The IREN revenue miss is the least interesting number the company published this week. Wall Street spent Friday marking the stock down 12.53% to $35.45 because fiscal fourth-quarter revenue came in at $137.2 million rather than the roughly $157 million some models carried. That $20 million gap is a rounding error next to what IREN Limited disclosed on the same call: fiscal 2027 capital expenditure guidance of approximately $25 billion to $30 billion, against full-year FY2026 revenue of $707.0 million. IREN is telling the market it intends to spend between 35 and 42 times its annual revenue in a single fiscal year, and roughly twice its own $14.0 billion market capitalisation. This IREN stock prediction is therefore not a forecast about one quarter. It is a judgement on whether a company can finance that gap without destroying the per-share value of the equity that funds it.
Here is the part almost nobody has priced. IREN’s FY2026 headline revenue grew 41.1%, from $501.0 million to $707.0 million. Over the same year, adjusted EBITDA fell 8.9%, from $269.7 million to $245.7 million, and ordinary shares outstanding rose 47.3%, from 258,103,209 to 380,193,608, per the Form 10-K filed with the SEC on 27 August 2026. Divide revenue by the share count at each fiscal year end and the picture inverts: revenue per share went from $1.94 in FY2025 to $1.86 in FY2026, a 4.2% decline. The top line grew 41% and the shareholder’s claim on it shrank. That single arithmetic fact, not the missed consensus figure, is the honest starting point for any long-term view on IREN.
Key facts: IREN FY2026 in seven numbers
- Q4 FY2026 revenue $137.2m, down from $144.8m in Q3. AI Cloud Services contributed $70.5m and Bitcoin mining $66.7m, the first quarter in which AI revenue exceeded mining at 51.4% of the total. Source: IREN Limited FY26 results, SEC Form 8-K EX-99.1, 27 August 2026.
- FY2026 net loss of $702.6m versus net income of $86.9m in FY2025, including non-cash impairments of $638.8m, of which $450.4m landed in Q4 alone. Source: IREN FY26 results, 27 August 2026.
- FY2027 capex guidance of $25bn to $30bn, stated by CFO Anthony Lewis on the 27 August earnings call. Source: IREN Q4 FY2026 earnings call transcript.
- $4bn of contracted annualised run-rate revenue for 2026 capacity, of which approximately $1bn was operating as of 26 August 2026. IREN’s own footnote warns that “recognized revenue may be materially lower”. Source: IREN FY26 results.
- Long-term debt rose from $962.8m to $7,423.6m in twelve months, a 7.7-fold increase, against $5,895.6m of unrestricted cash at 30 June 2026. Source: IREN Form 10-K, FY ended 30 June 2026.
- 394,058,648 ordinary shares outstanding at 14 August 2026, up 52.7% from 30 June 2025, with roughly $3.5bn of a registered $6.0bn at-the-market facility still undrawn. Source: IREN Form 10-K.
- Consensus price target $78.31, range $43 to $131 across 17 analysts. The lowest published target sits 21.3% above the current share price. Source: StockAnalysis, 28 August 2026.
What actually happened in the FY2026 numbers
Strip out the noise and IREN’s fourth quarter did something structurally important: AI Cloud Services revenue more than doubled sequentially, from $33.6 million to $70.5 million, while Bitcoin mining revenue fell 40.0%, from $111.2 million to $66.7 million. For the first time, AI was the majority of the business. Total revenue slipped only 5.2% quarter on quarter because one line grew almost exactly as fast as the other shrank.
That is the mechanism people keep missing. IREN is not losing mining revenue to a weak Bitcoin price. It is deliberately switching mining hardware off to free grid capacity for GPUs, then writing that hardware down. The $638.8 million of FY2026 impairments is not a surprise loss; it is the accounting cost of a decision management made on purpose. Think of it the way a refinery operator thinks about a turnaround: you take the cash-generating unit offline, absorb a quarter of ugly numbers, and come back configured for a higher-margin product. The refinery analogy holds right up to the point where you notice the scale of the rebuild, and IREN’s rebuild is enormous.
Adjusted EBITDA is where the strain shows. It fell from $59.5 million in Q3 to $19.2 million in Q4, a 67.7% drop on a 5.2% revenue decline, because IREN nearly tripled headcount during FY2026 and added five C-suite appointments while the AI revenue ramp was still ahead of it. The company is carrying a full-scale operating cost base against a revenue line that has not yet arrived. That is what an investment year looks like, and it is also what a value trap looks like from the outside. The two are only distinguishable in hindsight, which is precisely why the bull and bear cases below sit so far apart. Readers who followed our earlier work on the sector will recognise the pattern from our AI data centre bear case, where backlog and debt grew faster than delivered capacity across the whole cohort.
