The AMD-Core Scientific Partnership: A Data-Driven Dissection of the Warrants Signal
On July 15, Core Scientific (CORZ) announced a partnership with AMD that sent the stock up 8% in pre-market trading. The press release painted a picture of a visionary pivot: 500 megawatts of computing capacity, expandable to 2.5 gigawatts, dedicated to advanced AI workloads. The market cheered. But data does not lie; it only reveals hidden patterns. A forensic look at the warrant structure — AMD received the right to purchase CORZ shares at market price with zero premium — tells a different story. This is not an endorsement. It is a hedge.
Core Scientific emerged from Chapter 11 bankruptcy in January 2024, burdened by debt and an outdated mining fleet. The company’s core business remains Bitcoin mining, but the post-halving environment has squeezed margins. Hashprice — the revenue per terahash per day — has fallen 18% over the past 90 days, while network difficulty hit an all-time high. In response, Core Scientific began converting its Texas and Kentucky sites to AI data centers, a capital-intensive move that requires new cooling systems, fiber connectivity, and most critically, a steady supply of high-performance chips. Enter AMD.
The partnership grants AMD access to over 500 MW of existing capacity for its Instinct GPUs, with an option to scale. In return, Core Scientific receives a warrant to purchase CORZ common stock at the prevailing market price — currently around $3.50 — with no additional premium or discount. On its surface, this looks like a standard strategic investment. But the terms reveal a lopsided risk profile. AMD does not commit to any minimum purchase of chips or guaranteed revenue. The warrants are entirely at AMD’s discretion to exercise within five years. If Core Scientific’s AI business flops, AMD walks away with zero cost. If it succeeds, AMD profits from equity appreciation. This is a free call option on another company’s execution.
The 500 MW figure is misleading. In Bitcoin mining terms, 500 MW at modern efficiency (~28 J/TH) translates to roughly 17.8 EH/s — a substantial but not dominant share of the 600 EH/s network. But AMD’s chips are GPUs, not ASICs, meaning this capacity is not for mining. It will be used for AI inference and training, a market already crowded by AWS, CoreWeave, and Lambda Labs. The 2.5 GW expansion target is even more speculative: it requires zoning approvals, grid interconnection agreements that can take 18–36 months, and an estimated $3 billion in capital expenditure. Core Scientific’s market cap is $700 million. To fund this, the company will likely issue new debt or equity, further diluting existing shareholders.
My 2024 Bitcoin ETF inflow study taught me that institutional capital flows tell the real story. In the 90 days leading up to this announcement, Core Scientific’s largest institutional holders — Vanguard, BlackRock — held steady, but hedge fund positions dropped 12%. Meanwhile, short interest rose to 15% of float. The warrants, if exercised, could increase the share count by up to 8% based on current outstanding shares (assuming AMD fully exercises a warrant block estimated at 10 million shares). That dilution is manageable, but it signals that AMD sees value only if Core Scientific hits its transformation targets.
Data does not lie; it only reveals hidden patterns. The hashprice decline, the short interest increase, and the zero-premium warrants form a consistent picture: the market is skeptical, and AMD is not paying a premium to participate. The partnership is an option, not an investment.
Background is essential here. Core Scientific’s financials, drawn from the latest 10-Q filing, show non-GAAP net income of -$45 million for Q1 2024. Mining revenue was $95 million, down 30% year-over-year. AI hosting revenue was $12 million — 12% of total. The company projects AI hosting to reach 30% of revenue by Q4 2025, but that assumes a 2x increase in contracted capacity. The AMD deal brings capacity, not demand. Core Scientific must still find paying AI customers, a challenge given that hyperscalers like Azure and Google are building their own data centers.
Contrarian analysis reveals a hidden risk: the AMD partnership may actually accelerate Core Scientific’s dependence on a single technology vendor. Bitcoin mining uses ASICs from Bitmain and MicroBT, which are commodity hardware. AI GPUs from AMD are proprietary, and once a data center is optimized for AMD’s ROCm software stack, switching to Nvidia becomes costly. If AMD lags in performance (as recent benchmarks suggest), Core Scientific’s AI business could become locked into suboptimal hardware. The warrants, meanwhile, give AMD an incentive to keep Core Scientific alive but not necessarily profitable. AMD could dilutely exit if the company falters.
My 2022 LUNA/UST post-mortem documented how centralized partnerships with asymmetric payoff structures often prefigured collapse. Terra’s early investors received warrants that they rapidly sold, masking insider distribution. Here, AMD’s warrants are not a vote of confidence; they are a calculated bet on a binary outcome: either Core Scientific transforms into a top-tier AI provider, or the warrants expire worthless and AMD loses nothing.
Hashrate is the ultimate validator, and it’s not yet convinced. Core Scientific’s Bitcoin hashrate has remained flat at 15 EH/s for six months, while competitors like Riot added 3 EH/s. The AI pivot is essentially an admission that Bitcoin mining alone cannot sustain the company. But the on-chain data for Bitcoin — exchange reserves declining, long-term holder accumulation rising — suggests a bullish cycle that could lift mining revenue in 2025. If Bitcoin rallies to $100k, Core Scientific’s mining business may outperform its AI aspirations, making the AMD deal an expensive distraction.
The market often confuses narrative with reality. This partnership will likely push CORZ stock to the $4.00–$4.50 range in the short term, driven by retail speculation. But the warrants cast a long shadow. My 2020 Uniswap liquidity mapping refined my understanding of how smart money positions before inflection points. In the derivatives market, put/call ratios for CORZ options have skewed 2:1 bearish since the announcement, indicating hedging against dilution.
Takeaway: The only numbers that matter are the ones in the SEC filings, not the press releases. The partnership is a necessary but insufficient step for Core Scientific’s survival. The next signal is Q3 2024 earnings: if AI hosting revenue fails to exceed 15% of total, the warrants become a deadweight. If it exceeds 25%, the warrants may be exercised and the dilution temporary. Data will reveal the truth. Until then, treat this as a strategic alignment of struggling entities, not a paradigm shift.