IREN: The power behind the AI race

From Bitcoin mining to AI infrastructure – a $9.7B Microsoft deal, a $3.4B NVIDIA deal, a 5GW+ power pipeline and the question that decides everything: what happens if the market starts valuing IREN as an AI infrastructure company?

NASDAQ: IREN | AI INFRASTRUCTURE | HIGH RISK / HIGH POTENTIAL

From Bitcoin mining to AI infrastructure. Why we believe IREN could be one of the more interesting opportunities in the next phase of the AI race.

Executive summary

The AI infrastructure race is entering a new phase.

For the past two years, the market has largely focused on GPUs, AI models and the companies developing the technology. But as AI adoption accelerates, another bottleneck is becoming increasingly important:

Electricity.

Without sufficient power, there are no large-scale AI data centers. Without data centers, there is nowhere to deploy GPUs. And without computing capacity, the AI economy cannot continue expanding at its current pace.

This is where IREN becomes interesting.

IREN is transforming from a Bitcoin mining company into a vertically integrated AI cloud and infrastructure provider.

The company combines power access, land, data center infrastructure, NVIDIA GPU deployments and multi-billion-dollar customer agreements.

It has signed a $9.7 billion agreement with Microsoft and a separate $3.4 billion AI Cloud agreement with NVIDIA.

At the same time, the share price has experienced a substantial correction.

We believe the market is still trying to determine what IREN should ultimately be valued as: a Bitcoin miner, an AI cloud provider or an infrastructure company controlling some of the resources needed to power the next generation of AI.

Our thesis is that the company could become considerably more valuable if the transition succeeds.

We also believe IREN offers an interesting risk/reward profile relative to CoreWeave and Nebius.

Not because those businesses are inferior, but because IREN combines several attractive characteristics at a valuation that we believe deserves closer attention.

Our central question: What happens if the market starts valuing IREN as an established AI infrastructure company rather than a former Bitcoin miner?

01 · The transformation – a completely different business

IREN's history is rooted in Bitcoin mining.

The company developed large-scale facilities in locations where it could access suitable electricity and infrastructure to operate mining equipment.

But the economics of the AI industry have created another opportunity.

The same underlying resources that make a location attractive for Bitcoin mining can also be valuable for AI infrastructure.

Land, electricity access, grid connections, substations and data center development capabilities are all relevant.

IREN has recognized that the long-term opportunity could be substantially greater in AI Cloud services.

Instead of simply continuing to expand its mining operation, the company is redirecting its infrastructure towards high-performance computing.

This is not a small adjustment to the business model.

It is a fundamental transformation.

The company is moving from generating revenue through Bitcoin mining towards providing computing capacity to AI developers, enterprises and hyperscalers.

The difference is important.

Bitcoin mining revenue is heavily influenced by Bitcoin prices, network difficulty and mining economics.

AI Cloud revenue is driven by GPU utilization, contracted pricing, customer demand and the ability to deliver reliable computing capacity.

Both businesses are capital-intensive, but their revenue drivers and long-term valuation potential are very different.

02 · The Microsoft deal – $9.7 billion in contract value

One of the most important milestones in IREN's transformation was its agreement with Microsoft, announced in November 2025.

The five-year agreement represents approximately $9.7 billion in total contract value.

It involves dedicated GPU cloud services delivered from IREN's Horizon data center facilities in Childress, Texas.

The project consists of four separate 50MW IT deployments, representing approximately 200MW of total IT capacity.

The first deployment, Horizon 1, was delivered and accepted by Microsoft in August 2026.

Horizon 2 is undergoing commissioning, while Horizon 3 and 4 are in late-stage construction.

The remaining deployments are targeted for delivery in phases during Q4 2026, with contractual grace periods extending into early 2027.

Why does this matter?

Because IREN is not simply building data centers and hoping customers will arrive.

It has secured a major customer willing to commit billions of dollars to its infrastructure.

Microsoft provides an important level of commercial validation.

However, contract value must be separated from recognized revenue.

The $9.7 billion is spread across five years. It is not immediate cash revenue, and the remaining phases must be delivered and accepted.

That is why execution remains critical.

03 · NVIDIA – a strategic relationship

In May 2026, IREN entered into a five-year AI Cloud services agreement with NVIDIA representing approximately $3.4 billion in total contract value.

The agreement supports NVIDIA's internal AI and research workloads.

IREN also entered into a strategic partnership intended to support the deployment of up to 5GW of NVIDIA DSX-aligned AI infrastructure across its global development pipeline.

