Here is the misconception surrounding IREN:
The company is either a Bitcoin miner pretending to be an AI company, or it is the next great AI infrastructure winner.
Both descriptions are too simplistic.
IREN is attempting something far more interesting.
It is converting a portfolio of power-connected data-center sites, originally built for Bitcoin mining, into infrastructure capable of hosting some of the most valuable computing equipment in the world.
That transformation could create enormous value.
It could also consume enormous amounts of capital.
That is the tension.
At approximately $47.05 per share as of September 23, 2026, investors are no longer being asked whether IREN has a credible AI business. Microsoft and several AI companies have already answered that question.
The question now is whether IREN can turn contracted demand into durable cash flow without giving too much of the upside away through debt, dilution and relentless capital spending.
That is the investment case.
IREN's real asset is not Bitcoin
IREN was built as a Bitcoin miner, but Bitcoin was never its most strategically valuable asset.
The scarce asset was always the infrastructure underneath the miners:
- Large blocks of secured power
- Land near electrical infrastructure
- Grid interconnections
- Data-center development experience
- The ability to operate extremely power-dense equipment
Bitcoin mining was simply the first way IREN monetized that infrastructure.
AI offers a potentially more valuable use.
A Bitcoin miner uses power to solve cryptographic puzzles. An AI cloud provider uses that same power to rent advanced computing capacity to Microsoft, AI laboratories and enterprise customers.
Same electrons.
Much higher-value workload.
That is why IREN's transition matters.
The Microsoft contract changed the story
In November 2025, IREN signed an approximately $9.7 billion agreement to provide Microsoft with dedicated NVIDIA GB300 infrastructure over an average five-year term.
The deployment covers four 50 MW Horizon facilities at IREN's Childress campus, representing approximately 200 MW of IT capacity. Microsoft agreed to prepay 20% of each tranche, while IREN estimated that the associated GPU equipment would require approximately $5.8 billion of capital expenditure. IREN's Microsoft filing
Why does that matter?
Because Microsoft was not buying a PowerPoint presentation.
It was contracting for physical infrastructure with specific GPUs, delivery schedules and performance requirements.
That validates IREN's ability to compete for hyperscaler workloads.
But a signed contract is not the same thing as delivered revenue.
Microsoft retains remedies if IREN misses delivery deadlines, and each tranche must pass an acceptance process.
That is the part investors cannot ignore.
The contract validates demand.
Execution determines the economics.
The gap that matters most
IREN currently reports approximately $4 billion of contracted annualized run-rate revenue associated with 2026 capacity.
But only around $1 billion of ARR was operating at the time of its August update. The company also said that its 2026 capacity was largely sold out. IREN's FY2026 update
That sounds spectacular.
And it is.
But the $3 billion gap between contracted ARR and operating ARR is the entire investment thesis.
Facilities must be completed.
GPUs must arrive.
Cooling systems must work.
Customers must accept the deployments.
Revenue must begin.
Margins must survive depreciation, financing costs and eventual hardware replacement.
Until that happens, contracted ARR is evidence of demand, not evidence of completed economic value.
A megawatt is not a data center.
A contract is not cash flow.
The historical numbers still belong to the old IREN
For fiscal 2026, IREN generated approximately $707 million of total revenue.
Bitcoin mining still produced $578 million of that amount, while AI Cloud Services contributed approximately $129 million.
AI revenue grew nearly eightfold, but the historical income statement still looked much more like a Bitcoin miner than a hyperscale AI cloud company.
IREN also reported a $703 million net loss, although that included approximately $639 million of largely noncash impairments, primarily associated with decommissioning Bitcoin-mining equipment.
Why does that matter?
Because the loss does not necessarily mean the new AI strategy is failing.
It does show how expensive the transition is.
The company is writing down yesterday's equipment while simultaneously borrowing and issuing shares to purchase tomorrow's equipment.
Investors are financing both sides of the transformation.
The financing is clever, but it is not free
IREN arranged approximately $3.6 billion of GPU financing at a 6% rate for the Microsoft deployment. Combined with Microsoft prepayments, the company says this covers approximately 96% of the associated GPU capital expenditure.
For other deployments, IREN arranged approximately $2.8 billion of GPU financing, including $2.4 billion priced at a 9% fixed rate. Those facilities are expected to finance approximately 90% of the relevant GPU spending.
This is constructive because it reduces the amount IREN must finance directly with common equity.
A 9% fixed rate is a significant financing cost.
And financing the GPU does not eliminate the cost of buildings, electrical systems, cooling infrastructure, networking equipment and future hardware replacement.
