Bitcoin miners are turning mining sites into AI data centers

Bitcoin $BTC ( ▼ 0.73% ) miners are starting to reinvent themselves. Instead of relying purely on block rewards, they’re branching into AI infrastructure and building massive colocation data centers to ride the next tech boom.

These are well-known Bitcoin miners listed on Nasdaq, and what stands out is the move since the sharp pullback in April.

We’re talking mid to high triple-digit returns, which is strong by any standard.

Honestly, it’s hard to find many sectors that have outperformed them recently. That forces the real question: why?

⛏️ The Tough Business of Bitcoin Mining

Firstly, What is bitcoin mining?

Bitcoin runs on a proof-of-work system. That means miners have to do real computational work to earn bitcoin rewards.

Those rewards are built into the protocol, and they get cut in half roughly every four years.

Right now, miners earn 3.125 BTC per block they successfully mine. The next halving is expected in April 2028, and this cycle keeps going until the network reaches the hard cap of 21 million BTC.

So what do miners actually do to earn those rewards?

In simple terms, they solve a puzzle. They use specialized machines called ASICs (built specifically for mining). These machines run an enormous number of guesses per second, trying to find the correct solution.

The first miner to get it right wins the block reward.

In the long run, mining is a simple rule: the more you contribute, the more you earn.

Your payouts will eventually line up with the computing power you contribute to the network. If you provide A% of total hashrate, you should expect roughly A% of the rewards over time.

But here’s the problem. Bitcoin mining is a brutal race where the winners keep running.

• First, running ASICs 24/7 consumes a massive amount of electricity. The biggest edge almost always goes to whoever has cheap, stable power costs.

• Second, you can’t stand still. To stay competitive, you have to keep deploying capex to upgrade into newer ASIC generations that deliver more hash per kWh.

• Third, your revenue is tied to one of the hardest variables to forecast: BTC price. It can expand your margins or crush them within weeks.

And finally, the protocol hits you with a recurring “tax” every four years: the halving. When revenue gets cut in half but your costs don’t automatically fall, simply surviving becomes the real test.

If you look at the “average cost to mine 1 BTC” chart, mining can cost around $99K per bitcoin, while BTC is trading closer to $90K. On the surface, that looks like a losing business.

But that “average cost” number is the trap. There’s no single universal cost to mine one bitcoin.

Costs vary massively from miner to miner depending on electricity rates, ASIC efficiency, location, cooling setup, and how well the operator runs the business day to day.

That’s why some miners don’t run full throttle all the time. They’ll only switch machines on when the math works.

In practice, profitability comes down to two big levers:
- BTC price: The higher it is, the more miners can mine profitably.
- Network hashrate: The lower it is, the easier it is to earn rewards with the same machines.

When BTC drops or competition gets too intense, some operators will curtail or shut down to stop bleeding and wait for better conditions.

And because block rewards are still random in the short term, mining can feel like gambling if you’re solo. That’s why most miners join mining pools, which smooth the variance and turn “lumpy” rewards into steadier payouts.

Right now, a lot of miners are hovering near breakeven, and many show negative free cash flow even when revenues look decent.

That’s because the money gets recycled straight back into the operation. New ASIC purchases, cooling upgrades, infrastructure buildouts, maintenance, and other ongoing capex.

So it’s no surprise they’re looking at AI. It’s not just “AI is hot”, it is because miners already have a head start most companies can’t buy quickly: power access, industrial sites, cooling, and infrastructure that can be repurposed into data center capacity.

🔌 The Opportunity: Bitcoin Miners as Attractive AI Playssiness of Bitcoin Mining

For decades, power demand stayed relatively flat, so there was little urgency to expand the grid or build new capacity. Now AI is changing that. Demand is accelerating, and the gaps in power generation and grid infrastructure are getting exposed fast.

Big tech can absolutely build new data centers. The issue is speed.

A greenfield build takes years, and grid interconnection approvals alone can drag past five years.

That’s exactly why bitcoin miners are suddenly valuable. They already have what AI needs most to scale a colocation data center business:

• Energized sites and grid connections

• Industrial-scale facilities with cooling

• Reliable fiber connectivity

And we’re already seeing early movers lock in large AI infrastructure contracts, with deals reaching hundreds of millions to billions.

This miner-to-AI shift isn’t a side quest. It’s a direct response to a structural power shortage shaping the entire AI economy.