The partnership included a potential NVIDIA investment of up to $2.1 billion through a conditional right to purchase IREN shares.

This is strategically important.

NVIDIA is positioning itself across the infrastructure ecosystem required to expand AI computing.

IREN benefits from being aligned with one of the most important companies in the AI industry.

The relationship potentially provides:

• Access to NVIDIA's technology ecosystem.
• Alignment with next-generation GPU infrastructure.
• Greater credibility when approaching large AI customers.
• Opportunities to expand future deployments.
• Potential strategic capital support.

We should be precise: the $2.1 billion investment right is conditional and should not be treated as capital already received.

But the strategic relationship is an important part of the investment thesis.

04 · The financials – the transition is already visible

IREN's FY2026 financial results illustrate how quickly the business is changing.

Fiscal year ended June 30, 2026.

AI Cloud revenue increased from $16.4 million in FY2025 to $128.8 million in FY2026.

That is approximately eight times growth.

Bitcoin mining still represented the majority of annual revenue, demonstrating that the transition is not yet complete.

The reported net loss of $702.6 million also needs context.

A substantial portion was driven by non-cash impairment charges, including $638.8 million primarily related to decommissioning Bitcoin mining hardware as facilities are converted to support AI Cloud growth.

This does not mean the loss is irrelevant.

Depreciation, financing expenses, capital expenditure and operating costs all matter when assessing the long-term economics of the business.

But we believe investors should focus particularly on how the revenue mix evolves over the next several quarters.

The important question is whether AI Cloud can become the dominant and profitable revenue engine.

05 · The $4 billion ARR opportunity

One of the most interesting figures in IREN's August 2026 business update was its approximately $4 billion contracted annualized run-rate revenue associated with 2026 capacity.

The company also reported approximately $1 billion in operating ARR at the time.

This distinction is extremely important.

Operating ARR: approximately $1 billion.

This represents annualized revenue from commissioned GPUs under contract.

Contracted ARR: approximately $4 billion.

This represents the broader contracted capacity targeted to become operational by December 2026.

ARR is not the same as recognized accounting revenue or cash received.

The additional capacity still needs to be commissioned, tested and accepted by customers.

IREN also reported that its 2026 capacity was largely sold out.

For recent three-year contracts, pricing exceeded $20 million per MW of IT capacity, with active discussions around approximately $25 million per MW.

Customer prepayments on recent contracts represented approximately 45–55% of associated GPU capital expenditure.

This is a particularly interesting development.

It suggests IREN may be able to expand its AI Cloud business without financing every new GPU deployment entirely through its own balance sheet.

But the difference between contracted ARR and actual operating revenue remains one of the most important execution risks.

06 · The real moat – electricity and infrastructure

This is where we believe IREN has one of its most interesting competitive advantages.

The AI infrastructure industry is increasingly becoming an electricity business.

A company can have access to capital and purchase thousands of GPUs, but without sufficient electricity, grid connections, cooling and suitable data center locations, those GPUs cannot generate revenue.

IREN has spent years developing the infrastructure needed to operate at scale.

Its global development pipeline exceeds 5GW of data center capacity.

This includes locations and projects in:

• Childress, Texas.
• Sweetwater, Texas.
• Kiowa, Oklahoma.
• Mackenzie, British Columbia.
• Canal Flats, British Columbia.
• Prince George, British Columbia.
• Bundey, Australia.
• Badajoz, Spain.

The company is targeting cumulative delivered IT capacity of approximately:

These figures should not be confused with the 5GW development pipeline.

A pipeline does not mean 5GW of operational AI computing capacity, nor does it mean IREN owns 5GW of electricity generation.

IREN is securing power access, grid infrastructure and development opportunities.

The distinction matters.

Nevertheless, the ability to develop large-scale data centers in locations with access to substantial power could become a major competitive advantage.

In a market where electricity and grid connections are increasingly difficult to secure, this is a valuable position.

We believe IREN's power portfolio is one of the most important parts of its long-term investment thesis.

07 · Bitcoin mining – they have not abandoned it yet

An important part of understanding IREN is that Bitcoin mining has historically been the foundation of the company.

During FY2026, Bitcoin mining generated $578.2 million in revenue.

That business has helped IREN establish its power infrastructure, develop large-scale facilities and build operational experience.

The company is now transitioning away from Bitcoin mining, with a stated objective of substantially exiting the business by December 31, 2026.

This creates an interesting transition.

