The vote was decisive. Core Scientific shareholders rejected a $9 billion acquisition offer. The AMD partnership announcement that followed is not a validation of that decision. It is a test — of whether the company can transform a distressed Bitcoin miner into a viable AI infrastructure provider. Successful execution creates a new floor for the stock. Failure means the $9B was the peak.
Context: From Bankruptcy to Pivot
Core Scientific emerged from Chapter 11 in early 2024, shedding nearly $400 million in debt. The company’s core asset: massive, low-cost power contracts secured during the 2021 mining boom. These contracts, locked in at sub-$0.04/kWh, are the envy of the industry. But the post-halving environment compresses mining margins. Bitcoin’s hashrate keeps climbing, and the block reward is halved. The arithmetic is brutal: more competition for less revenue.
The natural escape route is diversification. Hosting AI workloads turns power into compute, not just hash. Core Scientific signed a multi-year hosting deal with CoreWeave in 2024, committing to deliver 200 MW of GPU-ready capacity. The AMD partnership, announced just days after the shareholder vote, extends that strategy. AMD will supply Instinct GPUs to Core Scientific’s data centers. The press release promised “joint engineering optimization” and “shared go-to-market.” No dollar figures. No minimum purchase commitments. No timeline for capacity delivery.
The shareholder vote adds a layer of tension. The $9 billion bid was not disclosed in detail, but likely came from a private equity consortium or a larger tech player. The rejection sends a clear signal: the board believes the AI pivot can unlock more than $9 billion in long-term equity value. The market now watches whether that belief is rational.
Core Analysis: The AMD Partnership Through a Structural Lens
I have spent the last decade analyzing capital-intensive infrastructure plays — from ICO tokenomics to DeFi yield farms to AI data centers. The common thread: execution risk is always understated in press releases. The AMD partnership, as currently structured, is a loose handshake, not a binding commitment.
Technical Viability: Mining to AI Conversion
Converting a Bitcoin mining facility to host AI GPUs is not a simple retrofit. Mining rigs are ASIC-based, air-cooled, and low-latency-tolerant. AI clusters require liquid cooling, high-density power distribution, and InfiniBand or RoCE networking with sub-microsecond latency. The facility must support 30-40 kW per rack, not the 5-10 kW typical for miners. Core Scientific claims to have solved this with a modular design, but no independent audit of their conversion capability exists. The press release mentions “joint engineering optimization” with AMD, but no specifics on which sites will be converted, at what cost, or by when.
AMD’s Instinct GPUs are the second piece. The MI300X series offers competitive raw compute, but the software ecosystem — ROCm — lags Nvidia’s CUDA in maturity. Many AI workloads, especially large language model training, are optimized for CUDA. Porting them to ROCm requires engineering effort. Core Scientific’s partnership with AMD may include joint optimization of these workloads, but that is speculative. The risk is that even if the hardware is installed, the utilization rate will be low due to software incompatibility. If utilization drops below 60%, the entire economic model collapses.
Economic Sustainability: The Power Edge
Core Scientific’s true competitive advantage is its power cost. At $0.04/kWh, they are 50-70% cheaper than a typical hyperscale data center. This is the core of the thesis: use cheap power to offer AI compute at a discount. The math works if the GPUs are fully utilized. But AI compute is not a commodity market. The price of compute is set by the hyperscalers — AWS, Azure, GCP — who have massive scale and vertical integration. Core Scientific must compete on price, but also on reliability, network speeds, and customer support. The partnership with AMD helps with the supply side, but does not guarantee demand.
The Shareholder Vote as a Value Anchor
The $9 billion rejection sets a floor for the company’s intrinsic value. In a typical M&A scenario, the board’s fiduciary duty is to accept the highest bid unless they can demonstrate that the standalone plan creates more value. The AMD partnership is the central piece of that plan. But the plan must be quantified. The market needs to see: How much capacity will be delivered? What is the expected revenue per MW? What is the margin profile? The press release provides none of this. The shareholder vote is a bet on the management team’s ability to execute, not on a proven track record.
Contrarian Angle: The Decoupling Thesis
There is a prevailing narrative that Bitcoin miners are uniquely positioned to capture the AI boom because of their power assets. This narrative is romantic but fragile. History shows that infrastructure pivots often fail because the skill sets are different. Mining is about managing ASIC fleets, optimizing hashrate, and hedging P&L. AI hosting is about managing GPU clusters, optimizing network latency, and maintaining SLAs with enterprise customers. The operational DNA is different. Core Scientific hired former CoreWeave executives to bridge that gap, but cultural integration is slow.
Another blind spot: the AMD partnership might be a distraction. AMD’s market share in AI GPUs is around 10% and growing, but the company lacks the end-to-end integration that Nvidia offers with its CUDA software stack, networking (Mellanox), and DPUs. Core Scientific could face a support gap if AMD’s chip supply or performance falls short. The partnership also locks Core Scientific into AMD’s product roadmap, which carries the risk of a generation gap if Nvidia’s next-gen Blackwell or Rubin chips leapfrog.
Finally, the shareholder vote itself may be a signal of overconfidence. The $9 billion offer was likely made by a sophisticated buyer who did due diligence. The rejection suggests that the board believes the standalone plan is worth more. But what if the plan fails? The opportunity cost of not selling is the entire $9 billion. In a bear market, capital preservation is paramount. The board is betting on growth, but the bear market rewards survival, not moonshots.
Takeaway: Cycle Positioning and the Real Test
The bear market is still in its middle innings. The “easy money” from Bitcoin’s recovery has been made. Now, the survivors are those who can generate sustainable cash flow without relying on token appreciation or equity dilution. Core Scientific’s pivot to AI is a rational response to the halving pressure, but it is not a guaranteed path to value creation. The AMD partnership is a step in the right direction, but it is a step on a long, capital-intensive road.
Volatility is the fee for entry. The stock has already surged 20% on the AMD news. That is the market pricing in the option value of success. But option value decays if the milestones are not met. The next six months are critical: Core Scientific must announce concrete capacity delivery numbers, have its first AI workload running on AMD hardware, and show that the margin profile is attractive. If they do, the $9 billion rejection will be remembered as a prescient call. If they don’t, the shareholders who voted against the sale will be the ones that paid the highest price.
Liquidity evaporates faster than hype. The AMD partnership is hype today. It becomes liquidity only when the GPUs are powering actual inference jobs for paying customers. Until then, it is a press release with no code, no audit, and no guarantee. The structural skepticism engine is running. The burden of proof is on Core Scientific.
