer, is positioned as the linchpin of this transformation.
But the devil, as always, lives in the technical details. Let me be clear: this is a strategic announcement, not a technical milestone. The press release contains no delivered megawatts of GPU capacity, no benchmark results, no software integration lab results, and no revenue guarantees. My 2026 AI-Crypto convergence audit—where I dismantled a $50 million valuation fraud by demanding access to smart contract logic—taught me that when a company claims to be building the next-generation infrastructure, the absence of verifiable technical data is the first red flag.
Core: The Facts and Immediate Impact Let’s reconstruct what we actually know:
- Shareholders rejected a $9 billion acquisition. The offer came from an undisclosed buyer (likely a private equity consortium). The rejection signals that the board believes the AMD partnership will unlock more than $9 billion in long-term equity value. This is a high-stakes bet.
- The AMD partnership exists. Core Scientific will deploy AMD Instinct GPUs in its data centers. The agreement includes joint engineering optimization—meaning AMD will help Core Scientific design the cooling, networking, and power systems for AMD hardware.
- No financial terms were disclosed. There is no minimum purchase commitment, no revenue-sharing agreement, no locked-in pricing. The partnership is a framework, not a contract with teeth.
- Core Scientific already has an existing hosting contract with CoreWeave (a pure-play AI cloud provider) for Nvidia-based infrastructure. The AMD deal is a diversification of chip supply, not a replacement of the existing partnership.
- The company carries debt from its 2023 bankruptcy. The restructuring plan involved converting some debt into equity, but the exact leverage ratio remains opaque. AI infrastructure buildout requires massive capital expenditure—estimates run at $5-10 million per megawatt for a high-density GPU cluster.
From a technical due diligence perspective, the immediate impact is binary: either the partnership leads to a verifiable increase in deployed AI capacity within 12 months, or it remains a marketing narrative. The market is currently pricing in the former. My conservative estimate, based on my experience auditing the 2020 DeFi Summer protocols, suggests that the probability of material delivery within 12 months is less than 30%. The reason is not malice—it’s execution complexity.
Contrarian: The Unreported Blind Spots The mainstream coverage has focused on the “strategic pivot” and “shareholder confidence.” But three critical blind spots are being ignored:
1. The AMD CUDA Gap. Nvidia’s dominance in AI is not just about hardware—it’s the CUDA software ecosystem. AMD’s ROCm stack has improved significantly but still lags in library support, tooling maturity, and developer mindshare. Core Scientific will need to build a support team that can handle ROCm-specific deployment issues. This is not a plug-and-play swap. My 2017 ICO audit sprint taught me that the most dangerous assumption in crypto is “it’s just a matter of scaling.” Code is not transitive. Neither is software stack compatibility.
2. The Power Infrastructure Mismatch. Bitcoin mining facilities are designed for ASICs, which have different power density, cooling requirements, and network topology than GPU clusters. A typical ASIC miner runs at 3000W per unit and can be air-cooled. A modern GPU cluster like Nvidia’s H100 rack requires liquid cooling, higher power density (up to 40kW per rack), and InfiniBand networking. Converting a mining facility to support this is not a simple retrofit—it often requires a complete rebuild of the electrical and cooling systems. Core Scientific has not disclosed how many of its sites are suitable for conversion, nor the capex required.
3. The Revenue Model Ambiguity. The AMD partnership is framed as a value driver, but the revenue model is unclear. Will Core Scientific sell GPU compute time on a pay-as-you-go basis? Will it secure long-term hosting contracts with AI startups? Or is this a zero-margin hardware integration deal where AMD pays Core Scientific to host its hardware for demonstration purposes? The press release is silent. Without a clear revenue structure, the $9 billion valuation anchor is pure speculation.
Takeaway: What to Watch Next Ledgers don’t lie. Press releases do. The only signal that matters for Core Scientific is the first quarterly report showing actual deployed AI capacity in megawatts—not promises, not partnerships, not strategic pivots. If the company can demonstrate 10-20 MW of GPU capacity live and generating revenue, the AMD narrative has legs. If not, the $9 billion rejection will be remembered as one of the most expensive votes of confidence in 2026.
I will be watching the next SEC filing for the “Property, Plant, and Equipment” line item. A significant increase in capital assets without a corresponding increase in revenue will tell me that the company is spending money on infrastructure that may not be economically viable. The market is betting on execution. I am betting on data.