The Billion Dollar Gamble on High Performance Computing
The era of the 5.75 per share lowball is officially a relic of a less informed market. On this Tuesday, October 21, 2025, Core Scientific (CORZ) stands as a monument to strategic defiance. Shareholders who once trembled at the prospect of a 1.02 billion dollar buyout from CoreWeave in mid-2024 are now staring at a market capitalization exceeding 4.1 billion dollars. The pivot from pure-play Bitcoin mining to high-performance computing (HPC) has not just occurred; it has redefined the valuation metrics for the entire data center sector. The 270-megawatt hosting contract with CoreWeave, which many skeptics viewed as a desperate survival tactic, has instead become the foundation of a 6.7 billion dollar revenue stream over the next twelve years.
Breaking Down the CoreWeave Hosting Expansion
Power is the new global currency. As of late 2025, the scarcity of energized data center space has reached a fever pitch. Core Scientific managed to leverage its existing infrastructure, specifically its access to 1.2 gigawatts of power, to secure terms that were unthinkable eighteen months ago. Per the latest SEC filings, the company has successfully converted three of its legacy mining halls into Tier 3 equivalent facilities capable of housing NVIDIA Blackwell B200 clusters. This is not a simple hardware swap. The technical overhaul required a total redesign of the electrical distribution, moving from 20kW per rack for air-cooled ASICs to a staggering 100kW per rack for liquid-cooled GPU deployments.
The market has responded with a violent upward re-rating. While pure-play miners struggle with the post-halving economics of 2024, Core Scientific has decoupled its stock price from the daily fluctuations of Bitcoin. Market data from Yahoo Finance indicates a 220 percent year-to-date return, significantly outperforming the broader Bitcoin Mining Index. The reason is simple: yield. An ASIC miner generates revenue based on the hash price, which is currently hovering around 45 dollars per petahash. In contrast, an AI-hosting rack generates upwards of 400 dollars per megawatt-hour in steady, non-volatile rental income.
The Valuation Gap and Shareholder Sentiment
The visualization above tells the story of the 3 billion dollar value creation that occurred when management told CoreWeave no to the initial acquisition. Reflecting on the original June 2024 proposal, it is clear that the bid was a predatory attempt to capture undervalued power assets before the AI infrastructure gold rush fully materialized. Today, the skepticism that once dominated shareholder calls has evaporated, replaced by a demand for faster conversion of the remaining 500 megawatts of available capacity.
Technical Moats and Grid Interconnection
Core Scientific’s primary advantage is not its hardware, but its place in the interconnection queue. In the current 2025 energy environment, securing a new 100-megawatt grid connection in states like Texas or Georgia can take five to seven years. Core Scientific already has the transformers on the ground. By utilizing front-of-the-meter assets, they avoid the transmission delays plaguing new-build data centers. Furthermore, the company has implemented a sophisticated curtailment strategy. During peak load events on the PJM or ERCOT grids, the company can switch off its remaining Bitcoin miners in milliseconds, selling power back to the grid at spot prices that often exceed 2,000 dollars per megawatt-hour, while keeping the mission-critical AI clusters online through redundant backup systems.
| Metric | June 2024 (Actual) | October 2025 (Current) |
|---|---|---|
| Stock Price (CORZ) | $5.75 | $18.42 |
| Total Operational Power | 745 MW | 1.12 GW |
| AI/HPC Revenue % | < 5% | 42% |
| Estimated Yearly EBITDA | $180M | $540M |
Critics who argued that Core Scientific was not a need to have for CoreWeave failed to account for the physical constraints of the AI revolution. There is no AI without power. There is no power without infrastructure. Core Scientific owns the infrastructure. The 200-megawatt expansion announced yesterday is the third such increase in the last six months, proving that the partnership is far more lucrative for Core Scientific as a landlord than it ever would have been as a subsidiary. The operational leverage here is immense: as the company retires older S19 miners and replaces them with B200 GPU clusters, the margin per electron increases by nearly 800 percent.
Management has successfully navigated the transition by focusing on Cooling Distribution Units (CDUs) and secondary loop piping, technical hurdles that have stalled many other mining firms attempting the same pivot. The ability to manage 50kW+ heat loads per rack is the new barrier to entry. Companies like Riot or Marathon, which remain more heavily weighted toward Bitcoin production, are now trading at significant discounts to Core Scientific on an EV/EBITDA basis. This valuation gap reflects the market’s preference for the predictable, high-margin cash flows of the AI sector over the high-beta volatility of the crypto market.
Looking toward the first quarter of 2026, the market is laser-focused on the January 15, 2026, activation of the 112-megawatt Phase 3 expansion at the Denton, Texas facility. This milestone will represent the first time the company’s AI revenue officially eclipses its Bitcoin mining revenue on a quarterly basis. Investors should watch the finalized PJM capacity auction results due in December, as these will dictate the baseline power costs for the 2026 fiscal year and determine if Core Scientific can maintain its current 60 percent plus gross margins in the HPC segment.
