To The Moon

The global economy is in the middle of the largest technology revolution in human history, as AI promises to change the world as we know it.

However, if we want AI to be able to do more than just create silly cat videos, it needs to be properly trained on lots of high-quality data.

Furthermore, for this data to be processed, the world needs to build more data centers than it has ever built before!

Researchers at McKinsey forecast that by 2030, data center capex will reach $7T and capacity will almost triple to 219 GW!

To meet this insatiable demand for computing power, capacity will come from clever and previously overlooked places. And increasingly, a trend has emerged of bitcoin mining companies turning into AI data center businesses.

For years, aspiring crypto entrepreneurs built ever larger and more sophisticated bitcoin mining operations. These bitcoin farms went from small basement servers to dedicated data centers with thousands of GPUs, custom-designed cooling equipment, and access to cheap energy.

It so happens that an AI data center also needs thousands of GPUs, cooling, and cheap energy!

Thus, many of these clever crypto mining entrepreneurs are taking their crypto mines and turning them into AI data centers.

One of these crypto miners turned AI data center operators is $Cipher Mining Inc.(CIFR)$ .

Let’s dig in.

1. Business Model

Cipher Digital (CIFR) was founded in 2021 with the goal of developing and operating large-scale, efficient Bitcoin mining data centers across the United States.

From inception, the company focused on vertical integration, low-cost power procurement, and purpose-built infrastructure to achieve industry-leading operating efficiency.

However, as the AI technology landscape evolved and demand for high-density compute accelerated, it became increasingly clear that the company’s expertise in power, land, and data center development could potentially generate a much higher ROIC by serving the AI industry.

The company has currently developed 807MW of gross operating and contracted capacity!

A chart of a pie chart AI-generated content may be incorrect.

Cipher Digital Q4 2025 Earnings Presentation.

• 26% of its capacity is used in Bitcoin mining.

• 74% is used in AI Data Centers.

Let’s expand on each segment.

1.1. Bitcoin Mining

As of Q4 2025, had an installed Bitcoin mining capacity of 23.6EH/s, which in 2025 generated $225M in revenues.

Bitcoin miners use EH/s to measure mining capacity, which stands for exahash per second. The Bitcoin mining process works by performing complicated computations that solve digital cryptographic puzzles. These steps are needed to validate Bitcoin transactions and ensure the security of the network.

As a reward for this task, miners are given newly minted Bitcoins!

In real mineral mining, the type and purity of the end mineral determine the final price. However, in crypto mining, there is no differentiation, as a Bitcoin is a Bitcoin, Ethereum is Ethereum, etc. As there is no differentiation in the final product, all miners sell their bitcoin for the same price.

This means that costs ultimately determine a miner’s profitability!

That is why crypto miners have a lot of experience in not only finding the location with the cheapest electricity but also designing energy-efficient data centers. A key metric used to measure the efficiency of a Bitcoin miner is joules per terahash (J/TH), which essentially shows how much electricity a miner uses to complete Bitcoin mining puzzles.

Image

Bitcoin Miners Efficiency

In the picture above, we can see the efficiency of the leading Bitcoin miners. Iren is the most efficient with 15 J/TH, while most of the others use around 20 J/TH.

Looking at CIFR, we see that their efficiency is competitive.

The weighted average efficiency of the portfolio was about 17.2 J/TH, but the company is upgrading equipment in older facilities, which should further improve the fleet efficiency.

Cipher Digital Q3 2025 Earnings Presentation.

The Odessa, Texas, facility has an efficiency of 17.6 J/TH, which is higher than Iren.

However, the Black Pearl location is incredibly efficient, with 13.9 J/TH. This is the newest and most advanced facility, indicating that the company has the skills to build highly efficient AI data centers.

The weighted average electricity cost per bitcoin is about $34,900. With the bitcoin price hovering at around $71,000, CIFR is well-positioned to generate stable bitcoin earnings to reinvest in its AI business.

However, the ultimate plan is to fully transition from Bitcoin mining to AI high-performance data centers!

A graph of a bitcoin mining AI-generated content may be incorrect.

Cipher Digital Q4 2025 Earnings Presentation.