🏗️ How to Spot the Winners (Investment Factors)

To pick the right play here, you don’t need a PhD. You just need to focus on a few variables that actually move outcomes.

Capacity

Because energy is the bottleneck, the first question is simple: how many megawatts (MW) does this operator control.

But don’t stop at “nameplate” capacity. What matters is what they can energize on a real timeline. If they can’t bring MW online, they can’t bill for it.

The rule is: more energized MW = more revenue potential.

A few second-order details help you separate signal from noise. Higher data center density usually means better efficiency. A single large site is often easier to run than multiple scattered sites. And location matters because proximity to major metros tends to mean better fiber connectivity.

Track record and financial health

Next, look at what you’re actually buying: a team and a balance sheet.

• Has the company been run well over time?

• Did it create shareholder value?

• Or did it survive by stacking debt and diluting equity?

Mining being hard is not the point. The point is whether management has proven they can operate through cycles without constantly punting the bill to shareholders. Because if they couldn’t execute in mining, why should you assume they’ll execute in AI infrastructure, which is even more capital-intensive and contract-driven.

Execution matters too as a slide deck is cheap but delivering energized MW on schedule is not.

So watch if they are hitting milestones, communicating clearly, and responding well when something breaks.

That’s usually where pretenders get exposed.

Strategy

Finally, you need to understand their operating model, because it defines both upside and risk.

Think of it like this: MW is the raw input. GPUs turn that MW into AI output.

There are two main approaches:

1) GPU cloud providers
They sell the full stack: data center plus GPUs. It’s attractive because customers can deploy fast.

• It’s risky that they own the GPU inventory, so they eat hardware obsolescence, upgrade cycles, and supply-chain access to the newest chips.

• The upside is they can charge more per MW because they’re selling a “ready to use” product.

2) Colocation operators
They provide the essentials: power, space, cooling, connectivity, security inside a colocation data center environment..

• The customer brings the GPUs. That lowers risk because the operator isn’t tying up capital in fast-depreciating hardware or betting on which chip generation wins.

• The trade-off is simpler economics: lower revenue per MW, but typically more stability.

In general, which one is “better” depends on you. If you want maximum upside and you’re comfortable with GPU cycle risk, GPU cloud can be the higher-beta bet.

If you want cleaner infrastructure exposure and fewer moving parts, colocation is usually the steadier lane.

Now that you’ve got the framework, we can actually compare the operators and see who looks like the smartest risk-adjusted bet.

⭐ Comparison of Colocation Operators

To keep the comparison clean, we need to line up operators that run the same business model.

So for this section, we’re focusing only on colocation operators running the colocation data center model and leaving out GPU cloud providers.

The reason is simple: GPU cloud is higher beta. You’re taking on GPU inventory risk, upgrade cycles, and chip availability dynamics. That’s a different game, and it’s harder to underwrite without going deep on the GPU market.

That also means a quick change to the earlier list. We initially highlighted IREN, WULF, and CIFR.

But IREN fits the GPU cloud model, so we’re removing it here. In its place, we’re adding Core Scientific (CORZ), which is aligned with colocation.

And we’re also adding Galaxy $GLXY ( ▼ 10.42% ) .

Yes, Galaxy isn’t a pure bitcoin miner. But for this analysis, the point is “AI data center exposure via colocation.” And Galaxy belongs in that bucket.

So our side-by-side set is: CORZ, WULF, CIFR, GLXY.

Now we start with the basic question every investor should ask first: what kind of company are you actually buying?

From the key metrics table, market caps are similar for the pure crypto miners, roughly around $5B but market cap alone doesn’t tell you the quality of the balance sheet.

Cash matters and CORZ is the lightest on cash at around $450M, while CIFR and GLXY are sitting north of $1B.

And for GLXY specifically, it also holds meaningful crypto assets. That’s not the same as cash, but it can become liquidity fast if they need it.

Regarding debt, Galaxy shows the highest debt number, roughly $1.47B, and at first glance that looks like higher risk. But in this case, that debt is tied to financing they’ve already secured for Phase I of their buildout. It’s essentially “funded progress,” not just financial stress.

For the other crypto miners, lower debt today doesn’t automatically mean healthier. It can simply mean they haven’t raised yet.

The typical sequence is: land contracts first → raise financing against those contracts → build and energize capacity.

By that logic, most of the crypto miners are still earlier in the cycle than Galaxy.

