I didn’t expect shareholders to say no to $9 billion.
Not in this market. Not when every AI infrastructure play is trading at 50x forward revenue. But Core Scientific’s boardroom just pulled the ultimate contrarian move. Shareholders rejected a $9B acquisition offer. Walked away from a 9-figure exit. The news hit my terminal at 2:47 PM EST, and my first thought was: Chaos isn’t the mining rig noise; it’s the boardroom silence.
This isn’t a story about a chip deal. It’s a story about a company that survived bankruptcy, emerged from Chapter 11 with scars, and now believes it can build something bigger than any acquirer is willing to pay. The AMD partnership is the headline. The shareholder vote is the subtext. The real story is how a Bitcoin miner turned AI host is betting its entire future on convincing the market that its power infrastructure is worth more than a $9 billion check.
--- ### Context: From Bankruptcy to the AI FOMO Machine
Core Scientific isn’t a new name. It’s one of the largest publicly traded Bitcoin miners in North America. But the company’s history is a cautionary tale written in carbon-heavy debt. In 2022, it filed for Chapter 11 bankruptcy after the crypto winter crushed its mining margins. The restructuring was brutal: creditors took equity, shareholders got diluted, and the company emerged leaner but scarred.
Then came the AI pivot. The same massive power contracts that once powered ASICs could now run NVIDIA H100s and AMD Instinct GPUs. The thesis was simple: Bitcoin mining is a low-margin commodity business; AI inference is a high-margin growth business. Core Scientific started converting its mining sites into GPU-ready data centers, signing multi-year hosting deals with CoreWeave and others. The stock, once trading at distressed levels, rallied.
But the market never fully priced in the next step: a strategic partnership with AMD to deploy its MI300X and future Instinct GPUs. When the news broke alongside the shareholder rejection of the $9B buyout, the narrative shifted. Suddenly, Core Scientific wasn’t just a mining operator anymore. It was a potential AI infrastructure challenger.
--- ### Core: The AMD Partnership – What’s Real and What’s Hype
Let’s cut through the press release. The AMD partnership is announced. Core Scientific will host AMD’s GPUs in its data centers. Full stop. There are no disclosed commitments on megawatts, GPU count, or revenue guarantees. The company hasn’t published a single benchmark, a single workload test, or a single customer deployment for the AMD hardware.
This is a strategic announcement, not a technical milestone.
I’ve covered enough mining-to-AI pivots to know the difference. A real technical milestone looks like: “We’ve deployed 10 MW of H100 capacity, achieving 90% utilization, with a 3-year contract attached.” That’s what CoreWeave does. That’s what Iris Energy does. Core Scientific’s AMD deal has none of that.
But here’s the nuance. AMD needs this deal more than Core Scientific does. AMD’s MI300X is a strong GPU, but its software ecosystem (ROCm) is still catching up to NVIDIA’s CUDA. AMD needs real-world deployment data to prove its chips can handle AI inference at scale. Core Scientific provides that—a live, production-grade testing ground. In exchange, AMD likely provided favorable pricing or engineering support. The deal is a mutual validation play.
From my experience analyzing ASIC miners and GPU racks, the biggest bottleneck isn’t the chip itself. It’s the power delivery, the cooling, the networking. Core Scientific’s existing sites are designed for high-density power consumption. But Bitcoin mining runs on air cooling. AI workloads need liquid cooling. The transition requires costly retrofitting. The company hasn’t disclosed how many sites are liquid-cooled ready.
The real value of the AMD partnership isn’t the GPU; it’s the power contract.
Core Scientific’s core asset is its long-term power purchase agreements (PPAs) locked in at below-market rates. In a world where AI data centers are struggling to find grid capacity, that’s gold. The AMD deal is a way to monetize that power at higher margins than Bitcoin mining. But the margins only work if utilization stays high. If the AMD GPUs sit idle, the economics collapse.
--- ### Contrarian: The $9B Rejection Might Be a Mistake
Here’s the uncomfortable angle. Shareholders rejected the buyout because they believe the AMD partnership will unlock more value than $9 billion. But what if they’re wrong?
The counter-argument: The AI infrastructure market is already crowded. CoreWeave, Lambda, and even traditional cloud providers like AWS are competing for the same GPU deals. Core Scientific’s advantage is its existing power contracts, but those contracts are not exclusive. Competitors can also negotiate PPAs. The AMD partnership is not a moat; it’s a supplier relationship. AMD can easily partner with 10 other hosting providers. The partnership is non-exclusive by default.
Moreover, the timing of the rejection is suspicious. The $9 billion offer came from a private equity consortium. The offer valued Core Scientific at roughly 20x its current EBITDA. That’s rich for a company in transition. But rejecting it means the management is now under immense pressure to deliver organic growth. If the AMD partnership stumbles, or if Bitcoin prices drop again, the stock could crater below the offer price.
I didn’t expect the market to cheer the rejection.
But it did. The stock popped 12% on the news. The narrative is powerful: “A mining company that survived bankruptcy is now an AI play.” The future isn’t written yet. But the track record of mining-to-AI pivots is mixed. Many have announced big partnerships but delivered little. The ones that succeed have clear metrics: MW delivered, utilization rates, contract lengths. Core Scientific hasn’t shared any of those for the AMD deal.
--- ### Takeaway: The Sprint to the Grid
The future isn’t in the GPU, it’s in the grid. Core Scientific’s edge isn’t its chip supplier; it’s its ability to secure and operationalize large blocks of power. The AMD partnership is a bet that the company can convert that power into AI revenue faster than competitors. But the shareholder rejection of the $9 billion buyout is a bet that the company can do it better than any acquirer could.
That’s a high-risk wager.
I’ll be watching three things:
- MW under contract for AI workloads – not just announced, but delivered and live.
- Liquid-cooled capacity – how many sites are ready for GPU clusters.
- AMD’s own execution – if AMD’s chip yields suffer, Core Scientific’s timeline slides.
Chaos isn’t the mining rig noise; it’s the boardroom silence when the next quarterly report misses analyst expectations. The company sprinted toward this pivot, one block at a time. Now it has to prove that sprint was worth rejecting $9 billion.