Just recently, the company divested its stakes in the smaller Alborz, Bear, and Chief sites and is retrofitting the Black Pearl data center to serve AI workloads.

As a result, their future BTC capacity has been reduced by half to 11.6 EH/S.

It is clear that Bitcoin mining is the past, and AI is the future.

Subscribe

1.2. AI Data Centers

AI requires fundamentally different infrastructure than general-purpose cloud data centers.

Simply put, AI data centers use a higher number of powerful GPUs that “eat” a lot of electricity, requiring meaningful grid upgrades. Additionally, these GPUs generate more heat than a conventional data center, demanding improved cooling systems. This means that, unfortunately, using existing general-purpose data centers for AI workloads is just not economically viable.

This is why we are witnessing this extreme data center buildout boom!

However, not all AI data center companies are the same, as there are 3 key ways in which they earn revenue.

• Colocation

• Bare Metal

• AI Cloud

In a colocation model, a company such as CIFR builds the data center, secures energy, and equips it with cooling, networking, and other equipment. Then they secure a customer who signs a long-term lease agreement of 5+ years and brings their own GPUs to the facility. This model has the lowest value added and thus has the worst economics, generating the lowest price per MW.

Meanwhile, in a bare metal model, the owner of the data center is responsible for fully equipping the facility with all equipment, including GPUs. It then finds a customer who negotiates access to a pre-agreed number of GPUs for a set per-hour rental price. Bare metal customers use their own software to run AI workloads on these GPUs. As the data center operator provides the GPUs, the bare metal model has a higher value added than colocation, and thus has better economics, generating a higher price per MW.

However, the most profitable AI data center model is the AI cloud provider!

In this model, the data center operator provides advanced AI training and inference software. Purpose-built AI training clusters, model hosting, managed inference services, AI APIs, storage, networking, developer support, and more. An AI cloud doesn’t simply rent out space, like colocation, or rent out GPUs like in bare metal. An AI cloud is essentially an AI development, optimization, inference, and deployment partner for AI start-ups and large enterprises.

As revenues are software-driven, this model has the highest value added and thus generates the highest price per MW of capacity!

Each of the operating models has its own advantages and disadvantages.

In colocation, the data center operator is essentially a landlord, and such a model can be highly profitable if operated well. The key advantage is that it requires less capital to start, is easier to operate, and generates a more stable and predictable rental income from a small number of key tenants.

A bare metal is similar to colocation. But, if we use the analogy of a rental apartment, colocation is comparable to renting an apartment without any furniture, whilst bare metal is a fully furnished apartment. Meanwhile, an AI cloud could be comparable to a fully serviced condo, with furniture, maids, gym, pool, personal chef, nanny, tutor, and chauffeur.

An AI cloud might have the highest margins in theory, but it is also the most difficult model to operate. Companies must employ highly skilled software engineers to build the digital environment that AI start-ups need. This requires a lot of upfront capital and years of research and development. Furthermore, AI clouds must have big sales and customer management teams that manage a large number of small customers.

Most importantly, this area of the industry is highly competitive, filled with highly skilled start-ups such as Nebius, and it is led by Hyperscaler clouds with hundreds of billions of dollars at their disposal, such as Google, Amazon, and Microsoft.

This is why CIFR has no intentions of competing in this area. Their core competency is not AI software, but

1. Finding locations with cheap energy.

2. Building energy-efficient data centers.

So far, the company plans to operate as a provider of colocation services.

Share

2. Data Center Locations

As I said earlier, CIFR’s contracted and operating capacity as of Q4 2025 is 807MW.

477MW of that capacity was fully operational in bitcoin mining, across 5 data centers.

A screenshot of a graph AI-generated content may be incorrect.

Cipher Digital Q3 2025 Earnings Presentation.

The Alborz, Bear, and Chief data centers had a combined capacity of 120MW and were what I would describe as the smaller Bitcoin data centers, so it is no surprise that they were sold.

The Odessa data center has a capacity of 207MW, its 100% owned by CIFR, and is fully dedicated to Bitcoin. The current plan is to use it for Bitcoin mining till July 2027, after that, it will likely be converted to manage AI workloads.

