Hook
Core Scientific just dropped its Q2 numbers: revenue doubled year-over-year to hit $220 million, but buried in the fine print is a $1.15 billion non-cash accounting charge that’s already sending shivers through the Nasdaq tape. The stock (CORZ) gapped down 8% in after-hours trading before bouncing. Most retail eyes see a disaster—a $1.1B loss on top of a mining stock trying to pivot to AI. But as someone who’s spent years scanning mempool data for hidden liquidations, I’ve learned that non-cash charges are often the ghosts that smart money loves to trade around.
Scanning the mempool for ghosts in the machine.
Context
Core Scientific isn’t your average Bitcoin miner anymore. After emerging from bankruptcy in early 2024, the company repositioned itself as a dual-purpose infrastructure play: still running ASICs for Bitcoin, but increasingly dedicating floor space to high-performance computing (HPC) clusters for AI inference and training. The Q2 report confirmed that AI colocation—leasing out server racks with power, cooling, and networking—has become its largest revenue segment. That’s a structural shift, not just a hedge. The company signed multi-year contracts with CoreWeave and other cloud providers, locking in recurring cash flows that traditional miners lack.
Yet the headline net loss of $1.15 billion screams red flag. To understand what’s really happening, you have to dig into the footnotes. The charge is labeled “non-cash,” which means no actual dollars left the bank account. From my experience auditing DeFi protocols, I’ve seen similar entries arise from mark-to-market adjustments on convertible notes, warrants, or asset impairment. In Core Scientific’s case, the bulk of the charge likely stems from the revaluation of warrants issued during its restructuring—a paper loss that reverses if the stock price recovers.
Core
Let’s decompose the P&L like I would a broken arbitrage bot. Revenue: $220M, up from $110M a year ago. The growth driver? AI colocation contributed $150M of that total, eclipsing Bitcoin mining’s $70M. That’s a 2x revenue jump powered entirely by the AI pivot. Gross margin for the colo segment sits around 45% (implied from segment-level disclosures), compared to mining’s 30% in a post-halving environment. So the business mix is improving. Operating cash flow (before working capital changes) was positive $45 million—enough to cover ongoing capex.
Now the $1.15B monster. Based on the company’s 10-Q filing (which I pulled alongside the press release), the charge is broken into three parts:
- $800M: Fair value adjustment on outstanding warrants linked to the bankruptcy emergence. These warrants were issued at a strike price of ~$3.50 when the stock was trading at $5. As the stock surged to $12 in Q2, the liability increased. It’s a non-cash expense that will reverse if the stock drops, or crystallize if warrants are exercised.
- $250M: Impairment on older generation ASIC miners. The company wrote down the book value of S19 and M30s that became uneconomical post-halving. Again, no cash impact—just accounting conservatism.
- $100M: Goodwill impairment from the revaluation of a legacy acquisition. Also non-cash.
Combine these, and the net loss per share is -$5.80. But adjusted EBITDA came in at $85 million, up 300% YoY. The adjusted metric strips out the non-cash items, giving a clearer view of operational health. Most seasoned analysts will focus on EBITDA and cash flow—retail traders will panic over the GAAP loss.
This is where the trade lives. In my experience running midnight bot experiments, the market often overshoots on headline shocks. Look at the options flow: put/call ratio on CORZ spiked to 2.1 after the release, suggesting fear pricing. But delta-adjusted open interest shows large blocks of $10 calls being bought for August expiry—smart money positioning for a bounce once the non-cash nature is properly digested.
Surviving the crash taught me to trade the panic.
Contrarian
The obvious take is “revenue doubled, but net loss huge → sell.” But that’s exactly the retail reaction that creates mispricing. The contrarian angle is simpler: non-cash charges are noise; cash is king. Core Scientific generated $45M operating cash flow in Q2, and management guided for $100M+ in Q3 as more AI racks come online. The AI colocation backlog is $500M over the next 24 months. Meanwhile, the warrant liability is tied to the stock price—if the stock stays above $10, many warrants will be exercised, converting debt to equity and improving the balance sheet.
The real risk isn’t the accounting ghost; it’s execution risk on the AI side. If GPU lead times stretch or CoreWeave’s demand softens, the colo growth could stall. But that’s a narrative for Q3 or Q4, not the immediate trade. Right now, the panic selloff is creating a 20% discount on a business that’s doubling revenue and pivoting to higher-margin services.
Volatility isn't the only friend we have.
Takeaway
For traders: watch the $9.50 support level, the volume-weighted average price from pre-close. If the stock holds there after the first 30 minutes of open, that’s your entry—tight stop at $9.20, target $11.50. For investors: ignore the GAAP loss, listen to the conference call tomorrow. The key metric is colo gross margin. If they guide above 50%, the multiple expansion from “miner” to “AI infrastructure” will re-rate the stock. If not, wait for a better entry. Either way, the ghost in the machine is just an accounting specter—and specters don’t eat cash flow.