IREN can redirect resources and infrastructure towards AI Cloud services while replacing exposure to Bitcoin mining economics with longer-term computing contracts.

The transition is not without costs.

Decommissioning mining equipment has already resulted in substantial impairment charges.

But we believe the opportunity is worth examining.

IREN is not a startup trying to build an AI infrastructure business from nothing.

It is an established infrastructure operator attempting to reposition existing assets for a different and potentially more valuable market.

08 · Financing and debt – the most important risk

This is where we need to be particularly careful.

IREN is investing enormous amounts of capital into its expansion.

GPUs, data centers, cooling systems, substations and grid infrastructure all require significant financing.

At June 30, 2026, IREN reported:

• $5.90 billion in cash and cash equivalents.
• $1.72 billion in restricted cash.
• $13.81 billion in contractual capital commitments.

The company also secured substantial GPU financing.

The Microsoft deployment includes approximately $3.6 billion in investment-grade GPU financing at a weighted interest rate of approximately 6%.

Together with customer prepayments, this structure was expected to fund approximately 96% of the associated GPU capital expenditure.

Additional GPU financing of approximately $2.8 billion was arranged for other customer deployments.

One facility, led by Blue Owl and PIMCO as adviser to certain investors, represented approximately $2.4 billion at a fixed interest rate of 9%.

This is important because it shows how IREN is attempting to finance its expansion.

Rather than funding every GPU deployment through ordinary corporate borrowing or issuing new shares, it is using a combination of customer prepayments and asset-backed financing.

However, this does not eliminate risk.

The company still has substantial capital commitments, financing expenses and potential dilution.

There are also risks associated with customer acceptance, GPU utilization and the residual value of hardware.

Our view: IREN's financing structure is interesting, but the company must demonstrate that the returns from its infrastructure justify the capital invested.

09 · Why IREN over CoreWeave and Nebius?

This is perhaps the most important section of our entire investment thesis.

We are not saying that CoreWeave or Nebius are bad companies.

Both have built impressive businesses in a rapidly expanding market.

But when comparing these three companies, we believe IREN offers an interesting combination of infrastructure assets, power access, strategic partnerships and potential valuation upside.

Power figures are not perfectly comparable: they refer to different stages and definitions of capacity, and should not be interpreted as equivalent operational IT capacity.

CoreWeave – a great business, but a different risk profile

CoreWeave has established itself as one of the leading specialist AI cloud providers.

Its revenue growth is impressive and its customer relationships are significant.

But its expansion is extremely capital-intensive.

The company has a large financing structure involving debt, leases and ongoing capital expenditure.

At the same time, its enterprise value is considerably higher than its market capitalization, reflecting the financing obligations investors need to consider.

We see CoreWeave as a serious competitor, but we believe IREN offers an interesting alternative through its power portfolio, infrastructure history and Microsoft/NVIDIA agreements.

Nebius – explosive growth, but a higher expectation

Nebius is another company we follow closely.

It has demonstrated extraordinary revenue growth and is aggressively expanding its AI infrastructure.

The market has already assigned a substantial valuation to its future potential.

We like the business, but the valuation leaves less room for disappointment in our opinion.

Nebius also has significant financing requirements as it expands its global infrastructure.

IREN, on the other hand, offers a different entry point into the AI infrastructure market.

Its transformation is still underway, and we believe that uncertainty may create an opportunity if management executes successfully.

10 · The valuation – what are we actually paying?

As of October 30, 2026:

These figures show why valuation needs to be examined carefully.

IREN's enterprise value is approximately $11.4 billion, compared with $76.3 billion for CoreWeave and $44.4 billion for Nebius.

However, IREN also has considerably less historical revenue than CoreWeave.

On trailing enterprise value to revenue, IREN is not cheaper than CoreWeave. Its multiple is approximately 16x, compared with approximately 10x for CoreWeave.

Nebius is considerably more expensive at approximately 33x.

This is an important distinction.

We do not want to claim that IREN is the cheapest company on every traditional valuation measure.

Our argument is different.

IREN's historical revenue still contains a substantial Bitcoin mining component, while its AI Cloud business is expanding rapidly.

That means historical revenue multiples may not fully capture the future business mix.

The more interesting valuation question is what the company could look like once the contracted AI capacity becomes operational and contributes to recognized revenue.

We believe there is potential for a valuation re-rating if the transition succeeds.

Valuation data: October 30, 2026. Enterprise value and market capitalization are market-data estimates and can change with share prices, financing and reporting updates.

11 · The asymmetry – where we see the opportunity

This is the central reason we prefer IREN for our November case.

