Shareholders vote against a $9 billion exit. That is not a vote of confidence. It is a bet on a narrative shift. The data from the proxy statement tells a story of a company that sees itself as something more than a mining farm. The question is: does the technology stack support the valuation?
I traced the capital flows back to the genesis block of this deal. The only way to understand the rejection is to map the liquidity pools of the company's balance sheet and the technical reality of its pivot.
Context: From Bankruptcy to AI Hosting
Core Scientific emerged from Chapter 11 in early 2024 after a brutal crypto winter. The restructuring left it with a legacy debt load and a management team that needed to prove the mining model was not obsolete. The pivot to AI data center hosting was not a choice—it was a survival mechanism.
In 2023, the company signed a multi-year hosting contract with CoreWeave, a pure-play AI cloud provider. That deal provided a revenue stream not tied to Bitcoin price. It also signaled that the 200+ megawatts of power capacity, originally built for ASICs, could be repurposed for GPUs.
Then came the AMD partnership. The news broke alongside the shareholder vote to reject a $9 billion acquisition offer from an unnamed buyer. The market interpreted the rejection as a signal that management believed the AMD deal would create more value.
But the on-chain data—if we consider the company's financial statements as a public ledger—shows a different story. The rejection is a high-risk bet on technical execution.
Core: The Technical Reality of the Power-to-GPU Conversion
Every transaction leaves a scar on the ledger. The scars on Core Scientific's balance sheet are deep. The company carries debt from the restructuring, and the capital expenditure required to convert a mining facility to a GPU data center is not trivial.
Let me break down the technical requirements based on my experience auditing DeFi protocols and infrastructure projects.
First, the power infrastructure. Mining facilities are designed for high-density ASIC loads, but ASICs are air-cooled and operate at lower power densities per rack than modern GPUs. An NVIDIA H100 GPU consumes 700W, and a standard rack can hold 10-20 GPUs. That is 7-14 kW per rack, requiring liquid cooling for high-density configurations. Core Scientific's existing facilities may have the total power capacity, but the distribution, cooling, and networking need a complete overhaul.
Second, the network. Mining operations use a simple TCP/IP stack for Stratum protocol. AI workloads require InfiniBand or RoCE v2 for low-latency GPU-to-GPU communication. The networking infrastructure must be rebuilt from scratch, including new switches, cables, and possibly new fiber runs.
Third, the software stack. AMD's ROCm ecosystem is improving, but it is not a drop-in replacement for CUDA. Most AI training frameworks are optimized for CUDA, and porting to ROCm requires engineering effort. Core Scientific will need to provide engineering support to its customers or risk underutilization of the GPUs.
The AMD partnership announcement lacked any specific technical milestones. No hash rate numbers. No power usage effectiveness (PUE) targets. No deployment timelines. This is a strategic announcement, not a technical one. Based on my experience in 2017 auditing ICO whitepapers, I know that narrative value often diverges sharply from technical reality.
The liquidity pool is a mirror, not a reservoir. The liquidity here is the company's ability to raise capital for the conversion. The shareholder vote to reject $9 billion means they are betting the company can achieve a higher valuation through the AI pivot. But the capital markets are not a reservoir of infinite appetite. Core Scientific will need to issue equity or debt to fund the conversion. The dilution risk is real.
Let me isolate the behavioral pattern of the management team. They rejected a sure exit in favor of a risky pivot. This is typical of a management team that overvalues its own ability to execute. The pattern is visible in the proxy statement: the board argued that the AMD partnership would create "superior long-term value." But the data on AMD's market share in AI is clear. AMD holds less than 10% of the data center GPU market. The partnership is a bet on AMD's ability to compete, which is itself a bet on execution.
Contrarian Angle: The AMD Partnership is a Distraction
The contrarian view is that the real value of Core Scientific is not in its GPU capacity, but in its power infrastructure. The company has long-term power purchase agreements (PPAs) that lock in low electricity rates. These PPAs are the true asset. The GPUs are a commodity.
Whales don't swim in shallow water. The large institutional investors who voted against the sale are likely betting on the power infrastructure, not the AI pivot. They see the AMD partnership as a means to utilize the power capacity, but the actual value creation will come from the power contracts, not the GPU racks.
Consider the alternative: if Core Scientific had accepted the $9 billion, the buyer would have acquired the power contracts. The shareholders are essentially saying that the power contracts are worth more than $9 billion in the long run. That is a bold claim, given the historical volatility of power markets and the regulatory risks around cryptocurrency mining.
Another blind spot: the AMD partnership could actually be a liability. If AMD GPUs underperform relative to NVIDIA, Core Scientific could be locked into a contract that forces them to use inferior hardware. The partnership terms were not disclosed, but typical OEM agreements include minimum purchase commitments. If the ROCm software stack is not ready, the GPUs will sit idle, and the company will still have to pay for them.
Takeaway: The Next Signal
The next signal to watch is not the stock price. It is the deployment of the AMD GPUs. I will be tracking the on-chain data of the company's wallet addresses to see if they start purchasing large quantities of AMD hardware. But more importantly, I will be watching for the actual utilization rates of the data centers. Without that data, the AMD partnership is just a press release.
Tracing the ghost coins back to the genesis block. The genesis block of this story is the bankruptcy restructuring. The ghost coins are the hidden liabilities. The shareholders have made a bet that the power contracts are worth more than $9 billion. The data will tell us if they are right.
