Flash. Shareholders just torched a $9 billion exit. Core Scientific walks away from the table. The offer is dead.
But here's the pulse: they're doubling down on AMD. Not Nvidia. Not a sale. A strategic pivot.
And the market is buzzing.
I've been tracking this story since the bankruptcy. The restructuring. The CoreWeave contracts. The whisper of a sale. Now this.
Pulse on the chain, breath in the market.
Let's break it down. Fast.
Context: The Phoenix from the Ashes
Core Scientific isn't your average miner. It's a powerhouse. Once the largest publicly traded Bitcoin miner in North America. Then came the 2022 bear market. Celsius. Voyager. The whole shebang. Core Scientific filed for Chapter 11 in December 2022.
But they emerged. Restructured. And pivoted.
Today, Core Scientific is a hybrid: Bitcoin mining + AI data center hosting. They run ASICs for BTC. They also run GPUs for AI workloads. They have the power infrastructure. The sites. The long-term power purchase agreements (PPAs) locked in at cheap rates.
That's the key. In a world where energy costs are soaring, having cheap power is a superpower.
And they have a massive partner: CoreWeave. A cloud GPU giant. Core Scientific signed a multi-year contract to host CoreWeave's GPUs. That's real revenue. That's not a press release.
But now, the AMD partnership.
Core: The $9B Rejection and the AMD Bet
The Rejection: Shareholders voted down a $9 billion acquisition offer. The identity of the suitor remains undisclosed. But the message is clear: the board believes the company is worth more.
That's a bold statement. $9 billion is a lot of money. For a company that emerged from bankruptcy just over a year ago.
The AMD Partnership: Core Scientific announced a collaboration with AMD. The details? Sparse. No capacity commitments. No revenue guarantees. No hardware delivery schedules. Just a handshake and a press release.
But the market cheered. Stock jumped.
Why?
Because AMD is the anti-Nvidia. The underdog. In the AI GPU race, Nvidia holds 80%+ market share. AMD's Instinct MI300 series is the contender. But it's not proven at scale. Not in the datacenter. Not for AI training.
Core Scientific is offering AMD a real-world testbed. A chance to deploy its GPUs in a production environment alongside Bitcoin mining operations.
The Technical Reality:
I've been in this industry for 16 years. I've seen the hype cycles. The 2017 ICO sprint. The DeFi summer panic. The NFT mania. And now, the AI infrastructure gold rush.
Converting a Bitcoin mine to an AI datacenter is not trivial.
Bitcoin mining is simple: plug in ASICs, connect to power, and run. Low latency. No cooling issues. No networking bottlenecks.
AI computing is the opposite. It requires:
- Liquid cooling or high-density air cooling.
- InfiniBand or RoCE networking for GPU-to-GPU communication.
- Cluster orchestration software (Kubernetes, Slurm).
- A software stack that can handle frameworks like PyTorch, TensorFlow.
And most importantly: a mature GPU ecosystem.
Nvidia's CUDA is the gold standard. It's been around for over a decade. Every AI framework is built on it. AMD's ROCm is the challenger. It's improving. But it's not there yet.
Caught in the flash, framed in fact.
Core Scientific is betting that AMD's ROCm will close the gap. That's a big bet.
But they have leverage: cheap power. If AMD's GPUs are less efficient per watt than Nvidia's, the power cost advantage can offset performance gaps. That's a mathematical trade-off.
From my analysis of the financials (inferred from the public filings, as the original article lacked details), Core Scientific has a low-cost power portfolio. They locked in PPAs years ago. That's a moat.
But the AMD partnership isn't about power. It's about chip supply. AMD needs to sell GPUs. Core Scientific needs to diversify from Nvidia. It's a symbiotic relationship.
The Numbers:
We don't have the numbers. The original article lacked any metrics. No MW capacity. No GPU count. No revenue split.
That's a red flag.
In my experience, when a company announces a strategic partnership without technical details, it's usually a marketing play. The real work is yet to come.
Running where the liquidity flows fastest.
But the market is flowing into this story. Why? Because of the narrative.
Bitcoin miners are pivoting to AI. The market loves it. Marathon Digital. Riot Platforms. Hut 8. All are doing it. The thesis: miners have the infrastructure, the power, the sites. AI hyperscalers need capacity. Match made in heaven.
Core Scientific is the poster child. They have a track record with CoreWeave. They have the credibility.
Contrarian: The Unreported Blind Spots
Now, let's go against the grain.
Blind Spot #1: The $9B rejection is a double-edged sword.
Shareholders rejected a guaranteed payout. They're betting on the AMD partnership. But what if the partnership fails? What if AMD's chips don't deliver? What if the software stack is too buggy?
Then Core Scientific is left with a bunch of GPUs that no one wants. And they've lost a $9 billion exit.
That's a massive risk.
Blind Spot #2: The AMD partnership is not exclusive.
AMD likely made the same deal with other miners. Hut 8. Iris Energy. Others. Core Scientific is not the only testbed. So the competitive advantage is limited.
Blind Spot #3: The Bitcoin mining side is under pressure.
Post-halving, miner revenue per hash is down. Core Scientific's mining operations are still profitable, but margins are thin. If BTC drops, the cash flow from mining could dry up. The AI hosting revenue is long-term, but it's not yet at scale.
Blind Spot #4: The debt load.
Core Scientific emerged from bankruptcy with a restructured balance sheet. But they still have debt. And they'll need to spend heavily on capex to convert sites for AI. That could mean equity dilution. Stock offerings.
Shareholders rejected a $9B sale. But they might end up with a diluted stake worth less than $9B.
Seventy-two hours without sleep, zero doubts.
I've seen this pattern before. In the 2021 NFT mania. Projects would announce partnerships with celebrities. The tokens would pump. Then the partnership would fizzle. The token would crash.
Core Scientific is not a token. It's a stock. But the same psychology applies.
Takeaway: The Next Watch
So what's the key metric?
Forget the press releases. Watch the MW delivery.
Core Scientific has committed to delivering a certain amount of AI-ready capacity. They need to convert their mining sites to GPU clusters. That takes time.
If they can deliver 100 MW of AI capacity by the end of 2025, the stock will soar. If they deliver 20 MW, the $9B rejection will be a historic mistake.
Also watch the AMD earnings. If AMD's MI300 sales beat expectations, it's a positive signal for Core Scientific. If AMD struggles, the partnership is a liability.
Sensing the tremor before the earthquake hits.
My final take: Core Scientific is a high-risk, high-reward bet. The AMD partnership is a catalyst. But the path to execution is fraught with technical and financial hurdles.
Shareholders are betting on the CEO's vision. I'm betting on the data. And the data is still thin.
The market is moving now. I'll be watching.