Next, comparing Enterprise Value:

EV = market cap + debt − cash.

EV matters because it normalizes leverage. Two companies can have similar market caps but very different levels of debt and cash, which changes what you’re really paying for the business.

Then there’s Book Value, which is the simplest “downside lens.”

Book value = assets − liabilities.
It’s what belongs to shareholders on paper.

Higher book value gives you more of a floor. Lower or negative book value means there’s less protection if sentiment turns.

In this table, CORZ shows a negative book value. That’s the kind of structure where, in a true downside scenario, equity can get wiped.

GLXY, on the other hand, shows a large positive book value, which gives you a much more defensible balance sheet profile.

Now zoom out to “accumulated profits.” This is the scorecard of how brutal bitcoin mining has been historically.

For the crypto miners, accumulated profits are negative. That means over their operating history, they’ve burned more than they’ve earned.

Galaxy is the outlier here. It shows positive accumulated profits, which is a very different track record versus companies that have been structurally loss-making.

And when you look at more recent profitability metrics like adjusted EBITDA and EPS, for most miners, EBITDA is around flat and EPS tends to be negative.

Galaxy stands out because its EBITDA and EPS reflect a broader crypto business that already generates cash and can fund growth.

That’s the key investor takeaway from the first table. GLXY looks like a company with a stronger balance sheet and a history of profitability, while the crypto miners still carry a higher risk premium because they have more to prove.

Now we move from history to future, because the AI pivot is all about forward capacity.

The most important concept in the capacity table is the difference between gross capacity and critical IT capacity.

Gross capacity is what the facility can handle in total, including overhead like cooling and power losses.

Critical IT capacity is the MW you can bill for, and it’s always smaller than gross capacity.

• Row 1 is energized capacity today. That’s what they can sell right now.

• Row 2 is projected billable capacity by the end of 2028. That’s the forward number that drives revenue.

This one metric is basically the spine of the whole thesis: where they are today versus what they can realistically energize over the next few years.

Pipeline MW matters too, but it’s lower conviction until approvals are secured. Site concentration matters because power in one concentrated location is generally more valuable and operationally simpler than the same MW spread across multiple distant sites.

Then come contracts, and this is where the game becomes real.

The best setup is when projected 2028 energized MW is already contracted. That turns “future potential” into “future revenue visibility.”

In this table, CORZ and GLXY appear to have high contract coverage into their 2028 capacity.

But there’s a trade-off. If 100% of projected capacity is tied to a single tenant, your revenue becomes more fragile. One customer issue becomes a whole-company issue.

That’s why tenant quality matters as much as MW when you’re underwriting a colocation data center operator.

In this case, the key tenant is CoreWeave. The risk looks lower because CoreWeave has strong strategic positioning in the GPU ecosystem, including Nvidia as an equity holder, plus serious software advantages in GPU workload orchestration.

That helps explain why hyperscalers are willing to work with them, and why these colocation contracts can be credible.

Now look at the unit economics. Revenue, margin, EBITDA, EBITDA per MW.

Galaxy leads on margin and also leads on EBITDA per MW, which implies it’s extracting more earnings from each unit of energized power than peers.

But you can’t stop at “best deal per MW.” You need to apply those economics to projected 2028 capacity to estimate the full earnings power of each operator

Once you have EBITDA, you can sketch valuation using a peer multiple. Data center businesses often trade on EBITDA multiples, so using something like 26x EBITDA creates a comparable baseline for Enterprise Value.

Then you adjust for the capital reality. Enterprise Value includes debt, and scaling MW requires a lot of capex.

Using a capex assumption around $13M per MW gives you a reasonable industry yardstick for what buildouts cost.

The big uncertainty is funding strategy. Some will finance with debt, some will issue equity, and some will do a mix. But if a company already has weaker balance sheets, higher dilution history, and negative cash flow, it’s hard to assume they’ll raise on better terms than the strongest operator.

That’s how you arrive at the final comparison: projected 2028 value versus today’s market cap.

Some names show big upside on paper but you need to interpret that upside correctly.

For the pure bitcoin miners, the AI data center move is often a replacement trade. They’re swapping one volatile business line for another that they hope becomes more stable and higher margin.

For Galaxy, it’s different.

The AI infrastructure buildout is additive. It’s a new line layered on top of an existing crypto platform that already throws off cash flow. That’s why even if the model shows a certain percentage upside, the strategic value can be larger.

