When shareholders reject a $9 billion buyout, they're not just saying no to a price. They're placing a bet that the market is mispricing the future. Core Scientific just did that. The Nasdaq-listed Bitcoin miner turned AI infrastructure provider voted down a $9 billion acquisition offer, reportedly from a consortium of private equity firms, while simultaneously announcing a strategic partnership with AMD to deploy Instinct GPUs for AI workloads. On the surface, it's a bullish signal: management believes the company is worth more than $9 billion, and the AMD partnership validates their pivot from mining to high-performance computing. But peel back the layers, and this is not a story of technological triumph—it's a high-stakes gamble on infrastructure conversion, supply chain dependencies, and the ability to execute a transformation that has never been done at scale. The market cheered initially, sending CORZ up 12%. Then the questions started. What exactly is Core Scientific selling? How much of the AMD partnership is real? And what happens if the pivot fails?
Let me be clear: I've spent the last decade analyzing crypto infrastructure, from the ASIC farms of 2017 to the GPU clusters of today. I've audited the balance sheets of six publicly traded miners and watched three of them file for Chapter 11. Core Scientific is no exception. It emerged from bankruptcy in early 2024 after a brutal crypto winter, burdened by legacy debt but gifted with a unique asset: a portfolio of long-term, low-cost power purchase agreements (PPAs) at its mining sites across the United States. Those PPAs are the real prize. In a world where AI compute demand is skyrocketing and energy costs are volatile, having locked-in, sub-5-cent-per-kilowatt-hour power is a strategic moat that traditional data center operators envy. But converting a Bitcoin mine into an AI data center is not just a matter of swapping ASICs for GPUs. It requires liquid cooling, high-density racking, low-latency networking (InfiniBand or RoCE), and a software stack that can orchestrate thousands of GPUs for training and inference. The AMD partnership is the linchpin of this transformation, but it's a linchpin made of untested metal.
The technical reality of the AMD-Core Scientific marriage
First, the good news. AMD's Instinct MI300X and MI350 series GPUs are credible alternatives to Nvidia's H100 and B200 for certain AI workloads, particularly inference and fine-tuning of large language models. AMD has made significant strides in its ROCm software ecosystem, closing the gap with CUDA, but the gap is not closed. For a company like Core Scientific, which is targeting AI hyperscalers and enterprise customers, the software compatibility question is existential. When a client brings a PyTorch model to a Core Scientific data center, will it run seamlessly on AMD hardware? The answer is: sometimes yes, sometimes no. Many popular AI frameworks, such as TensorFlow with mixed precision training, still have suboptimal performance on ROCm. The engineering teams at Core Scientific will need to invest heavily in custom kernels and optimization, which is not a one-time cost—it's a continuous liability.
I've seen this movie before. In 2018, a major mining farm I advised tried to pivot to GPU-based rendering for animation studios. They bought AMD GPUs because of the favorable pricing, but the software stack was a nightmare. The rendering software had bugs on AMD hardware, the drivers crashed, and the clients walked. The farm ended up selling the GPUs at a loss and going back to ASICs. The lesson: hardware is only as good as the software that runs on it. Smoke signals, not foundations. The AMD partnership announcement is a smoke signal of intent, but the foundation is yet to be built. Core Scientific has not disclosed any technical benchmarks, no power efficiency numbers, no latency measurements for the AMD clusters. That's a red flag. When a company announces a partnership of this magnitude without a single technical milestone, it's usually because the technical work hasn't started yet.
The capital structure elephant in the room
Core Scientific is a publicly traded company with no native token, so the tokenomics framework doesn't apply directly. But the capital structure dynamics are even more revealing. The company has a history of dilutive financing. In 2023, during its bankruptcy restructuring, it issued convertible notes and warrants that could convert into common shares. The exact terms are not public, but analysts estimate that full dilution could increase the share count by 30-40%. That means the $9 billion valuation that shareholders rejected is not a firm floor—it's a moving target. If the company needs to raise additional capital for the AMD infrastructure buildout, it could issue more shares, diluting existing holders. The AMD partnership likely requires significant upfront investment: ordering GPUs, retrofitting facilities, hiring engineers. AMD is not giving away the chips for free. Core Scientific will need to pay billions of dollars in capital expenditure over the next two years. Where will that money come from? Debt? Equity? Dilution? The company hasn't answered that question.