“We started IREN with a simple observation: the digital world can scale almost instantly, but the physical world cannot,” said Daniel Roberts, Co-Founder and Co-Chief Executive Officer of IREN, in the results release. “Exponential AI consumption growth has fueled demand for compute capacity well beyond the available supply of infrastructure. IREN was built for this moment.”
The bull case: $92, and why the customer list is the evidence
The strongest argument for IREN is not the ARR headline. It is that the counterparties signing the contracts have done their own diligence, and their names are now on the record. During the quarter IREN delivered Horizon 1 to Microsoft, the first of four 50MW liquid-cooled deployments at its Childress site in Texas, and achieved NVIDIA Exemplar Cloud status on GB300 NVL72 systems. Horizon 2 is commissioning, with Horizons 3 and 4 in late-stage construction targeting delivery in Q4 2026. Alongside Microsoft, IREN disclosed recent signings with Cohere, Prometheus, Perplexity, Figure AI, Fal AI and Higgsfield AI, plus a new multi-year AI Cloud contract with an unnamed leading frontier AI lab.
Pricing is the second pillar, and it moved in IREN’s favour while the stock fell. “Recent three-year contracts are pricing in excess of $20 million per megawatt of IT load, paying back the compute investment in around 2 years,” said Kent Draper, Chief Commercial Officer, on the earnings call. “Active discussions are now at around $25 million per megawatt.” Customer prepayments now represent 45% to 55% of GPU capital expenditure, which materially reduces the equity IREN must raise per megawatt delivered. That is a genuinely different capital structure from the one the market applied to Bitcoin miners, and it is the same structural argument we set out in our coverage of Nebius and CoreWeave.
Financing is the third. Anthony Lewis, Chief Financial Officer, told analysts that “over the past 12 months, we have secured circa $19 billion in funding,” including a $3.6 billion investment-grade GPU facility supporting the Microsoft contract at a 6.0% weighted average rate, and $2.8 billion of new GPU financings for non-investment-grade customers. The largest of those is a $2.4 billion facility led by Blue Owl and Pacific Investment Management Company, as adviser to certain investors, at a 9.0% fixed rate for the Mackenzie air-cooled expansion. When lenders of that calibre underwrite a data centre build at 6%, they are expressing a view on contract quality that no retail investor can replicate.
How we get to $92. IREN targets cumulative delivery of approximately 0.8GW of IT load by the end of 2027, against roughly 0.3GW in 2026. Assume 60% of that 2027 capacity is contracted at the new $20 million per megawatt pricing rather than the blended rate embedded in today’s $4 billion figure: 480MW at $20 million gives roughly $10 billion of contracted ARR. Apply 4.5 times enterprise value to ARR, modestly above the 3.9 times the market pays IREN today on an enterprise value of about $15.5 billion, and you get a $45 billion enterprise value. Subtract net debt grown to roughly $6 billion as the build is financed, spread the remaining $39 billion of equity over roughly 425 million shares after a partial ATM draw, and the result is approximately $92 per share. That sits above the $78.31 consensus and well below the $131 street high, which is where a genuine bull case should sit.
The bear case: $16, and the arithmetic nobody on the sell side has published
Start with an uncomfortable observation. Across the 17 analysts compiled by StockAnalysis as of 28 August 2026, the lowest price target on IREN is $43, which is 21.3% above where the stock closed. There is no published bear case. Every covering analyst models a higher share price than the one the market is paying. When the entire sell side is positioned on one side, the useful work is on the other, so here is the arithmetic they have not shown.
First, the financing that funds the bull case has a running cost. Take only the two facilities IREN has quantified: $3.6 billion at 6.0% and $2.4 billion at 9.0%. That is $216 million plus $216 million, or $432 million of annual interest, against FY2026 adjusted EBITDA of $245.7 million. Before a single dollar of the remaining $2.8 billion programme or the roughly $8 billion of additional FY2027 financing the CFO flagged, the announced facilities alone carry an interest bill equal to 1.76 times last year’s entire adjusted EBITDA. The AI revenue has to arrive not merely to justify the multiple, but to service the debt that bought the GPUs.
Second, the ARR figure and the revenue figure are not the same species. IREN’s own footnote is explicit that ARR “is an operating metric, not a U.S. GAAP measure, and differs from, and is not a substitute for, GAAP revenue; recognized revenue may be materially lower.” The gap between $4 billion of contracted ARR and $707.0 million of FY2026 recognised revenue is 5.7 times, and the bridge across it is commissioning, testing and customer acceptance. Those are exactly the steps that slipped this quarter.
