Core Scientific shareholders just said no to a $9 billion exit.
That's a bet on AMD chips, a gamble on AI infrastructure, and a signal that the market is still pricing mining infrastructure as a commodity. The vote was a rejection of the easy path. The question now: is the AMD partnership a real diversification play, or just a headline to keep the stock afloat?
I've spent the last decade watching mining companies pivot. Most fail. The ones that succeed don't just swap ASICs for GPUs—they rebuild their entire power architecture. Core Scientific is trying to do that. But the details are thin, and the stakes are high.
Context: The Infrastructure Layer
Core Scientific is a Nasdaq-listed company (CORZ) that started as a bitcoin mining operator. After a bankruptcy restructuring in 2023, it emerged with a plan to repurpose its existing power infrastructure for AI compute hosting. The core thesis: a mining facility is a data center with a different chip. The power, cooling, and security are already there. The switch is mostly about the hardware and the networking stack.
This is not a new idea. Hut 8, Riot, and others have tried similar pivots. The difference is that Core Scientific is now partnering with AMD to bring Instinct GPUs into their facilities. The partnership was announced alongside the shareholder vote that killed the $9 billion acquisition offer from an undisclosed buyer.
According to the company's press release, the AMD collaboration will "accelerate the deployment of high-performance computing (HPC) infrastructure" for AI workloads. But the release offered no specific numbers—no megawatts of capacity, no GPU count, no timeline for first delivery. This is a strategic announcement, not a technical milestone.
Why does that matter? Because in the crypto-infrastructure space, the gap between announcement and delivery is where most projects die. I've seen it firsthand: in 2021, I audited a mining firm that claimed to be building a 100MW AI data center. Two years later, they had 10MW operational and the rest was a PowerPoint deck. Core Scientific is a larger, more credible player, but the same pattern applies.
Core: The Technical Reality of the AMD Pivot
Let's get into the technical weeds. Core Scientific's existing infrastructure is built for ASICs—application-specific integrated circuits that run bitcoin mining algorithms. These chips are power-dense, but they don't require the same networking and cooling as GPUs. ASICs can run in ambient air with simple racks. GPUs for AI training need liquid cooling, high-bandwidth interconnects (InfiniBand or RoCE), and a completely different software stack.
AMD's Instinct GPUs are a competitive alternative to Nvidia's H100 and B200 series, but they come with a catch: the ROCm software ecosystem. ROCm has improved, but it's still not as mature as CUDA. For a client running PyTorch or TensorFlow, the difference is real. Any AI startup that has tried to switch from Nvidia to AMD knows the pain of debugging driver issues and missing libraries.
Core Scientific is not a software company. They are a power and facilities operator. Their value proposition is cheap electricity, not GPU optimization. The AMD partnership is a hardware supply deal, not a software engineering collaboration. That means the burden of making ROCm work falls on the client or on AMD itself. Core Scientific is just the landlord.
This is where the hidden angle emerges. AMD needs testbeds. They need real data centers to deploy their GPUs and prove their performance against Nvidia. Core Scientific offers that: a ready-made power infrastructure at scale, with long-term power purchase agreements that lock in low rates. The partnership is as much about AMD getting a foothold in the AI hosting market as it is about Core Scientific diversifying revenue.
But the numbers don't lie. The article that broke this news—the one I'm analyzing—contained zero data on delivered capacity, utilization rates, or contract terms. The only concrete fact is that shareholders rejected a $9 billion buyout. That sets a valuation floor: the board believes the company is worth more than that. The AMD partnership is the justification for that belief.
Is the AMD partnership worth $9 billion?
Let's look at the math. Core Scientific's market cap before the vote was around $4 billion. The $9 billion offer represented a 125% premium. To justify that premium, the company needs to show that the AI hosting business will generate at least $500 million in annual EBITDA within 3-5 years. That's a high bar. Most AI hosting companies with similar capabilities are trading at 10-15x EBITDA. Core Scientific would need to add $1-2 billion in enterprise value through the AMD deal alone.
Is that possible? Yes, if the AI compute demand explosion continues. But it's not guaranteed. The market is already pricing in a lot of optimism. The stock has rallied since the rejection, but it's still below the implied offer price. That tells me the market is skeptical about the AMD partnership's immediate impact.
Contrarian: The Real Value Is in the Power, Not the Chips
Here's the contrarian take that no one is talking about: the AMD partnership is a distraction. The real asset is the power infrastructure and the long-term power contracts. Bitcoin miners have a hidden advantage: they locked in low electricity rates during the 2020-2021 bull run, often below $0.03/kWh. Those contracts are now gold. AI data centers are desperate for power, and the grid is constrained. Core Scientific's existing facilities are already built and connected.
If the AMD partnership fails to deliver, the company can still sell its power capacity to other AI providers or even back to the grid. The chips are replaceable. The power is not.
This is the same logic that drove the Ethereum merge narrative: proof-of-stake doesn't need energy, but proof-of-work's infrastructure can be repurposed. The same applies here. The mining facilities are the moat, not the GPU choice.
Shareholders who rejected the $9 billion offer are betting on the management team's ability to execute the pivot. But management has a mixed track record. Core Scientific filed for bankruptcy in 2022 after over-leveraging on mining equipment. The current CEO, Adam Sullivan, has been pushing the AI narrative hard, but the company's financials still show a heavy reliance on bitcoin mining revenue.
In the first quarter of 2024, Core Scientific reported $150 million in revenue, with 80% coming from mining. The AI hosting segment contributed just $20 million. The AMD partnership is supposed to flip that ratio, but it will take time and capital. The company will need to raise additional funds—either through equity issuance, debt, or a joint venture. That dilutes existing shareholders.
The mint button was a lever, not a purchase.
This line from my past analysis applies here. The AMD partnership is a lever—a tool to enable growth, but not a guarantee of it. The purchase is the actual deployment of GPUs and the signing of long-term hosting contracts. Until we see those contracts, the partnership is just a press release.
Takeaway: What to Watch Next
For investors and traders, the next few quarters will be critical. I'm watching three things:
- Megawatt deployment: Core Scientific needs to announce how many MW of AI-capable capacity they are building. The industry standard is to report contracted capacity, deployed capacity, and utilization. If they don't start reporting these numbers by Q3 2024, the partnership is likely dead.
- AMD GPU supply: AMD's Instinct MI300X is facing supply constraints. If Core Scientific can't get chips, the pivot stalls. Watch for any delays in AMD's shipment timelines.
- Client announcements: The company needs to announce anchor tenants for the AI compute. A headline like "Core Scientific signs 5-year hosting deal with xAI" would be more valuable than any partnership announcement.
Volatility is just fear wearing a disguise.
The stock will swing wildly on every rumor. The $9 billion rejection sets a high bar, and the market will punish any sign of missed execution. But if the pivot works, Core Scientific could become a dominant player in the AI infrastructure space. The risk is real, but so is the upside.
I've been through these cycles before. In 2017, I watched mining companies pivot to AI and fail. In 2020, I saw the DeFi yield farms collapse. The winners are the ones who focus on the fundamentals: power, contracts, and execution.
Core Scientific has the power. Now they need to prove they can execute.