Let me put this in perspective. In 2021, during the mining boom, Core Scientific was spending over $100 million per quarter on ASIC purchases. Now, they need to spend even more on GPUs, networking, and cooling. The average cost of a single AMD MI300X GPU is around $15,000. A typical AI cluster for a hyperscaler requires 10,000 GPUs. That's $150 million for just one cluster. Multiply that by multiple sites, and you're looking at billions. High APY is just delayed pain. In the crypto world, high yields often mask unsustainably high risk. Here, the high expectations of the AMD pivot mask the capital pain that's coming. The shareholders who rejected the $9 billion buyout are implicitly betting that Core Scientific can generate more than $9 billion in equity value through the AI pivot. But that requires massive capital deployment, which will dilute their ownership. The math doesn't work unless the revenue from AI hosting is substantially higher than mining revenue.
The macro context: Why this matters beyond Core Scientific
This is not just a company story. It's a macro story about the intersection of crypto infrastructure and AI compute demand. The traditional narrative is that Bitcoin miners have a unique advantage: they own power capacity that can be repurposed for AI. But that narrative is being tested. The global liquidity environment is shifting. The Federal Reserve's rate cuts in late 2024 and early 2025 have lowered the cost of borrowing, but the market is now pricing in a potential resurgence of inflation. If long-term rates rise, the cost of debt for capital-intensive projects like AI data centers will increase, squeezing margins. Core Scientific's PPAs are fixed, but their debt financing is not. The company's ability to service its debt obligations depends on generating consistent cash flow from AI hosting. If the AI demand softens—or if the AMD hardware doesn't deliver the promised performance—the cash flow could fall short, triggering a liquidity crisis.
I've seen this pattern before in the crypto mining industry. In 2022, many miners overleveraged themselves to buy ASICs, expecting Bitcoin to stay above $50,000. When it dropped to $16,000, they couldn't pay their loans. Core Scientific itself filed for Chapter 11 in December 2022. The lesson is that capital-intensive infrastructure companies are vulnerable to macro shocks that affect both their revenue and their financing costs. The AI pivot is supposed to diversify revenue, but it also introduces new risks: technology risk, execution risk, and market risk. Systemic risk doesn't care about your partnership agreements. The AMD partnership is a partnership, not a guarantee. If AMD's GPU supply chain faces disruptions—due to TSMC capacity constraints, packaging issues, or geopolitical tensions—Core Scientific's deployment schedule will slip. Their clients will go elsewhere. The company will be left with empty racks and stranded power contracts.
The contrarian angle: What if the decoupling thesis is wrong?
The prevailing narrative in crypto is that Bitcoin miners are reinventing themselves as AI infrastructure providers, and that this will decouple their stock performance from Bitcoin's price. Core Scientific is the poster child for this thesis. But I'm skeptical. The decoupling thesis assumes that the core competency of mining—managing energy-intensive hardware—is directly transferable to AI compute. It's not. Mining is a linear, single-task operation: run ASICs, solve SHA-256 hashes, earn Bitcoin. AI compute is a multi-dimensional, software-intensive operation: schedule jobs, manage memory, optimize interconnects, handle failures, ensure security. The skills required are vastly different. Core Scientific will need to hire hundreds of engineers with expertise in distributed computing, GPU programming, and networking. That's a talent pool that is already in high demand from Amazon, Google, Microsoft, and Nvidia. The company is competing for the same engineers, but it can't offer the same salaries or prestige.
