Observe the headline: Core Scientific shareholders reject a $9 billion sale. The market applauds. The stock jumps. The narrative shifts to an AMD partnership as the new growth engine. Stop. Look at what is missing.
Silence in the code is the loudest warning sign. Here, the silence is not in code but in operational data. No megawatt capacity delivered. No utilization rates. No revenue share from the AMD deal. Just a press release and a rejected buyout. The market is betting on a transformation that has not yet passed a single technical audit.
Context: Core Scientific is a Nasdaq-listed Bitcoin mining company that survived a 2023 bankruptcy. It now pivots to AI data center hosting. The pivot is real: they signed long-term contracts with CoreWeave in 2024. The AMD partnership, announced in early 2025, is the next step. The narrative is seductive: reuse mining infrastructure, deploy AMD GPUs, and capture AI compute demand. The shareholders rejected a $9 billion buyout, implying they believe the company is worth more. The question is not whether the pivot is possible, but whether the market has already priced in success without evidence.
Core: Systematic Teardown
Technical Feasibility vs. Technical Reality
Converting a Bitcoin mining facility into an AI data center is an engineering challenge, not a plug-and-play upgrade. Bitcoin mining demands cheap power and simple ASIC rigs. AI training demands high-density racks, liquid cooling, low-latency networking (InfiniBand or RoCE), and massive GPU clusters. Core Scientific has experience in power procurement and facility management, but AI hosting requires a different skill set.
Based on my audit experience, I have seen projects overestimate their ability to bridge infrastructure gaps. In 2017, Tezos claimed formal verification solved all security issues. The reality was different: type-safety vulnerabilities in the implicit liquidity pools proved that cryptographic proof did not equal functional safety. Similarly, Core Scientific’s AMD partnership is a statement of intent, not a proof of capability. The article provides zero technical details: no benchmark results, no cluster size, no cooling solution, no timeline for first delivery. The market is buying a story, not a spec.
The AMD Dependency Risk
AMD’s Instinct GPUs are competitive on paper, but the software ecosystem lags Nvidia’s CUDA. ROCm, AMD’s open-source stack, has improved, but production-grade AI workloads—especially for large language models—still overwhelmingly run on CUDA. Core Scientific will need to either accept a limited market of customers willing to optimize for AMD or invest heavily in software engineering to bridge the gap. Complexity is often a veil for incompetence. Here, the complexity of the AMD stack is a real engineering barrier, not a marketing gimmick.
Moreover, AMD’s supply chain is not immune to constraints. If wafer allocation or packaging yields fall short, Core Scientific’s capacity commitments become unfulfillable. The article does not disclose any minimum purchase quantities or take-or-pay clauses. The partnership is a handshake, not a contract.
Financial Structure: The Missing Token
Core Scientific has no native token. There is no staking, no yield farming, no governance token to inflate. The value capture is purely through equity. Shareholders rejected $9 billion, setting a floor on their expectations. But how will that value be created? The company needs massive capital expenditure to retrofit mining sites. The likely mechanism is equity dilution or debt. Both hurt existing shareholders. The AMD partnership, if it is merely a procurement agreement, does not guarantee revenue. It only guarantees that Core Scientific will buy AMD hardware. The revenue depends on finding customers who will pay for that compute.
Trust is a variable, verification is a constant. The article presents the AMD deal as a value driver, but it is a cost center until customers sign contracts. The market has not seen those contracts.
Contrarian: What the Bulls Got Right
The bulls have a case. The underlying asset—cheap power—is real. Core Scientific’s mining sites are often located in regions with stranded energy or fixed-price power purchase agreements. That is a structural advantage. Pure AI cloud providers like CoreWeave or traditional data centers pay market rates. Core Scientific can offer lower cost compute, which is a competitive edge.
Second, the AI compute demand is not a bubble. It is a secular trend. Even if the AMD partnership takes time to yield results, the long-term thesis of owning and operating infrastructure is sound. The 2024 CoreWeave contracts provide a revenue base. The AMD partnership adds diversification. If AMD’s market share grows, Core Scientific will be early.
Third, the rejection of the $9 billion sale could be interpreted as a signal of board confidence. Perhaps the board has inside knowledge of a pipeline of high-margin AI contracts. If that is true, the market is correctly pricing in a premium.
But these are assumptions. The burden of proof is on the company.
Takeaway: Accountability Call
The market is treating Core Scientific’s pivot as a fait accompli. It is not. The AMD partnership is a blank check on engineering execution. The shareholders have rejected a known exit for an unknown future. That is a bet, not a strategy.
Watch for the following signals: first, a published benchmark of AMD Instinct GPUs in a Core Scientific facility. Second, a concrete customer announcement for the AMD workload. Third, the company’s capital expenditure plan and how it will be funded. Until then, the silence in the data is the loudest warning.
I have seen this pattern before. In 2021, Axie Infinity’s dual-token model looked like a flywheel. I calculated the inevitable hyperinflationary spiral. The community called me a bear. The crash came. In 2022, Terra’s algorithmic stability mechanism relied on infinite liquidity. I verified the math. The collapse followed. Core Scientific is not a DeFi protocol, but the same principle applies: verify the mechanism, not the narrative.
Complexity is often a veil for incompetence. The AMD partnership is complex. The engineering challenges are real. The market is ignoring the complexity because it wants the story. I prefer to wait for the data.
Final judgment: The rejection of the $9 billion sale is a high-risk signal. The market has priced in a successful transformation without proof. That is a dangerous asymmetry. The buyer was willing to pay $9 billion in cash. Shareholders said no. They now need to deliver more than that. The margin for error is zero.
Postscript: A Personal Note
In 2024, I re-audited EigenLayer’s slashing conditions. I found edge cases where restaked assets could be doubly slashed under specific network partitions. The developers acknowledged the issue. The ecosystem was grateful. But the lesson is that even the most elegant designs have hidden failure modes. Core Scientific’s infrastructure pivot is no different. The failure modes are not in smart contracts but in supply chains, software stacks, and customer adoption.
Check the math, ignore the hype. The math on Core Scientific is incomplete. The equations are missing variables: capacity, utilization, revenue per MW. Until those variables are filled, the model is nonsense.
Tags: Core Scientific, CORZ, Bitcoin Mining, AI Data Center, AMD Partnership, Infrastructure Analysis, Due Diligence, Cold Dissection