Third, dilution is not finished. IREN issued 47,165,838 shares under its new prospectus supplement for approximately $2.5 billion gross as of 14 August 2026, out of a registered $6.0 billion at-the-market programme. Roughly $3.5 billion of that authority remains unused. At $35.45 that is close to 99 million additional shares, or a further 25% of the current count, available to management without a further shareholder vote. Our earlier reporting on IREN’s $3 billion convertible notes sale flagged the direction of travel; the 10-K confirms how far it has run.
How we get to $16. Assume the $4 billion of 2026 ARR converts into materially lower recognised revenue, as IREN’s own footnote permits, and that 2027 contracting slips past the late-stage discussions management describes. Apply 2.0 times enterprise value to contracted ARR, roughly half today’s multiple and the sort of compression the market imposed on the whole neocloud cohort in early 2026. That gives a $8 billion enterprise value. Subtract current net debt of $1.53 billion, spread the remaining $6.5 billion of equity across the 394,058,648 shares outstanding at 14 August 2026, and you get approximately $16 per share. That is below the $22.63 twelve-month intraday low, and it should be: that low was struck when the share count was near 258 million and the balance sheet carried under $1 billion of long-term debt.
The long-term view: what has to be true by 2028
Because the fiscal year is the wrong unit for this business, the long-term question resolves into three testable conditions rather than a price target.
Condition one: the 2027 book closes at the new pricing. Management describes 2027 as being in “late-stage discussions with a range of new customers over a significant portion of capacity”. Roberts framed it deliberately on the call: “2027 and beyond is deliberate runway for us. This is capacity we are continuing to convert into a broader customer base.” Deliberate is a defensible word when demand exceeds supply and a worrying one when it does not. The observable test is simple. If IREN announces 2027 contracts at or above $20 million per megawatt over the next two quarters, the bull case is intact. If it announces volume at a lower price per megawatt, the entire $92 derivation collapses, because that figure is built on price, not just capacity.
Condition two: the physical build lands on schedule. IREN is developing at Sweetwater in Texas, Kiowa in Oklahoma, Bundey in Australia and Badajoz in Spain, with new liquid-cooled deployments planned at Mackenzie, Canal Flats and Prince George during 2027 to use power headroom at existing sites. The pipeline exceeds 5GW. Roberts is right that “power, land, data centers, these things take years to permit, finance, and build,” and that is genuinely IREN’s competitive moat, since grid interconnection queues in Texas and Oklahoma now run for years. It is also the risk: the moat that protects IREN from new entrants protects nobody from IREN’s own construction schedule. Horizons 3 and 4 are targeted for Q4 2026, and that delivery date is the single most checkable near-term claim in the entire release.
Condition three: power politics does not intervene. This is the tension the sector consistently under-weights. IREN’s expansion is concentrated in jurisdictions where large-load interconnection is becoming politically contested, and the regulatory conversation has shifted from whether data centres should be welcomed to who pays for the transmission they require. Texas Senate Bill 6, enacted in 2025, gives ERCOT authority to order loads of 75MW or more to curtail or switch to backup generation during grid emergencies, and requires facilities interconnecting after 31 December 2025 to build in remote-disconnect capability. That is manageable for a Bitcoin miner, which switches off by design and is paid to do so. It is a materially harder constraint for an AI cloud running contracted training workloads under a service level agreement, and IREN’s Childress, Sweetwater and Kiowa build-out sits squarely inside that threshold. Converting interruptible mining load into non-interruptible AI load does not just change IREN’s revenue mix. It changes its regulatory risk profile, and it does so at exactly the moment the company is adding fixed obligations to service $7.4 billion of long-term debt. The same dynamic is now shaping capital allocation across the sector, as Utility Dive reported on the Texas law and as our reporting on MARA’s own pivot rationale and Applied Digital has shown.
There is one further structural point worth making. IREN completed the acquisitions of Mirantis and Nostrum during FY2026, adding software and managed services and extending the platform into Europe. Managed services carry higher margins than raw GPU hours and, more importantly, they are harder for a customer to move. A tenant who has merely rented capacity leaves when a cheaper megawatt appears. A tenant whose orchestration layer runs on your software does not. If IREN’s long-term multiple is going to hold above the level the market awards commodity compute, the Mirantis integration matters more than any single contract announcement, and we covered the strategic logic when IREN acquired Nostrum in June.
What happens next: three predictions
One: the December quarter is where the ARR claim gets tested, not the September quarter. Management expects more than $4 billion of ARR operating by the December 2026 quarter, up from roughly $1 billion in late August. That requires Horizons 2, 3 and 4 to commission, test and gain customer acceptance inside four months. We expect at least one of those four deployments to slip into the March 2027 quarter, and we expect the market to treat that slip as far more significant than the $20 million revenue miss it punished this week, because a delivery slip moves the ARR bridge rather than a single quarter’s print.
