Core Scientific, once one of the largest publicly traded bitcoin miners in North America, has signed a massive 15-year infrastructure deal with AMD that could reshape the company’s identity forever — and the bitcoin mining industry is taking notes.
The agreement, announced on July 28, locks in 529 megawatts of U.S.-based AI computing capacity for AMD across Core Scientific facilities in Texas and Oklahoma. The total base contracted revenue from the deal? More than 14 billion over a decade and a half. And if AMD exercises its options to reserve additional capacity through 2028, the partnership could scale to roughly 2.5 gigawatts.
The Deal That Accelerated a Pivot
This is not a side project. AMD directly leased 377 MW across Core Scientific sites in Pecos and Hunt County, Texas, plus Muskogee, Oklahoma. An unnamed cloud provider leased another 152 MW in Auburn, Alabama, and Dalton, Georgia, under agreements supported by AMD. The infrastructure is expected to support customer deployments beginning in 2027.
The collaboration goes beyond simple real estate. Core Scientific and AMD will work together on data-center design and the deployment of AMD Instinct GPUs, EPYC processors, and ROCm software. In other words, Core Scientific is transforming from a company that once packed warehouses with bitcoin-mining rigs into a purpose-built AI infrastructure provider.
AMD also received warrants to purchase up to 30 million Core Scientific shares at 23.47 per share. About 6.5 million vested immediately upon signing, with more vesting as additional capacity gets contracted. That aligns AMD’s financial interests directly with Core Scientific’s success.
The Numbers Tell the Story
The financial impact is already visible. The new agreements lift Core Scientific’s total leased customer capacity to approximately 1.1 gigawatts, representing more than 24 billion in potential contracted revenue. As of mid-July, the company was billing customers for 437 MW — equivalent to about 635 million in annualized colocation revenue.
The second-quarter breakdown is even more revealing. Colocation generated 136.7 million, accounting for 83% of the company’s 164.2 million in Q2 revenue. Meanwhile, self-mining revenue collapsed by 66% year-over-year, falling to just 21.5 million. The message from the financials is unambiguous: bitcoin mining is no longer the core business at Core Scientific.
Block Deal Terminated — A Clean Break
Perhaps the starkest signal of the pivot is the termination of Core Scientific’s agreement with Block to purchase the payments company’s custom bitcoin-mining chips. The 2024 deal covered 3-nanometer chips representing roughly 15 EH/s of hashrate — a significant chunk of mining power. Core Scientific recorded a 41.9 million charge to walk away from it.
That termination is not just a financial write-down. It represents a strategic decision to stop investing in next-generation mining hardware and instead redirect capital and energy capacity toward AI infrastructure. The bitcoin network will lose that hashrate; AI data centers will gain the power.
What This Means for Bitcoin Mining
With BTC trading at 63,421 as of this writing, the economics of bitcoin mining remain under pressure. Energy costs, regulatory uncertainty, and post-halving block rewards have squeezed margins across the industry. Core Scientific’s response — pivoting to AI — is a playbook that other miners are watching closely.
The company held 848 BTC on its balance sheet as of June 30, up from 547 BTC three months earlier. But it also sold 2,385 BTC for 208.2 million in the first quarter. That selling pressure, combined with the termination of the Block chip deal, suggests Core Scientific is converting its bitcoin treasury into AI infrastructure capital.
For the broader mining sector, the implications are significant. If one of the largest publicly traded miners is willing to abandon next-generation hardware investments and eat a 41.9 million charge to exit mining commitments, others may follow. The total network hashrate could face headwinds if more large operators redirect their energy capacity toward AI workloads rather than securing the bitcoin network.
The AI Computing Arms Race
The Core Scientific–AMD partnership exists within a broader context. Demand for AI computing infrastructure has exploded, and the companies that control large-scale energy capacity — which is exactly what bitcoin miners built over the past decade — are uniquely positioned to capitalize. Data centers need power, cooling, and physical space. Bitcoin mining facilities have all three in abundance.
The deal also gives AMD a foothold in the infrastructure layer, not just the chip layer. By securing long-term capacity at Core Scientific sites, AMD ensures that its Instinct GPUs and EPYC processors have homes where they can run for paying customers. That vertical integration — from silicon to data center — could prove to be a competitive advantage against rivals like Nvidia.
CORZ shares responded positively, rising 5.6% in pre-market trading on the announcement. AMD shares were lower by 4% as the broader chip sector continues to face selling pressure — but the strategic value of this deal extends well beyond daily stock movements.
