Core Scientific shareholders just rejected a $9 billion buyout. The same day, they announced a partnership with AMD. These two events are not separate. They are the same narrative: a company betting that its power grid is worth more than its current market cap. But the market is ignoring the technical gap between a mining rig and an AI GPU cluster.
The market is euphoric. The stock jumped. Analysts cheered. But I’ve spent the last decade auditing infrastructure projects. I’ve seen mining companies pivot to AI before. Most fail. The reason is not the story—it’s the engineering. Code does not lie. People do. But here, there is no code—only promises.
Let’s cut through the hype. Core Scientific is a Bitcoin mining company that emerged from Chapter 11 bankruptcy in early 2024. It owns a portfolio of power infrastructure—sites with cheap, long-term power purchase agreements. That’s its real asset. Not the GPUs. Not the mining rigs. The power. The company is now pivoting to AI data center hosting, signing contracts with CoreWeave and others. The AMD partnership is the latest headline.
But a partnership is not a technical validation. It’s a supply chain agreement. AMD needs to deploy its Instinct GPUs to compete with Nvidia. Core Scientific needs a chip supplier. That’s it. No performance benchmarks. No delivered megawatts. No revenue guarantees. The article I analyzed mentioned the partnership as a strategic move, but provided zero technical details. Zero. I’ve seen this pattern before. A company announces a deal with a big name, the stock pumps, and then the real work begins. The real work is hard. And it’s not in the press release.
Context: The Narrative Cycles
We’ve been here before. In 2021, every mining company wanted to pivot to AI. The narrative was that cheap power and data center experience would translate. Most failed. The reason is that Bitcoin mining is a simple, high-volume, low-margin operation. AI data centers are high-complexity, high-performance, and demand 99.999% uptime. The conversion is not trivial. You need liquid cooling, InfiniBand or RoCE networking, high-density racks, and a software stack that supports CUDA or ROCm. AMD’s ROCm is years behind Nvidia’s CUDA. That’s not an opinion. That’s a fact. I’ve run tests on both. The developer ecosystem, the libraries, the tooling—all heavily skewed toward CUDA. Yield is a tax on ignorance. The yield here is the expected return from AI hosting. But it’s a tax on the ignorant who buy the hype without checking the infrastructure.
Core: The Technical Breakdown
Let’s examine the technical challenges that the market is ignoring. First, the data center conversion. A typical Bitcoin mining site runs ASICs that require air cooling and simple power. An AI cluster requires GPUs that generate massive heat. You need liquid cooling or advanced air handling. That requires capital expenditure. The article didn’t disclose how much Core Scientific is spending on retrofitting. Second, the networking. AI training requires high-bandwidth, low-latency interconnects like InfiniBand. Traditional mining sites use simple Ethernet. The network upgrade is a multi-million dollar project. Third, the software stack. AMD’s ROCm is open-source and improving, but it still lacks support for many popular AI frameworks. NVIDIA’s CUDA is the de facto standard. Any customer using AMD GPUs will face compatibility issues. That’s a risk they’re accepting.
Based on my experience in the 2022 bear market, I saw several mining companies try to pivot to AI. They raised money, bought GPUs, and then struggled to find customers. The reason was that the market for AI compute is not infinite. It’s dominated by hyperscalers like AWS, Google, and Azure. Core Scientific’s customers are likely smaller AI startups or enterprises that need on-premise-like hosting. But the competition is fierce. CoreWeave, the company Core Scientific hosts for, is also a competitor. That’s a complex relationship.
The real value of the AMD partnership is not the chips—it’s the engineering validation. AMD needs real-world deployments to prove its GPUs can handle AI workloads. Core Scientific gets access to engineering support. But this is a two-way street. If AMD’s chips underperform, Core Scientific’s reputation suffers. The article didn’t mention any performance guarantees or minimum purchase commitments. That’s a red flag.
Contrarian: The Blind Spots
The market is treating the AMD partnership as a validation of Core Scientific’s AI pivot. But the contrarian view is that the partnership is a distraction. The company’s core asset is its power infrastructure. The AI pivot is a way to leverage that asset, but it’s not a sure thing. The shareholders rejected the $9 billion buyout, implying they think the company is worth more. That sets a high bar. The stock is now priced for perfection. Any execution miss will be punished.
Consider the debt load. Core Scientific emerged from bankruptcy with a restructured balance sheet, but the details are not public. The AI pivot requires significant capital. The company may need to issue equity or take on debt. That dilutes existing shareholders. The article didn’t mention the capital structure.
The narrative that AMD is the savior is flawed. AMD is not winning the AI chip war. Nvidia holds 80%+ of the data center GPU market. AMD’s Instinct MI300 series is competitive in raw performance, but the software ecosystem is still catching up. Core Scientific is essentially betting on AMD’s success. If AMD fails to gain market share, Core Scientific will be stuck with underutilized GPUs.
Check the supply schedule. Always. But here, the supply is not tokens—it’s megawatts of compute. And they are not yet delivered. The article did not disclose any committed capacity from the AMD partnership. No MW, no GPUs, no timeline. That’s a fundamental lack of information.
Takeaway: The Next Proof Point
The next catalyst for Core Scientific will be the next earnings report. The market will look for AI revenue growth. But the real test is in the operational metrics: how many megawatts are under management? What is the utilization rate? What is the average contract length? Without these numbers, the AMD partnership is just a story. The stock is trading on sentiment, not fundamentals.
I’ve seen this before. The narrative is set. The market is buying the dream. But the dream will crash on the rocks of engineering reality. When the next earnings report shows no material AI revenue, will the narrative hold? Or will the market wake up to the fact that converting a mining site to an AI data center is harder than a press release suggests?
The bottom line: Core Scientific has a valuable asset—cheap power. But that asset is not a moat. It’s a commodity. The AI pivot is a bet on execution. The AMD partnership is a bet on AMD’s ecosystem. Both are risky. The market is ignoring the risk because the bull market is euphoric. But euphoria is the exit liquidity. I’m not buying the narrative. I’m watching the data.
Yield is a tax on ignorance. The yield from AI hosting is not guaranteed. It’s a tax on the ignorant who buy the story without checking the infrastructure. The infrastructure is not ready. The partnership is not a technical breakthrough. It’s a supply chain agreement. And the market is paying a premium for a press release.
I’ll wait for the proof. Not the promise. The proof is in the code, the power, the utilization. Until then, this is just another narrative in a bull market. And narratives have a shelf life.