The Black Pearl data center is currently operating as a Bitcoin mining facility with plans to retrofit into an AI data center. The data center is getting expanded from 150MW to 300MW capacity and is already contracted by Amazon to be used as a colocation AI data center for AWS AI workloads.

For its AI business, CIFR currently has 6 new high-performance AI data centers under development in Texas and 1 in Ohio, with a potential capacity of 2.87GW!

• 200MW Ulysses, Ohio

• 70MW Reveille, Texas

• 100MW Stingray

• 500MW Mikeska

• 500MW Milsing

• 500MW McLennan

• 1,000MW Colchis

A graph of energy sources AI-generated content may be incorrect.

Cipher Digital Q4 2025 Earnings Presentation.

Meanwhile, the total capacity pipeline, including existing facilities, is 3.37GW!

As you can see in the picture above, the vast majority of the development pipeline is expected to be energized in 2028-2029, with 2.5MW getting connected to the grid. The 100MW Stingray data center is scheduled to be energized in 2026, whilst the smaller 70MW Reveille is in 2027.

In data center development, energization means that all the electrical equipment is installed and the facility is ready to receive electricity from the utility. It doesn’t mean that the data center is fully ready for production, as all GPUs, networking, and cooling equipment still need to be installed.

CIFR begins recognizing revenue when the colocation facility is fully given to the tenant!

Depending on the contract, that likely means that it will be a few months after energization before revenue recognition begins, and it won’t happen all at once, it will be in stages as more capacity becomes operational.

3. Large Customer Deals

To summarize in a single sentence:

The CIFR business model consists of finding a perfect location for a data center, building it, and signing a large, long-term agreement with a tenant.

So far, the company has signed 2 large data center deals, so let’s analyze them.

3.1. AWS

On October 29, 2025, CIFR signed a 15-year $5.5B colocation agreement with Amazon Web Services.

A black and white website with text AI-generated content may be incorrect.

Cipher Digital Q3 2025 Earnings Presentation.

AWS is the biggest and most powerful cloud infrastructure company in the world. As such, the company has immense bargaining power, and they certainly used it to get this deal with CIFR.

Looking at this deal from a purely economic perspective, it might not seem that stellar at first, but once you look deeper, it actually is quite good.

A $5.5B deal might seem large, but it only amounts to $366.67M per year on average. Furthermore, the agreement is for 300MW of gross capacity, implying an ARR of $1.2M per gross MW, which is quite low.

For instance, WhiteFiber recently signed a 10-year $865M agreement for 52MW gross capacity (40MW IT), implying an ARR of $1.66M. That’s almost 40% higher.

However, the advantage is that this is a large 15 year long term contract with a stable counterpart such as AWS, enabling CIFR to get incredible financing terms.

A close-up of a document AI-generated content may be incorrect.

A month ago, on February 4, 2026, the company did its largest offering to date, raising $2B in secured senior debt.

These notes are due in 2031 and carry a 6.125% interest rate.

Crucially, these notes don’t convert into equity, as they are secured by the AWS Black Pearl data center, essentially an asset-backed loan.

The company has indicated that it will cost them $9-11M per IT MW to build the data center that has a useful life of 25 years. They target a PUE of 1.4, which means that the data center needs 0.4MW of additional energy capacity for each 1MW used for GPUs to cover cooling and other equipment needs.

Therefore, the cost per gross MW is around $6-8M, so let’s model $8M.

We get a total capex of $2.4B and yearly depreciation of $96M.

The $2B note the company just raised to complete the construction of this data center implies that CIFR was able to secure an 83% loan to cost financing. Often, companies can only secure a 70% LTC loan, but the AWS contract was the primary reason CIFR secured 83%. Debt holders were able to value the facility based on the estimated EBITDA that the data center will generate, and apply a multiple.

Amazon will pay all operating expenses for the facility, so we can model a 100% EBITDA margin.

If debt holders used an EBITDA multiple of 8, they would value the data center at $2.9B, and give a 70% loan to value loan.

That gives us a yearly EBIT of around $271M, a margin of 73.8%.

