Shareholders said no to $9 billion. Then AMD stepped in. The spread was real, but the exit was imaginary.
On paper, the math is simple. Core Scientific, a Bitcoin mining operator that crawled out of bankruptcy in early 2024, was offered a clean exit at $9 billion by CoreWeave. The board recommended it. The shareholders rejected it. Now the same company is announcing a partnership with AMD to supply GPUs for AI workloads. The stock jumped. The narrative flipped. But I’ve been in enough trading rooms to know that when the market celebrates a headline without asking for the data, the real risk is already priced in.
Context matters. Core Scientific is not a DeFi protocol or a Layer 2. It’s a physical infrastructure play. It owns land, power contracts, and racks of ASICs for Bitcoin mining. The pivot to AI data center hosting is a natural evolution—mining facilities are already built for high-density power, cooling, and 24/7 uptime. But the transition from ASICs to GPUs is not a software patch. It’s a full rebuild. I’ve audited similar infrastructure projects, and the gap between “announced partnership” and “delivered compute” is wider than most retail investors realize.
Let’s look at what the AMD deal actually contains. The press release mentions a “strategic partnership” to deploy AMD Instinct GPUs for AI workloads. No dollar amounts. No number of GPUs. No timeline for delivery. No mention of which data center sites will be converted. No power capacity commitments. The market treated this as a competitive win against Nvidia, but Nvidia’s CUDA ecosystem has a decade of optimization, while AMD’s ROCm is still catching up. I’ve tested both stacks in a quant context—ROCm works, but the tooling is fragmentary. For a company that needs to deliver enterprise-grade reliability to AI clients, that’s a tangible risk.
Alpha decays faster than the code that finds it. The same principle applies here. The initial surge from the AMD announcement is a temporary arbitrage between narrative and reality. The real value will be determined by execution metrics: how many megawatts of GPU-ready compute go live, what utilization rates they achieve, and whether the contracts are signed with fixed pricing or variable terms. As of today, none of those numbers are public. The market is trading on hope, not data.
I’ve been tracking Core Scientific since their bankruptcy filing in 2022. They emerged with a clean balance sheet but a massive capital expenditure requirement. The AI pivot requires billions in hardware, cooling infrastructure, and networking—InfiniBand, not just Ethernet. The company has already signed a 12-year hosting deal with CoreWeave, which provides a revenue floor, but the AMD deal is separate. It’s not a revenue guarantee; it’s a supply agreement. The difference is critical. If Core Scientific has to buy the GPUs upfront, they’re taking inventory risk. If AMD is providing them on consignment, the risk is lower, but the margin is thinner. The press release doesn’t clarify.
This is where my own experience with infrastructure arbitrage comes in. In 2020, I built a yield farming strategy that looked perfect on paper—140% APR, audited contracts, collateralized loans. But I ignored the operational risk of third-party vaults. When a minor exploit drained $2 million from a similar protocol, I withdrew everything. I saved 60% of my capital because I was watching the log of on-chain activity, not the hype. The same discipline applies here. Core Scientific’s stock is not a bet on AMD’s chips; it’s a bet on the company’s ability to transform a mining facility into a Tier 3 data center without blowing the budget.
Let’s dismantle the bullish case. The arguments for the AMD partnership are: 1. Diversification of GPU supply reduces dependency on Nvidia. 2. Core Scientific’s power contracts are cheap and long-term. 3. The AI infrastructure market is growing at 40% CAGR. 4. Shareholders rejected $9 billion, implying they believe the company is worth more.
Each of these points has a counter. First, GPU supply diversification is only valuable if the chips are in high demand. AMD’s Instinct MI300 series is competitive, but enterprise adoption is slow because software migration is painful. Core Scientific will need to offer engineering support—a cost not factored into the stock price. Second, cheap power is a commodity advantage, but converting that power into compute requires a different cost structure. Mining ASICs are plug-and-play; GPUs require liquid cooling, high-speed networking, and rack-level redundancy. I’ve seen mining companies underestimate this by 30% or more.
Third, the AI infrastructure market is growing, but the supply of compute is also growing. Core Scientific is competing with CoreWeave, Lambda, and even AWS. The differentiation is not just power; it’s latency, reliability, and software stack. AMD’s ecosystem is not as mature as Nvidia’s, which means Core Scientific’s clients will face higher integration costs. That reduces the addressable market. Fourth, the rejection of the $9 billion offer is not a signal of intrinsic value. It’s a signal of hubris. Shareholders are betting that the company can execute better than a private equity exit. But execution risk is high, and the timeline is uncertain.
I trust the log, not the hype. The log of Core Scientific’s operational history shows a company that was pushed into bankruptcy by over-leverage and market volatility. They have a new CEO and a new strategy, but the scars remain. The balance sheet is stronger, but the capital needs are massive. The AMD deal does not change the fact that the company needs to raise billions to scale. Dilution is likely. The stock price today does not reflect the future share count.
Let’s compare to a similar case: the transition of a mining company to AI. In 2023, Hive Blockchain rebranded to Hive Digital Technologies and pivoted to GPU hosting for AI and gaming. They had a head start with existing GPU infrastructure. Core Scientific is starting from ASICs—they need to buy or repurpose entire facilities. The capital expenditure per megawatt is roughly $3-5 million for AI-grade data centers, versus $1-2 million for mining. The revenue per megawatt is higher, but the lead time is longer. Investors are pricing in the revenue but not the time lag.
From a trading perspective, the stock is a volatility play. The AMD announcement created a gamma squeeze on options, but the underlying fundamentals are unchanged. The stock is trading at a premium to its book value based on future AI revenue that hasn’t been booked. The contrarian angle is that the market is mispricing the risk of execution failure. If Core Scientific fails to deliver the first 50 MW of GPU compute by Q2 2025, the stock will reprice downward. The millennials chasing the narrative will be caught in the liquidation.
Liquidity is a mirage during the storm. When the market realizes that the AMD partnership is not a binding contract for revenue, the bid will thin. The real money is in the data: watch for quarterly filings where they disclose “data center capacity under transformation” and “AI compute revenue.” Until those numbers appear, the stock is a speculation.
My own experience with the Terra Luna collapse taught me that data-driven exits are the only way to survive. In May 2022, I held $15,000 in UST. I monitored Dune Analytics hourly, watching the supply mechanics. I sold in stages, losing 40% but saving 60%. The emotional traders held to zero. Core Scientific’s stock is not going to zero, but the path is uncertain. The key metric is not the partnership announcement; it’s the power utilization rate. If the company can deploy 100 MW of AI compute by end of 2025, the stock is worth $15-20 based on projected EBITDA. If not, it’s a $5 stock.
I’ll be watching the log, not the hype. The AMD partnership is a necessary step, but it’s not sufficient. The blind spot is where the money hides. The market is blind to the operational complexity of the conversion. The retail crowd sees “AMD” and thinks “AI.” The smart money is asking: “What’s the power purchase agreement price? What’s the cooling system design? What’s the network topology?” Those questions aren’t answered in the press release.
Takeaway: Core Scientific’s stock is a binary option on execution. The AMD partnership is a catalyst, but the real value is in the engineering. I’m not buying until I see a quarterly report showing at least 50 MW of AI compute in production. Until then, the stock is a trading vehicle, not an investment. The bot didn’t fail; the market changed rules. The same applies here. The market is rewarding the narrative, but the rules of infrastructure require time and capital. Patience is a strategy.
We optimize for edges, not comfort. The edge here is to wait for the data, not the press release. The comfort is in the hype. I’ll take the edge.