Two: the equity raise arrives before the AI revenue does. With $25 billion to $30 billion of FY2027 capex guided, roughly $8 billion of additional GPU financing and prepayments targeted, and $3.5 billion of ATM authority already registered and unused, IREN has both the need and the mechanism. We expect meaningful ATM issuance during the December and March quarters, and we would treat any quarter in which the share count does not rise materially as a genuine positive surprise rather than the base case.
Three: adjusted EBITDA troughs in the September 2026 quarter and inflects from December. Q4’s $19.2 million reflects a full cost base against a partial revenue base, with mining revenue still falling as sites convert. Once Horizon 1 revenue is recognised for a full quarter and mining decommissioning annualises, the mix stops working against the margin. If adjusted EBITDA is still below $50 million in the March 2027 quarter, the bear case at $16 becomes the working assumption rather than the tail risk.
The uncomfortable conclusion is that both cases got stronger this week. The bear case was confirmed by the impairment, the EBITDA collapse and the share count. The bull case was strengthened by Microsoft’s acceptance of Horizon 1, by pricing moving from $20 million towards $25 million per megawatt, and by investment-grade lenders funding the build at 6%. A stock can fall 12.53% on a day when the long-term evidence improved, and that is roughly what happened. The question for a long-term holder is not whether IREN’s AI pivot works. On the evidence of the customer list, it is already working. The question is how many shares will exist by the time it finishes working. For broader context on how the compute supply chain is repricing, see our analysis of Nvidia and Vertiv.
Frequently asked questions
Did IREN actually miss revenue in Q4 FY2026?
Yes. IREN reported Q4 FY2026 revenue of $137.2 million against consensus estimates that ranged from roughly $136 million to $157 million depending on the compiler, which is why coverage of the print has been inconsistent. Full-year FY2026 revenue of $707.0 million was up 41.1% from $501.0 million in FY2025. The larger issue was the $702.6 million full-year net loss, driven by $638.8 million of non-cash impairments on decommissioned Bitcoin mining hardware.
What is the difference between IREN’s $4bn ARR and its $707m revenue?
Annualised run-rate revenue multiplies contracted GPU hourly pricing by 8,760 hours and adds storage and ancillaries. It is a forward operating metric for capacity that is contracted, not necessarily commissioned. GAAP revenue records what was actually delivered and recognised. IREN states directly that ARR “differs from, and is not a substitute for, GAAP revenue; recognized revenue may be materially lower.” The 5.7 times gap between the two is the central uncertainty in any IREN stock prediction.
Why is the IREN bear case below the 52-week low?
Because the share count is not the same. IREN traded at an intraday low of $22.63 in late August 2025, when roughly 258 million shares were outstanding and long-term debt was under $1 billion. At 14 August 2026 there were 394,058,648 shares outstanding and $7.42 billion of long-term debt. A similar enterprise value spread across 52.7% more shares, net of a larger debt load, produces a lower price per share. The $16 bear case is not a forecast of a smaller company; it is the same company divided differently.
How much dilution can IREN still issue?
IREN registered up to $6.0 billion of ordinary shares under its sales agreement and had issued 47,165,838 shares for approximately $2.5 billion gross as of 14 August 2026. That leaves roughly $3.5 billion of registered capacity, equivalent to close to 99 million shares at the 28 August price of $35.45, or around 25% of the current share count.
Is IREN still a Bitcoin mining company?
Not by revenue mix. In Q4 FY2026, AI Cloud Services generated $70.5 million against $66.7 million from Bitcoin mining, the first quarter in which AI was the majority. Across the full year mining was still dominant at $578.2 million of $707.0 million, but management is deliberately decommissioning mining hardware to release grid capacity for GPU deployments, which is what produced the $638.8 million impairment.
What should investors watch next?
Three things, in order: whether Horizons 2, 3 and 4 achieve customer acceptance by the December 2026 quarter as guided; whether new 2027 contracts are signed at or above $20 million per megawatt of IT load; and whether adjusted EBITDA inflects from the $19.2 million Q4 level. The first two determine whether the $92 bull case is reachable. The third determines how much dilution is required to get there.
This article is for information only and is not investment advice. The $92 bull case and $16 bear case are FinanceFeeds estimates derived from IREN’s own filings and stated guidance, not price targets. Figures are as of the 28 August 2026 close. Primary sources: IREN Limited FY26 results, SEC Form 8-K Exhibit 99.1; IREN Limited Form 10-K for the fiscal year ended 30 June 2026; IREN Q4 FY2026 earnings call transcript; StockAnalysis analyst forecast data.