If the AI segment alone can generate massive standalone value, it can re-rate the entire company beyond what a simple “miner comp” framework captures.

Now we can wrap this section and move into the final takeaways with a cleaner lens!

⚡ Key Takeaway

Let’s run through the four names quickly, with the same lens: upside math vs execution risk.

Terawulf $WULF ( ▼ 9.48% ) :

WULF screens as the highest upside on paper, around 421%, so it’s the one that grabs attention first.

But that upside is coming from a very aggressive capacity ramp. They’re projecting 1,094 MW of energized critical IT load by 2028, the largest target in the group.

The problem is where they’re starting from. With only ~18 MW online today, they’d have to add roughly 350 MW per year for the next three years.

That’s doable in theory, especially if approvals are in place, but execution across multiple sites is the hard part. That’s where timelines slip and costs creep.

Then there’s the shareholder history. Over the last three years, WULF has diluted shareholders by 305%+, which tells you how they’ve funded survival.

Pair that with -$870M in accumulated losses and you get the full picture: the upside exists, but the risk premium is huge. For me, this is a watchlist name, not a “sleep well” position.

Core Scientific $CORZ ( ▼ 5.0% ) :

CORZ shows a decent upside case, roughly 122%, and it has real traction in the colocation narrative.

But the track record is the red flag you can’t ignore. Accumulated profits are around -$4.2B, and they filed for bankruptcy in 2022.

Yes, restructuring kept them alive. But that history matters because it influences negotiating power and financing terms.

You can see that show up in economics too. Their CoreWeave deal is priced around $1.44M per MW, which is meaningfully lower than Galaxy’s $1.90M per MW.

So while CORZ may execute, the risk-adjusted setup doesn’t look like the best bet in the group. It’s not a top pick for me.

Cipher Mining $CIFR ( ▼ 9.69% ) :

CIFR is the “quietly interesting” one.

On the model, upside looks small, around 5%, but the balance sheet is healthier relative to the other crypto miners.

They still have negative accumulated profits (about -$260M), but the latest quarter showed a real improvement: $40M in adjusted EBITDA, the strongest recent operating read among the three crypto miners.

The bigger bull case is capacity expansion. Management has talked about targeting ~1 GW energized by 2028, but approvals are still a gating factor, so you can’t underwrite it as certain.

And if you extend the horizon to 2029, they’ve suggested a path toward ~2,500 MW across 2028–2029. If that starts becoming concrete, the entire valuation could re-rate.

So CIFR isn’t a screaming buy off the spreadsheet today, but it’s absolutely one you keep close. The upside isn’t in the base case, it’s in execution + approvals turning the roadmap into reality.

Galaxy $GLXY ( ▼ 10.42% ) :

GLXY is the cleanest setup here, and it’s my best pick.

Galaxy has a strong balance sheet and positive cash flow from its existing crypto business, which lowers the need for dilutive financing.

On the AI data center side, they’re also ahead in timeline. Phase I is already funded, and early cash flows are expected to begin around Q1 next year, while others are still working through the “sign contracts → raise money” phase.

And the unit economics are simply better. Galaxy has secured the strongest pricing in the group at roughly $1.90M per MW, which tells you tenants are willing to pay a premium for their execution and site quality.

The most important strategic point is that for the crypto miners, AI often replaces mining economics. For Galaxy, it’s additive. AI infrastructure becomes a new revenue engine layered on top of the existing business, not a swap.

That’s why GLXY is the one I’d be most comfortable holding through noise. If the market keeps chasing “bitcoin miners pivoting to AI,” Galaxy is positioned to benefit without wearing the same balance-sheet fragility as the pure miners.

⚠ This newsletter is for informational purposes only and should not be considered investment advice. Traders should conduct thorough research, understand the risks, and carefully evaluate their decisions before investing in cryptocurrency.

If you’re interested in other topics and want to stay ahead of how Crypto are reshaping the markets, from whale strategies to the next major altcoin narrative, you can explore more of our deep-dive articles here:

• Macro Environment: The AI Race is Hotter*

• The Crypto Sell-Off and Get Ready for the Old Coins

• Latest Crypto News: Bitcoin Reawakens, TAO Halving Hype Builds

• AI in Trading: How ChatGPT Atlas Could Redefine Trading Strategy*

• Zcash Pumps 30%, and Bitcoin Prepares for a Breakout

*indicates premium insights available to Pro readers only.