When shareholders of Core Scientific voted down a $9 billion acquisition, they didn’t just reject a buyout. They made a statement: the sum of our parts is worth more than the whole you’re offering. But in a market that rewards short-term liquidity over long-term resilience, that statement is a fragile one.
Hook: The numbers are stark. Over the past week, as the news broke that Core Scientific’s shareholders had blocked the sale to a consortium led by a private equity firm, the stock (CORZ) saw a modest uptick—but not the euphoric rally one might expect from a company that just rejected a massive premium. The market is asking: what are they building that’s worth more than $9 billion? The answer isn’t in the press release. It’s in the grind of converting Bitcoin mining sheds into high-density GPU clusters.
Context: Core Scientific is no stranger to chaos. The company emerged from Chapter 11 bankruptcy in 2023, a phoenix from the ashes of the crypto winter. Their core business: operating massive Bitcoin mining facilities in North America, with contracts that lock in low-cost power for years. But the post-ETF landscape has shifted. Institutional money demands not just hash power, but compute power for AI. So Core Scientific pivoted. They signed a multi-year hosting deal with CoreWeave, a GPU cloud provider, and now—this week—they announced a partnership with AMD to deploy Instinct GPUs for AI workloads.
But here’s the catch: the AMD partnership is a strategic alliance, not a revenue contract. No firm commitments on volume, no pricing terms, no delivery milestones. It’s a handshake, not a covenant. And the shareholders just said they trust that handshake more than a $9 billion check.
Core: Let’s talk about the numbers that matter. A $9 billion valuation implies a multiple of roughly 10x on Core Scientific’s projected 2025 EBITDA. That’s a premium for a company transitioning from mining to AI hosting. But the real value isn’t in the GPU chips themselves—it’s in the power infrastructure. Mining sites are designed for 24/7 energy consumption, with access to cheap, stranded power from hydro, nuclear, or natural gas plants. AI data centers need the same thing: constant, low-cost electricity. The difference is that AI workloads require liquid cooling, high-speed networking (InfiniBand or RoCE), and GPU-compatible software stacks.
Based on my audit experience of infrastructure projects, I’ve seen this play before. The promise of repurposing mining sites is technically sound, but the execution is brutal. You need to retrofit existing facilities with cooling systems that can handle 50kW per rack, install fiber optic backbones, and hire a team of network engineers who understand cluster scheduling. Core Scientific has the land and the power. The question is whether they can scale the engineering.
The AMD partnership is a double-edged sword. On one hand, AMD’s MI300 series GPUs offer competitive performance for AI inference, and their ROCm software stack is improving. On the other hand, Nvidia’s CUDA ecosystem is still the gold standard. A 2024 study by MLCommons showed that AMD GPUs required 20% more engineering effort to achieve comparable training throughput on large language models. That’s not a dealbreaker, but it’s a friction point. Core Scientific will need to invest in software optimization to make AMD chips work seamlessly for their AI customers.
Contrarian: The market narrative is bullish: shareholders reject lowball offer, company partners with AMD, future is bright. But let’s be contrarian for a moment. Rejecting a $9 billion offer is a vote of confidence, but it’s also a bet that the company can deliver on its AI pivot without diluting existing shareholders. Remember, Core Scientific still has debt from the restructuring. They’ll need capital for the retrofitting—likely through equity offerings or convertible bonds. That could dilute the very shares that just rejected the sale.
Moreover, the AMD partnership is not a revenue guarantee. It’s a supply agreement. Core Scientific buys AMD chips, sells compute to clients. If the AI market softens—and we’ve seen signs of overcapacity in cloud GPU providers—Core Scientific could be left with expensive hardware and no customers. The same risk applies to their CoreWeave contract: it’s a multi-year deal, but if CoreWeave itself struggles to fill its own capacity, the revenue stream could falter.
Takeaway: Tech changes. Values remain. What Core Scientific’s shareholders are saying is that they value the long-term vision of a decentralized infrastructure provider over a nine-figure exit. They’re betting that the convergence of Bitcoin mining and AI hosting will create a new asset class: compute-as-a-service powered by stranded energy. But the path from here to there is paved with technical debt, capital requirements, and market cycles. Bulls react. Bears reflect. We build. The question is whether Core Scientific can build fast enough to justify the trust its shareholders just placed in it.
In the end, this is a story about covenant over code. The covenant is the trust between shareholders and management. The code is the hardware and software that makes the vision real. Both must hold. If they do, Core Scientific becomes a template for how mining companies can evolve into the backbone of decentralized AI. If they don’t, the $9 billion will look like a missed opportunity. The next 18 months will tell us which narrative wins.
Verify the code, trust the community. But for a public company, the community is the shareholders. And they just made a very loud statement.