Furthermore, the AMD partnership is a double-edged sword. AMD is a strong competitor, but it's not Nvidia. The enterprise AI market is dominated by Nvidia's CUDA ecosystem. Most AI startups and hyperscalers have built their workflows around CUDA. Switching to AMD is not a trivial decision. It requires retraining, compatibility testing, and performance trade-offs. Core Scientific is essentially betting that the market will embrace AMD's hardware at scale, which is a bet that has failed multiple times in the past. In 2023, AMD's market share in the data center GPU market was around 5%, compared to Nvidia's 90%. Even with the MI300X, AMD's share is expected to grow to only 10-15% by 2026. That's a niche. Core Scientific is tying its future to a niche supplier. Thesis broken. Capital preserved. If the thesis doesn't hold, the capital deployed in this pivot will be lost. The shareholders who rejected the $9 billion buyout are essentially saying they believe the company can create more value than that. But the risks are substantial, and the margin for error is thin.
The experience signal: What I've learned from auditing infrastructure plays
In 2017, when I was auditing the whitepapers of Layer-1 projects, I saw a pattern: projects that prioritized hype over technical substance were the first to fail. The same pattern applies to infrastructure companies today. Core Scientific's announcement lacks substance. It's a press release, not a technical report. There are no architecture diagrams, no performance benchmarks, no customer commitments. The only concrete data point is the rejection of the acquisition offer, which tells us more about the shareholders' expectations than about the company's prospects. I've seen this dynamic before in the 2020 DeFi yield trap: protocols that promised high returns without showing the underlying mechanics were the ones that collapsed. Here, the promise is high returns from AI hosting, but the mechanics are opaque.
I recall a specific incident from 2022 when I was analyzing a mining company that claimed to be pivoting to AI. They had a partnership with a major GPU manufacturer, similar to Core Scientific's AMD deal. They announced it with great fanfare, and the stock rallied. But six months later, they had deployed only 10% of the planned capacity. The GPUs were delayed, the software stack was incompatible, and the clients backed out. The stock crashed 80%. The company was ultimately acquired at a fraction of its peak valuation. The lesson is that infrastructure pivots are hard, slow, and expensive. The market often prices in success before it's achieved, creating a valuation bubble that bursts when reality hits. Core Scientific is at risk of the same pattern.
The forward-looking judgment: What to watch in the next 12 months
For Core Scientific, the next 12 months are critical. The company needs to deliver on three key milestones: first, they must deploy at least 100 MW of AI-ready data center capacity by the end of 2025. Second, they need to sign at least two major AI hosting contracts with enterprise clients that are not related to CoreWeave. Third, they must demonstrate that their AMD clusters can achieve performance parity with Nvidia-based clusters for popular AI workloads like Llama 3 and GPT training. If they hit these milestones, the stock will likely outperform, and the rejection of the $9 billion offer will be vindicated. But if they miss any of these milestones, the market will revise its expectations downward, and the stock could fall below the implied acquisition price.
The bottom line: Core Scientific's story is a case study in the convergence of crypto infrastructure and AI compute, but it's too early to declare victory. The AMD partnership is a step in the right direction, but it's a step, not a leap. The shareholders who rejected the acquisition are betting on execution, but they're also betting against the odds. Based on my experience auditing infrastructure projects, I would say the probability of a successful full-scale pivot within the next two years is around 30%. The remaining 70% includes scenarios of partial failure, dilution, or acquisition at a lower price. The thesis is not broken yet, but it's fragile. Capital preserved—for now. The real test will come when the first AMD cluster goes live and the clients start benchmarking. Until then, this is all smoke signals, not foundations.
As a macro watcher, I see this as a microcosm of a larger trend: the desperation of Bitcoin miners to find new revenue streams after the 2024 halving. The block reward halved from 6.25 BTC to 3.125 BTC, squeezing margins. AI hosting is the white knight, but it's a knight with a rusty sword. The industry is rushing to convert, but the conversion is capital-intensive, technically complex, and fraught with execution risk. Core Scientific is the bellwether. If it succeeds, other miners will follow. If it fails, the narrative will shift. Either way, it's a fascinating experiment in infrastructure adaptation. I'll be watching the power meters and the ROCm compatibility logs. That's where the truth lies.
Final thought: The market has priced in a successful pivot. The technical reality is still unproven. The next 12 months will separate the signal from the noise. And as I always say, high expectations are just delayed pain. The pain may come, or it may not. But the risk is real, and it's not reflected in the current price.