The Verdict
Core Scientific’s AMD deal may come to be seen as a defining moment for the bitcoin mining industry’s evolution. The 14 billion in potential base revenue from AI colocation dwarfs what the company could earn from mining bitcoin at current prices. With ETH at 1,875 and SOL at 73.12, the broader crypto market offers no immediate relief for miners struggling with compressed margins.
The question for investors and industry observers is whether Core Scientific is an outlier or a trendsetter. If more mining companies follow this path, the bitcoin network’s security model could face structural changes — fewer professional miners, more fragmented hashrate distribution, and potentially higher transaction fees as block subsidy continues to decline post-halving.
For now, Core Scientific has made its bet. The company that once bet everything on bitcoin is now betting that AI will pay better. The next few quarters will reveal whether that gamble pays off — and whether the rest of the mining industry agrees.
Disclaimer
This article is for informational purposes only and does not constitute financial advice. Cryptocurrency investments carry significant risk, and readers should conduct their own research before making any investment decisions. The author holds no positions in the assets mentioned in this article.
529 megawatts redirected from BTC mining to AMD AI compute. core scientific basically just admitted mining isnt where the money is anymore
Muskogee, Oklahoma getting 377 MW of AI compute is wild. That town is about to become one of the most important data center hubs in the country.
529 MW for AMD is massive. Core Scientific basically looked at mining margins and said nah, AI pays better
$14 billion over 15 years for a company that was nearly bankrupt in 2022. The pivot to AI infrastructure might be the best timed move in mining history.
imagine being a BTC maximalist and your favorite miner just pivoted to GPUs for AMD lol. the hashrate exodus is gonna be rough
14 billion over 15 years. do the math on that, its under 1B a year. still huge but the headline makes it sound bigger than the NPV
also the expansion options through 2028 sit with amd, not corz. 2.5 GW of headroom sounds great until you read who actually controls the expansion clause. still a solid deal at any real discount rate
the flip side is amd holding the option blocks every competitor from that power. 2.5 GW of headroom locked behind their clause is a moat against the next miner trying this pivot in the same regions
amd holding the option blocks is also amd holding the risk. if ai demand softens those 2.5 gw of options quietly never get exercised and corz is left defending empty halls
true but amd pays reservation fees on those option blocks either way, cash shows up before any gpu lands. optionality with a floor beats hashprice roulette
meanwhile smaller miners are still holding onto S19s hoping BTC goes to 200k. read the room guys, the energy arbitrage is in AI now
Pecos and Muskogee getting repurposed. those were tier-2 sites back in 2021, now they are AI datacenters lol wild
529 MW for AMD across Texas and Oklahoma. Core Scientific basically looked at mining economics and said yeah no thanks
14 billion over 15 years makes this the biggest pivot in mining history. every other publicly traded miner is probably running the same math right now
Roshan D. the ones without Texas real estate and cheap power contracts are stuck. this pivot only works if you already have the infrastructure footprint
The math only works if you already own the land and the interconnects. Core Scientific spent 2021 buying sites next to substations in Texas. Most miners cannot even queue for transmission today.
Pecos and Hunt County are basically substations with land attached at this point. Anyone starting a GPU campus from scratch today is three to five years behind on the interconnect queue.
substations with land attached is the exact phrase. i drive past the pecos site monthly, the interconnect infrastructure alone took years. no greenfield competitor replicates that by 2028
three to five years is optimistic in ercot. some of the 765kV projects in west texas have been queued since before corz even exited chapter 11
ercot queues are brutal, some of those west texas 765kV lines predate the whole ai pivot. corz sitting on finished interconnects is the entire moat, the amd deal just monetized it
529 MW redirected from hashing to inference is the quiet headline. once half the public fleet follows, the btc security budget conversation stops being theoretical
Roughly 900 million a year for 15 years before any options get exercised. For a company that was in chapter 11 not that long ago, one signed contract did more for the risk profile than a decade of hashrate records.
The 900M a year figure is floor revenue too, before the 2.5 GW options even price in. Chapter 11 to investment grade paper in four years is quite the turnaround.
Fifteen years of contracted revenue from a chipmaker beats praying for hashprice every quarter. CORZ spent 2021 explaining dilution, now it explains optionality. Wild turnaround.
529 MW off the btc network barely dents hashpower today but picture corz, riot and marathon all doing this by 2028. the security budget conversation stops being hypothetical fast
The precedent here matters more than the revenue. Every mid-size miner board is now asking why they are still chasing hashprice when one hyperscale lease reprices the whole equity.