The opportunity is not based on one single catalyst.

It is the combination of several factors.

• 1. Power – A development pipeline exceeding 5GW gives IREN a substantial platform for future expansion.
• 2. Infrastructure – The company already owns and operates important parts of the physical infrastructure needed for AI computing.
• 3. Customer contracts – Microsoft and NVIDIA have committed billions of dollars in multi-year agreements.
• 4. Financing – Customer prepayments and GPU-backed financing can support expansion while reducing the need to fund every deployment through equity.
• 5. Business transformation – IREN is moving away from Bitcoin mining and towards AI Cloud services, potentially changing how investors value the company.
• 6. Valuation – The stock has fallen significantly from its 52-week high, while the company continues to develop its AI infrastructure business.
• 7. Potential re-rating – If AI Cloud revenue grows as expected, the market may increasingly value IREN as an AI infrastructure company rather than a Bitcoin miner.

This is the asymmetry we are interested in.

The downside is real.

But if the company executes, the upside could be substantial relative to the current valuation.

12 · What could go wrong?

We need to be equally clear about the risks.

• Delays in data center construction and commissioning.
• Microsoft or other customers delaying acceptance.
• Lower GPU rental prices.
• Reduced AI infrastructure spending.
• Rising electricity and operating costs.
• Financing becoming more expensive.
• Additional share dilution.
• Customer concentration.
• Competition from CoreWeave, Nebius and hyperscalers building their own infrastructure.
• Failure to generate sufficient cash flow from deployed GPUs.

Perhaps the biggest risk is that IREN commits billions to infrastructure, but the economics of AI computing deteriorate before those investments generate the expected returns.

A large contract does not automatically mean a profitable business.

The company must deliver, operate and generate attractive returns on its invested capital.

That is what we need to monitor.

13 · November catalysts – what we are watching

The next major checkpoint is November 5, 2026, when IREN is scheduled to report its Q1 FY2027 results.

This report will be particularly important because it will provide an update on the company's progress following its FY2026 results.

Our focus:

• AI Cloud revenue – Is the business accelerating as expected?
• Operating ARR – How much of the contracted capacity has actually become operational?
• Microsoft Horizon – Are Horizon 2–4 progressing towards customer acceptance?
• NVIDIA deployments – Are GPU deployments progressing towards the expected 2027 revenue ramp?
• Financing – How much additional capital is required, and how are financing costs developing?
• Bitcoin exit – How quickly is the company transitioning away from mining?
• New customers – Can IREN diversify its customer base and secure additional multi-year contracts?

These are the metrics that will determine whether the investment thesis is strengthening or weakening.

Grabit Labs – our final verdict

We believe IREN represents one of the more interesting speculative opportunities in the AI infrastructure market.

The company has several important ingredients:

• Access to substantial electricity and infrastructure.
• A development pipeline exceeding 5GW.
• Multi-billion-dollar agreements with Microsoft and NVIDIA.
• Rapidly growing AI Cloud revenue.
• Financing structures designed to support GPU expansion.
• A business transition that could fundamentally change its valuation.

We prefer IREN over CoreWeave and Nebius as our speculative AI infrastructure selection at current relative valuations.

Not because the other companies lack potential.

But because we see a particularly interesting combination of physical infrastructure, power access, customer commitments and potential valuation re-rating in IREN.

The market is currently focused on the risks associated with its transformation.

We understand those risks.

But we believe the potential opportunity deserves attention.

IREN is not simply a bet on more AI spending. It is a bet on the infrastructure that makes that spending possible.

If the company successfully converts its power portfolio and contracted GPU capacity into profitable, recurring AI Cloud revenue, the market could eventually value the business very differently.

That is what we want to own exposure to.

GRABIT SCORE: 7.4/10 · INVESTMENT VIEW: SPECULATIVE BUY · TIME HORIZON: 12–24 MONTHS

We see an attractive opportunity, but position sizing matters. This is a high-volatility growth investment, not a defensive holding.

Our conviction depends on management delivering on its infrastructure commitments and converting contracted capacity into real operating revenue.

The next phase of the AI race may not belong to the companies with the most GPUs. It may belong to the companies that can actually power them.

This report is produced by Grabit Labs for informational and educational purposes only. It does not constitute financial advice or a personal investment recommendation. All investments involve risk, including the potential loss of invested capital. Forward-looking statements, revenue targets and contract values are subject to execution, financing, market and customer risks. Figures are based on information available through October 31, 2026, unless otherwise stated.