Everyone cheered the AMD-Core Scientific announcement as a natural marriage of two computing giants. I saw a warrant-splattered deal that reeks of optionality. Code is law, but bugs are justice. The bug here is the asymmetry in who gets to benefit from the 'AI pivot' first. The headline reads: 'AMD to Acquire Core Scientific Warrants, Access up to 2.5GW of Computing Power.' But the market quickly priced in the spin—Core Scientific shares popped 8% in pre-market. I watched the order book and saw algos buying the narrative, not the numbers. That’s my cue to dig deeper.
Context: The Pivot from ASIC to GPU
Core Scientific, once a top-tier Bitcoin miner with tens of thousands of ASIC rigs, is now a landlord for AI workloads. The deal with AMD grants the chip giant warrants to purchase CORZ shares at a price tied to the market (no discount disclosed), in exchange for providing up to 2.5 gigawatts of computing capacity. The first 500MW is already live; the rest is a roadmap to 2.5GW. For context, 2.5GW is roughly the power consumption of a mid-sized US city. This isn’t about mining blocks anymore. It’s about hosting NVIDIA H100s and AMD Instincts for inference training.
But here’s the uncomfortable truth: Bitcoin miners are inherently bad at AI. Their infrastructure is built for high heat, ASIC-specific power curves, and minimal latency tolerance. AI data centers require precision cooling, fiber-dense networking, and uptime SLAs that make mining look like a hobby. Core Scientific has been experimenting with AI hosting for over a year, and the early results are mixed. They announced a partnership with CoreWeave last year, but that hasn’t materially moved the revenue needle yet.
Still, the market is hungry for any story that combines AI and crypto. The AMD deal fits neatly into that narrative. But as a trader who lived through the 2017 ICO audit cycle, I learned to read the contract, not the news release.
Core: The Warrants Are the Real Story
Let’s break down the structure. AMD receives warrants to buy Core Scientific shares at ‚Äúmarket price‚Äù—essentially a call option with zero premium. If CORZ runs to $20, AMD buys cheap. If CORZ drops to $2, AMD walks away. No capital at risk. This is a free one-way bet on the upside, paid for by the promise of future compute supply.
From a derivatives perspective, this is a synthetic long position with a free put. The implied volatility of CORZ options will now compress because the warrants act as capped upside for AMD but unlimited downside for existing holders if the shares are diluted. The trick is the warrants are “at-the-money” on issuance, but they are typically structured with a 5-year term and a cashless exercise provision. That means AMD can convert without injecting cash, further diluting shareholders.
Let’s run the numbers. Core Scientific currently has about 180 million shares outstanding. If AMD exercises all warrants—assuming the typical warrant cover ratio of 15% (not disclosed, but industry average)—that adds 27 million shares. Dilution of ~15%. That’s not catastrophic, but it’s a headwind. More importantly, the dilutive overhang depresses the stock by 5-10% in theory, as the market prices in future supply.
But the bigger risk is the required capital expenditure. To build 2.5GW of AI-ready capacity, Core Scientific needs to spend $2-3 billion. They have $100 million in cash. The rest will come from debt or equity. More dilution. Or they could use the warrants as a funding mechanism—a trick I saw in 2019 when some miners issued convertible notes with warrants attached. The problem is that once the market realizes the warrants are a down payment on future dilution, the stock becomes a delta-one trade on the success of the AI pivot, not a pure Bitcoin proxy.
Contrarian: The Retail Blind Spot
Retail traders see the AMD logo and think “big tech validation.” They buy the stock and the narrative. But institutional money reads the warrants as a floor on upside: AMD caps its exposure and gets free upside, while retail holds the bag for execution risk.
I remember the 2021 NFT floor price manipulation—traders on Twitter celebrated BAYC rising, while I was shorting Aave because the wash-trading pointed to a crash. Same dynamic here. The market is euphoric about the AI pivot, but the structural math says the stock is priced for perfection. Net dollar per megawatt in AI hosting is $1.5-2 million per MW per year, but after power, cooling, and staffing, margin is 20-30%. At 2.5GW, that’s $500-750 million EBITDA at best. That gives a 30x+ EBITDA multiple on the current market cap—rich for a company with no recurring revenue guarantee.
Furthermore, AMD isn’t giving anything away. They get a captive customer for their Instinct GPUs in a world where Nvidia dominates. But if AI demand slows, Core Scientific is stuck with empty racks and high debt. Meanwhile, AMDs warrants remain in the money on any positive stock movement.
This deal is structurally bullish for AMD, neutral for CORZ in the short term, and bearish for CORZ if you account for future dilution and capex. The contrarian trade is to short CORZ and long AMD—or better yet, trade the volatility. The CORZ options chain has been abnormally quiet; I set a strangle before earnings to capture the move when the market realizes the warrants are a hidden liability.
Takeaway: Actionable Levels
The real insight isn’t in the press release. It’s in the SEC filing that will detail the warrant terms. I’m watching for the exercise price—if it’s at a discount to the 10-day VWAP, the dilution risk is higher. If it’s at a premium, AMD is sending a signal of long-term commitment. Until then, the stock trades on sentiment.
Key level: $8.50 support. If CORZ breaks below that, the stock revisits the lows from last year. If it holds and closes above $9.50, the warrants become a buy signal for smart money. I’m small on the long side, but I’ve also hedged with puts on CORZ to protect against the realization that the AI pivot is a multi-year journey with high execution risk.
The market doesn’t want to hear this, but: the AMD deal is a structural hedge for AMD, not a vote of confidence for Core Scientific. With the Greeks, I can see the implied vol is pricing in a 15% move by next earnings. That’s where the real edge is. Not in the narrative, but in the options chain.
Final thought: Code is law, but bugs are justice. The bug in this deal is the asymmetric payoff for the two sides. If you’re trading the stock, you’re trading a binary bet on the success of the AI pivot. If you’re trading the warrants, you’re trading a free option. Know which side of the contract you’re on.
Based on my audit experience from 2017, I’ve seen how these partnerships often mask underlying risks. The ICOs had smart contract flaws; this deal has structural flaws. The NFT floor is a feeling, not a number—and so is the current market cap of Core Scientific. Wait for the 10-Q, then decide.