Description

Core Scientific is a Bitcoin miner that went into bankruptcy in late 2022 and reemerged with a new (messy) capital structure in January. I started looking at CORZ as a short along with other BTC miners and I am not the type to like a retail-driven, formerly bankrupt Bitcoin miner. But a recent deal to host Coreweave in their data centers makes the numbers here seem relatively simple to understand and very compelling.

The story changed significantly last week and even though the stock is up ~65% since the deal, I don’t think the market is fully capturing the opportunity. With their analyst day tomorrow on June 12th and more AI-hosting announcements likely, I think there is a chance this moves 50-100% in the next 6 months with relatively little downside – and downside that can be partially hedged by shorting IBIT or MARA.  

This will be a simple pitch by breaking down CORZ three components of value:

1. The Coreweave deal announced June 4 where CORZ will host Coreweave GPUs (200 MW)

2. The remaining 300 MW of High-Performance Compute that CORZ will likely lease to Coreweave (who has options for 60-90 days on some amount) or other AI providers

3. The remaining Bitcoin mining business

1. Coreweave Deal

Coreweave is a former crypto miner that is now a GPU-as-a-service company. They’ve raised over $12B (mostly debt) from Tier 1 investors because they have long-term contracts, mostly from Microsoft. Last week, they announced a 12-year, $3.5B deal to use Core Scientific facilities. The $300M required Capex will be paid by Coreweave and netted against that $3.5B payment (at 50% until repaid), and revenue should begin to show up for CORZ in 2025. 

CORZ management “called their shot” on their Q1 call (which I would highly encourage you to read):

“Based on industry data, we target Tier 1 HPC hosting revenue on the order of $1.4 million to $1.6 million per megawatt per year with gross margin of 75% to 80%. Power costs and utilities are direct pass-through to clients. The complete conversion of 500 megawatts of bitcoin mining infrastructure to HPC hosting would likely take 3 to 4 years, but we expect to begin generating revenue earlier as capacity comes online incrementally during that process.”

This deal is ~$1.5M per MW per year. Below is my estimate for the gross profit impact on CORZ considering a multi-year ramp, the $300M Capex netting, and 25% other COGS.

I assume an 8% cost of capital which seems low but is appropriate given this is essentially a hosting agreement with Microsoft – Coreweave has long-term contracts, is exceptionally well-financed, and is basically a creditor of Microsoft. With conservative assumptions about the terminal value of this facility (less revenue post 2036, higher costs, a mid-teens EBITDA multiple vs ~20x for DCs today), I arrive at $2.1B in value of this deal, or $5.20 per fully-diluted share (cap structure detail later). More aggressive terminal value assumptions gets you closer to $7 per share.  

There’s a question of “Why Core Scientific?” for these HPC workloads which is a good and complicated question. It’s hard to evaluate each individual data center for its readiness to host AI workloads, but there are differences across power access, geographic location, networking, rack density, and other variables. 

Chanos is famously short the legacy data centers because he believes they will require significant Capex to make the facilities “AI ready”. A huge re-rating for CORZ is deserved as the market reprices from skeptical to “contract-in-hand.” 

2. 300 MW of Additional HPC Compute

The second tranche of value is the remaining 300 MW of HPC compute they own and that Coreweave holds options on some portion for 60 to 90 days. This requires more guess work, and is the biggest risk to this investment. Can they lease the compute on the same terms? Better? Worse?

They clearly stated on the call that they believe they can get $1.4-$1.6M per MW per year. Ordinarily I discount what a bankrupt BTC miner says, but they delivered exactly that in the first deal. They had a small deal with Coreweave prior to the June 4th deal that was on worse terms, suggesting it’s possible the 300 MW will be on even better terms. 

I apply a 40% haircut - arbitrary and likely conservative. This yields a $1.9B value and $4.7 per share. A 0% discount, assuming the same value on the 300 remaining MW as the initial 200, yields $7-10 per share depending on terminal value assumptions. 

3. Bitcoin Mining Business

I ascribe very little value to the remaining BTC business. They mined approximately 450 BTC in May (post-halving) with post-halving cash costs around $37k per BTC. Let’s conservatively say that their mining drops by 50% (lease 500 MW entirely to AI out of their 1.2 MW footprint) and assign the mining business a 7x EBITDA multiple and you get ~$600M of value, or $1.5 per share. 

The market clearly values Bitcoin mining at a higher multiple than this due to the implicit option on Bitcoin price (by looking at comps like MARA or other memes like MSTR). In general, I think BTC mining is a terrible business and don’t ascribe much value, though I do think there is some commodity miner value here. 

If you wanted to be more optimistic, you could assume mining costs go down as high-cost miners exit post-halving or as capacity goes offline given the HPC pivot across the sector. 

Capital Structure and Other

Below is a picture of the capital structure which implies ~400M in FDSO and ~$1 per share in cash, assuming all the warrants convert and the convertible notes convert. 

There are some other provisions in the remaining debt, but they are largely immaterial to the investment thesis. Operating expenses are relatively low (~$50M GAAP annualized once you remove bankruptcy advisor costs). 

Summing it all up, we get the following:

1. Coreweave deal: ~$5 per share, potentially as much as $7

2. 300 remaining MW: ~$4.5 per share, potentially as much as $10

3. Bitcoin mining business: ~$1.5 per share

4. Cash + corporate overhead: Around -$1 per share

Net is $10 - $18 per share against a current price of ~$8. 

Sources of additional upside: (i) Anything AI related is trading at a premium to multiples here, (ii) Bitcoin mining is trading at premium to multiples here, and (iii) they start to develop new sites and grow HPC MW.

Sources of downside mainly center on management’s credibility, an inability to lease the remaining 300 MW, and Bitcoin prices imploding while the company continues to mine. 

Coreweave also offered to buy the company for $5.75 (management rejected), so I believe we have 50-100% upside with fairly knowable downside given the signed Coreweave deal and the offer to buy the business. 

All told, I think there is material upside to the current stock and you can hedge out the Bitcoin risks elsewhere (or even by shorting overvalued miners like MARA). 

Catalyst

Analyst Day

More Hosting Agreements