Assuming a 25% tax rate to calculate net operating profit after taxes, we get a NOPAT of $203M. That gives us a ROIC of 8.46%. That ROIC might look deceptively small, but as CIFR is using debt, levered returns are much higher.

Modelling a 25 year, 83% loan to cost loan at 6.125% interest rate that gets refinanced every 5 years, 3% maintenance capex, and an 8x EBITDA multiple to calculate terminal value, I get a levered IRR of 32%.

32% levered IRR might not be particularly exciting for some, but this is an exceptional result for a long and stable data center contract such as this.

Moreover, this deal should also be viewed more from a strategic lens.

With it, CIFR is taking a large step to move from being a Bitcoin miner to an AI data center infrastructure provider. It serves as a backbone on top of which future deals will be built. Whilst absolute dollar returns are not stellar, the company will still generate $271M in EBIT per year. These funds will be reinvested in future data centers, which will be larger in scope, achieving higher economies of scale.

The strategy is to use the AWS earnings to invest in these larger projects that will have better economics, higher ROIC and levered IRR.

Subscribe

3.2. Fluidstack/Google

On September 25, 2025, CIFR signed a 10-year $3B colocation agreement with Fluidstack, which is backed by Google.

A screenshot of a black and white chart AI-generated content may be incorrect.

Cipher Digital Q3 2025 Earnings Presentation.

This deal is different than the AWS one, as instead of signing the final deal with the user (AWS), CIFR signed a deal with a bare metal company Fluidstack.

Here is how it works:

• CIFR builds and owns the data center building and electrical equipment, and leases the facility to Fluidstack.

• Fluidstack pays CIFR rent and equips the facility with GPUs, cooling, and other equipment.

• Google pays Fluidstack a rate to use the GPUs for their AI workloads.

Let’s use a hotel analogy.

CIFR is a real estate owner that leases its building to a real estate company (Fluidstack), which is responsible for equipping the building with beds and other furniture. Then that real estate company finds a hotel operating company that will come with their brand (Google) and will actually run day-to-day operations to find guests (Google Cloud customers) for the hotel.

Google prefers to do it this way, as they are offloading GPU procurement and financing risk to Fluidstack. Essentially, Google is turning GPU capex into operating expenses. Google pays more in the end, but it preserves cash flow today by avoiding expensive GPU purchases, which could be about $5B for such a facility. It reduces the risk, whilst preserving cash flows.

Considering Google plans to spend $180B on capex in 2026, preserving and smoothing out $5B seems like a good idea!

Meanwhile, Fluidstack is partnering with CIFR for the same reason that Google is partnering with them, to lower risk and reduce immediate cash outflow. Instead of spending about $1.7B, it will cost CIFR to develop such a data center, Fluidstack will just pay $300M per year.

CIFR is financing this deal with a $1.73B, 7.125% interest 5-year note.

The economics of this deal are similar to those of AWS, but not as great as they expect a net operating margin of 86%, not 100% as in the AWS deal.

Modelling a 86% EBITDA margin, 25 year, 88% loan to cost loan at 7.125% interest rate that gets refinanced every 5 years, 3% maintenance capex, and an 8x EBITDA multiple to calculate terminal value.

We get a levered IRR of 25%!

4. Capex Requirements

A screenshot of a white and green screen AI-generated content may be incorrect.

Cipher Digital Q4 2025 Earnings Presentation.

CIFR operates in a very capital-intensive industry, and as such, it will need to raise a lot of money from various sources to succeed.

The company has 3.37GW of gross capacity under development, which is scheduled to be energized by around 2030.

Modelling a $8M capex per MW to bring all of this capacity online as colocation AI data centers could cost $27B!

That is a hefty sum for a company with a $5.5B market cap and $250M in revenues. Therefore, it is safe to presume that the company will resort to equity issuances or convertible note offerings to fund a major part of this build-out. Banks will likely not lend more than 70-85% of the buildout cost, so the remaining 15-30% need to come from somewhere.

If we assume that CIFR can finance 70% of the $27B with conventional loans, around $8B must be raised in another way!

This must be done in the next 5 years. If the CIFR stock price doesn’t rise meaningfully, the company might be forced to increase the total number of outstanding shares by 60-100%.