IREN ended June with approximately $5.9 billion of unrestricted cash and $7.6 billion of debt. It also disclosed roughly $13.8 billion of capital commitments. IREN's 2026 annual report
The company is well funded.
It is also exceptionally capital intensive.
Both can be true.
Dilution is not a side issue
IREN had approximately 258 million ordinary shares outstanding in June 2025.
By August 14, 2026, it had approximately 394 million.
That is an increase of roughly 53% in little more than a year.
Some of that capital helped create the AI opportunity investors are now buying. Raising equity is not automatically bad when the return on the new capital exceeds its cost.
But company growth and shareholder value are not the same thing.
Revenue can rise.
EBITDA can rise.
The share price can rise.
And existing owners can still capture less of the business than they expected if the share count keeps expanding.
For IREN, per-share execution matters more than corporate-scale execution.
What is the market already assuming?
At $47.05 and approximately 394 million shares, IREN's equity value is roughly $18.5 billion. Adding June debt and subtracting unrestricted cash produces a simplified enterprise value of approximately $20 billion. IREN price history
If the mature business eventually deserves an enterprise-value-to-EBITDA multiple of 10 to 12 times, the current valuation requires something close to $1.7 billion to $2 billion of sustainable annual EBITDA.
At a 35% to 40% EBITDA margin, that would require approximately $4.3 billion to $5.7 billion of revenue.
That is important.
The current price does not require IREN to fill its entire 5 GW development pipeline.
But it does require the company to convert most of its existing contracted AI capacity into real revenue at attractive margins.
The stock is not pricing in failure.
It is pricing in meaningful execution.
Three possible outcomes
These are analytical values, not price targets.
The bear case: roughly $14 per share
In this scenario:
- 2028 revenue reaches only $4 billion.
- Delivery delays, customer concentration and lower utilization hold the EBITDA margin near 28%.
- The market applies a 9 times EBITDA multiple.
- Net debt rises toward $4 billion.
- Diluted shares increase to approximately 450 million.
The resulting equity value is approximately $14 per share.
This is what happens if IREN builds aggressively but the per-share economics disappoint.
The base case: roughly $59 per share
In this scenario:
- 2028 revenue reaches approximately $7 billion.
- The EBITDA margin improves toward 38%.
- The business receives an 11 times multiple.
- Net debt is approximately $4 billion.
- Diluted shares rise more moderately to around 430 million.
That produces an analytical value close to $59 per share.
This outcome assumes successful execution, but not perfection.
The bull case: roughly $130 per share
In this scenario:
- 2028 revenue reaches approximately $10 billion.
- Strong pricing and utilization support a 45% EBITDA margin.
- The market awards a 13 times multiple.
- Net debt remains manageable at approximately $4 billion.
- Dilution is contained near 420 million shares.
That produces an analytical value around $130 per share.
This requires IREN to become a major AI cloud operator, not merely a data-center developer with several large contracts.
What could break the thesis?
The obvious risk is construction.
A facility delivered six months late is not simply six months of delayed revenue. It can trigger contractual remedies, raise financing costs and leave expensive GPUs underutilized.
The second risk is technological depreciation.
Today's premium GPU can become tomorrow's previous-generation GPU surprisingly quickly. Accounting depreciation schedules do not determine economic value. Customer demand does.
The third risk is capital allocation.
IREN has shown that it can raise money. Now it must show that each new dollar of debt and equity creates more than one dollar of shareholder value.
The fourth risk is customer concentration.
A large hyperscaler contract is powerful validation, but it also creates dependence. A healthier customer mix would include more hyperscalers, frontier laboratories and enterprise customers.
What to watch out for
Watch the conversion of contracted ARR into operating ARR.
Watch the acceptance and revenue commencement of Horizon 2, Horizon 3 and Horizon 4.
Watch AI Cloud revenue per deployed megawatt.
Watch EBITDA after corporate overhead, not just project-level economics.
Watch the relationship between debt, diluted shares and operating ARR.
Watch how much capital must be reinvested when the current generation of GPUs becomes less competitive.
Most importantly, watch whether each new phase creates more value per share, not merely more revenue for the company.
My conclusion
IREN is no longer just a Bitcoin miner with an AI presentation.
The contracts are real.
The customers are real.
The infrastructure is being delivered.
But at $47, the market already expects much of that contracted opportunity to become operational.
That makes IREN compelling, but speculative.
The upside could be substantial if management delivers capacity on time, maintains attractive margins and contains dilution. The downside could also be severe if the company mistakes access to capital for economic value creation.
The principle is simple:
Do not value IREN by the number of gigawatts it controls. Value it by the cash flow each diluted share ultimately owns.