The Warrants That Whispers: Deconstructing the Core Scientific-AMD Computing Pact

When AMD announced it would acquire 500 megawatts of computing capacity from Core Scientific with an option to expand to 2.5 gigawatts, the press release was read as a victory lap for the mining company's pivot to AI. The market nudged CORZ shares upward. But the warrants—priced at market value with no discount—tell a different story. They are not a premium on future growth. They are a hedge against failure. Proof exists; it is merely waiting to be verified. In this case, the proof lies in the fine print of the warrant agreement. AMD secured the right to buy Core Scientific shares at the prevailing market price at the time of exercise. No discount. No lock-in. That is not a vote of confidence. It is an option to exit if the AI compute narrative fails to materialize. Core Scientific emerged from bankruptcy in early 2024, burdened by debt and a fleet of ASICs that were becoming obsolete with the Bitcoin halving. The company's leadership pivoted to AI computing, repurposing its power infrastructure to host GPUs for inference workloads. The AMD deal was supposed to validate that strategy. Yet the warrant structure suggests AMD sees this as a low-risk trial, not a long-term bet. Context: The Bitcoin mining industry is in a contraction phase. Post-halving, miners with inefficient hardware are shutting down. The survivors are diversifying into AI compute, where margins are thinner but demand is growing. Core Scientific has 2.5 gigawatts of power capacity, most of it under long-term contracts. That power is the real asset—not the miners, not the GPUs. AMD needs that power to scale its data center footprint without building its own substations. But AMD is not paying upfront. It is paying with warrants that dilute existing shareholders only if the stock rises. Core: Let me walk through the numbers based on my forensic audit of similar partnerships. The warrants cover a number of shares equivalent to a percentage of the deal value. Assuming the 500 MW phase is valued at $1.5 billion (based on typical buildout costs of $3 million per MW for AI data centers), AMD likely received warrants for 5–10% of Core Scientific's diluted shares. At current market capitalization of roughly $2 billion, that is $100–200 million in potential dilution. If AMD exercises when CORZ is trading at $10, the company issues new shares and receives cash. But if the stock falls, AMD simply waits—or lets the warrants expire. The partnership becomes a one-way bet for AMD. The algorithm remembers what the witness forgets. The market memory of Core Scientific's bankruptcy is short. The company restructured its balance sheet, but the debt has been replaced by equity that is still vulnerable to Bitcoin price swings. The AI pivot is not a hedge; it is a second revenue stream with its own risks. GPU compute is competitive. Core Scientific will compete with established players like CoreWeave, Lambda Labs, and even hyperscalers like AWS. The company's advantage is cheap power, but that advantage erodes as more miners enter the same game. I have spent the past three years auditing mining companies' energy contracts and hardware procurement. The typical power purchase agreement for a mining site is structured with a fixed price per kilowatt-hour, but that price is often indexed to local rates. In Texas, where Core Scientific has several sites, winter storms have caused spot prices to spike. The company's backup generators can cover shortfalls, but at a cost. The AMD deal does not solve this; it merely shifts the computing load to higher-margin AI workloads that require near-100% uptime. That increases operational risk. Technical analysis of the infrastructure: To host AMD's GPUs, Core Scientific must retrofit existing mining halls. Mining uses air cooling and high-voltage DC power. AI compute requires liquid cooling and redundant network connectivity. Retrofitting costs about $2 million per MW, according to industry reports. For 500 MW, that's $1 billion in capital expenditure. Core Scientific will need to raise debt or issue more equity. The dilution from the AMD warrants is just the beginning. The contrarian angle: Bulls argue that the partnership gives Core Scientific a captive customer with deep pockets. AMD will use the capacity for training and inference of its own models, potentially becoming a long-term anchor tenant. The warrants, they say, align incentives. If AMD exercises, it shows confidence. But the lack of a discount indicates AMD is not paying for exclusivity. It can walk away without penalty. The real value for AMD is the ability to secure power without committing capital to construction. For Core Scientific, the partnership provides a floor on utilization—but at a price that may not cover the debt service on the retrofitting. Ledgers balance, but ethics remain uncalculated. The ethical question here is whether Core Scientific is transparent with shareholders about the true cost of the AI pivot. The press release highlighted the gigawatt-scale potential. It did not mention the retrofitting cap-ex, the power price risk, or the dilution from warrants. The algorithm remembers what the witness forgets: when mining companies overpromised on AI revenue in 2021, many ended up selling their hardware at a loss. Takeaway: The Core Scientific-AMD deal is a textbook case of financial engineering disguised as technological synergy. The warrants are not a signal of confidence; they are a free option for AMD. Shareholders should demand a clear accounting of the retrofitting budget and the power price sensitivity. Without that, the gigawatt story remains as hollow as the promises of the last bull run. The ledger will show whether this partnership created value or just another layer of complexity. And I will be watching the SEC filings for the first signal of exercise.