For example, in September, CIFR raised $1.3B from 2031 0% interest rate $16.68 net share settlement convertible notes, about 81M shares. It then spent $82.7M for a capped call option to increase the conversion price to $23.32. They secured the right to pay the principal back in cash and the premium above the conversion price in shares.

So if the share price is $40 in 2031, CIFR has to pay a premium of $16.68 ($40-$23.32) * 81M shares= $1.35B. In $40 shares, that’s about 33.8M shares.

They might be forced to make more raises like these.

However, if banks are willing to continue lending more than 70%, and CIFR is able to issue equity at a higher market cap, that dilution level falls significantly.

5. Valuation Model

With a market cap of $5.5B and revenues of $224M, CIFR trades for 25x sales.

Ordinarily, that would be an incredible multiple, but the company is growing extremely quickly.

So, let’s build a valuation model to see what kind of returns investors could experience.

First, looking at the pipeline capacity of 3.37GW and 800MW of existing capacity, I model CIFR having 3.2GW of gross active capacity in 2030.

$8M capex per MW, with total capex of $25.6B. 80% of that coming from debt, and the remaining 20% from equity, and internal cash flow.

Revenue per gross MW at $1.3M, slightly above Google and AWS deals.

Segment EBITDA of 90%.

We get a total 2030 revenues of $4B, and segment EBITDA of $3.6B!

Interest rate of 7%

Capex useful life of 20 years.

Other opex at 5%. That might seem low compared to other AI data center companies, but the CIFR’s colocation model means that no large R&D, sales, and tech support teams are needed.

The result is $706M in earnings before taxes.

With an effective tax rate of 20%, we get a net income of $565M!

Next, I model the total share count increasing by 25% by 2030!

Assuming an exit multiple of 30x, CIFR could be trading for $33.49 in 2030, and have a market cap of $17B.

That is an upside of 143% to today’s share price of $13.76!

Discounting that back at 16% per year, we get a fair value per share of $15.95, implying the current price of $13.76 is 14% below the fair value.

Furthermore, if the market assigns a higher multiple of 40-50, the upside increases to 225-306%. The multiple depends on the company’s execution and growth outlook past 2030.

6. Conclusion

Cipher is currently in a transition from a Bitcoin miner to a data center infrastructure provider!

If McKinsey’s forecast is even remotely accurate, 219GW of data center capacity will come online by 2030. This is an incredible growth, and traditional data center developers can’t scale their operations fast enough, making Cipher an important partner.

Furthermore, as the agreements with AWS and Google/Fluidstack demonstrate, there is a clear demand for scalable colocation infrastructure by Hyperscalers and large start-ups. Despite the size and power of Google and Amazon, CIFR was able to secure decent terms, generating an estimated levered IRR of 25% and 32%, respectively.

To secure more contracts such as these, the company will need to raise more than $20B in the next 5 years to bring all its pipeline online. If favorable conditions continue, enabling CIFR to fund 80%+ buildout cost with loans, the dilution could be minor.

As the valuation model showed, there is potential for meaningful upside of 143-306% if the AI industry continues to develop at a rapid pace, as the industry forecasts imply!

Lastly, with its colocation model having significantly lower capital requirements, the company can scale with less investment and lower risk, compared to other industry names such as Nebius and Iren.

However, with lower risk, there is also less potential for explosive upside. That’s the trade-off.

For SG users only, Welcome to open a CBA today and enjoy access to a trading limit of up to SGD 20,000 with unlimited trading on SG, HK, and US stocks, as well as ETFs.

🎉Cash Boost Account Now Supports 35,000+ Stocks & ETFs – Greater Flexibility Now

Find out more here.

Complete your first Cash Boost Account trade with a trade amount of ≥ SGD1000* to get SGD 688 stock vouchers*! The trade can be executed using any payment type available under the Cash Boost Account: Cash, CPF, SRS, or CDP.

Click to access the activity

Other helpful links:

• 💰Join the TB Contra Telegram Group to Get $10 Trading Vouchers Now🎉

• How to open a CBA. How to link your CDP account. Other FAQs on CBA. Cash Boost Account Website.