The $9 Billion No: Core Scientific’s Rejection and the AMD Pivot

The shareholders of Core Scientific looked at a $9 billion buyout offer and said no. That’s not a small number to walk away from—it’s roughly 30% above the company’s pre-announcement market cap. The acquirer, a consortium of private equity firms, had been circling for months, drawn by the company’s unique positioning: a fleet of Bitcoin mining rigs sitting on cheap power contracts, and a newborn AI data center business that had just landed a multi-year hosting deal with CoreWeave. Yet the vote was decisive. The offer was rejected. And the next day, the company announced a strategic partnership with AMD to supply Instinct GPUs for its AI workloads. The market reaction was muted at first, then confused. Did shareholders just bet that AMD’s GA300 series could outrun Nvidia’s CUDA moat? Or was this a rejection of the offer itself—a signal that the board believes the company is worth more than $9 billion? The answer, as always, lies somewhere in the structural tension between infrastructure and narrative. To understand the gravity of this decision, one must first map the landscape. Core Scientific emerged from Chapter 11 bankruptcy in early 2024, laden with legacy debt but carrying a formidable asset: 1.2 gigawatts of contracted power across seven mining sites in North America. Most of that power was locked in multi-year, fixed-price agreements, often below $0.04/kWh—a rare advantage in an era of rising energy costs. The company had spent 2023 pivoting from pure Bitcoin mining to what it calls “digital infrastructure hosting,” essentially converting a portion of its mining halls into GPU-ready data centers. The first tenant was CoreWeave, a cloud GPU provider that signed a 10-year, $1.2 billion contract in July 2024. That deal validated the thesis: mining power is AI power, if you can cool the chips. But the AMD partnership, announced concurrently with the shareholder vote, is a different beast. It is not a revenue contract; it is a supply agreement. Core Scientific will buy AMD’s Instinct MI300X and MI350 series GPUs, presumably to stock its own AI cloud offering or to lease to new clients. The terms are undisclosed. No minimum purchase volume, no exclusivity, no revenue guarantee. Just a press release and a handshake. From a technical perspective, the AMD partnership is a double-edged sword. The Instinct MI300X is a competent GPU for both training and inference, with 192 GB of HBM3 memory and a theoretical peak of 1.3 PFLOPS in FP8. It competes directly with Nvidia’s H100 and B200. But the ecosystem is where the battle is lost. AMD’s ROCm software stack, while improving, still lags CUDA in library support, debugging tools, and community-written kernels. For a data center operator like Core Scientific, the cost of switching is not just the GPU price—it’s the engineering time needed to port models, the risk of encountering undocumented bugs, and the potential for lower throughput in real-world workloads. I have seen this play out before. In my early days auditing DeFi protocols, I watched teams migrate from Solidity to Rust-based smart contracts, only to find that the developer tooling was immature, and they ended up rewriting half the codebase. The same principle applies here: software ecosystem maturity is a nonlinear multiplier on hardware performance. A 10% performance gap in benchmarks can become a 50% gap in production when the operators don’t have the right libraries. Core Scientific will need to hire a substantial team of AI infrastructure engineers—not just server wranglers—to make AMD work at scale. The company’s existing workforce is skilled in mining operations, which are vastly different: ASIC management is about power efficiency and heat management, while GPU clusters require InfiniBand networking, distributed training frameworks, and fault-tolerant job scheduling. The learning curve is steep. Yet the economic logic of the AMD pivot is undeniable. Bitcoin mining margins have been compressed by the April 2024 halving, which cut block rewards from 6.25 to 3.125 BTC. The hashprice—the expected revenue per terahash per day—has fallen from $0.12 in early 2023 to around $0.05 in early 2025. Miners with high-cost power are already bleeding. Core Scientific, with its cheap power and low debt, can survive, but growth is limited. The AI hosting market, on the other hand, is projected to grow at 40% CAGR through 2028, driven by enterprise adoption of generative AI. The bottleneck is not demand—it is supply of high-density data center space. Traditional data centers are designed for 5-10 kW per rack, but AI racks often require 40-80 kW, demanding liquid cooling and upgraded electrical distribution. Core Scientific’s mining sites, originally built for 100 kW+ ASIC racks, are physically suited for this. They have the power, the cooling (though mostly air-cooled, needing retrofits), and the security. The conversion cost is estimated at $5-10 million per megawatt, a fraction of building a new data center. So the AMD partnership is not just a hedge—it is a bid to capture the AI infrastructure wave before the traditional REITs and cloud providers catch up. But here is the contrarian edge: the market is treating this as a validation event, but it is actually a stress test. The $9 billion offer was rejected, and the stock price has not risen above that implied valuation. The market is saying, “Prove it.” The AMD partnership, absent committed revenue, is a bet on the company’s ability to execute a complex technical transition while competing against giants like CoreWeave, which is itself backed by Nvidia. CoreWeave has exclusive access to H100s and B200s, and its entire stack is optimized for Nvidia. Core Scientific, by partnering with AMD, is essentially choosing to fight with one hand tied behind its back—unless AMD delivers a step change in software maturity. That is possible, but it is not priced in. The shareholders are betting on a thesis that has not been proven: that the infrastructure layer (power + land + cooling) is more valuable than the hardware layer. In theory, yes. In practice, the switching costs for customers are high, and the GPU supply chain is volatile. I recall during the 2020 DeFi summer, many protocols touted their “multi-chain” strategy, but the ones that actually succeeded were those that focused on a single, dominant chain first. Diversification for the sake of hedge is often a sign of weakness. Core Scientific is diversifying its GPU supply, but it may be diluting its focus at a time when execution is paramount. Another layer of complexity lies in the shareholder psychology. The rejection of the $9 billion offer was not unanimous—it was a vote, and the dissenting shareholders likely believed that the company could generate more value independently. But what is the timeline? The company will need to raise capital for the AI conversion, either through debt or equity issuance. If it issues new shares, existing shareholders will be diluted. If it takes on debt, it risks a repeat of the 2022 crisis that led to bankruptcy. The AMD partnership does not solve that; it only increases the capital expenditure requirement. The board’s message is: “We are worth more than $9 billion, but only if we successfully execute a multi-year transformation.” That is a high-risk, high-reward bet. The market may reward patience, but it could also punish overreach. From a macro perspective, this is a case study in the decoupling of crypto-native assets from legacy infrastructure. Bitcoin mining is becoming a commodity business, and the survivors are those that can pivot to AI. But the pivot is not easy. The surface is chaotic. The headlines scream “partnership” and “rejection of lowball offer,” but underneath, the engineering timeline stretches years, the software stack is fragile, and the capital markets are unforgiving. I have seen this pattern before: a company with a strong base (cheap power, existing sites) announces a transformative partnership, the stock jumps, but then the complexities of integration surface, and the narrative fades. The true test will come in 12-18 months, when Core Scientific must report its first AI-driven revenue line and disclose GPU utilization rates. Until then, the $9 billion no remains a puzzle—a defiant stand by shareholders who believe in the story, but a story that has yet to be written. In the end, the question is not whether Core Scientific can survive; it is whether it can thrive in a world where every mining company is trying to do the same thing. Riot Platforms, Marathon Digital, and Hut 8 are all pivoting to AI. The differentiation will come down to site selection, engineering talent, and the ability to lock in long-term, high-margin contracts. Core Scientific has the power, but it is betting on AMD to provide the brains. The surface is chaotic. The structure is uncertain. The next chapter will be defined by execution, not announcements. And as the market waits, the shareholders who said no are holding their breath. s chaotic surface s chaotic surface s chaotic surface