Summary IREN Limited is transitioning from infrastructure build-out to execution, with a Buy rating maintained as revenue growth now hinges on delivery. AI Cloud Services revenue surged 6x YoY to $17.3M in Q2 FY26, with 86% gross margins, but future growth depends on converting contracted capacity to utilization. Microsoft's $9.7B five-year deal underpins visibility, but revenue recognition is tied to deployment acceptance; capital needs are largely prearranged. Valuation is anchored on IREN's ability to scale revenues from $1B to $3B, with execution timing and customer concentration as key risks. Elevator Thesis It appears the market is beginning to re-rate IREN Limited ( IREN ) as a part of the outsourced AI infrastructure market. My earlier piece covered IREN's foundation, its scale, structure, and partnerships. Interestingly, this perspective remains, but it has come into the prove-it phase. Here, execution will drive the pace of the revenue growth. Basically, now the focus moves from laying the groundwork to delivering on it. The stock has grown by 22.5% to around $50 (at the time of writing) since my last analysis, giving it about $16.8 billion in market capitalization. This is in line with the broader industry positioning, where companies such as CoreWeave ( CRWV ) and Nebius ( NBIS ) are external capacity providers to hyperscalers. CoreWeave has $5.1 billion in 2025 revenue , and Nebius closed 2025 with $1.25 billion ARR and is projecting $7-$9 billion in FY 26. Crucially, demand is determined by hyperscalers' capex rather than IREN's sales. For some color, Amazon Web Services is growing fast, with Amazon anticipating about $200 billion of capex in 2026, much of it driven by AI. Whereas Alphabet ( GOOG , GOOGL ) expects $175-185 billion . Microsoft ( MSFT ) adds further weight. It has $625 billion in commercial remaining performance obligations and 39% growth in Azure in FY26 Q2. Reuters also reported AI-related capacity constraints will extend to mid-2026. This means that the debate has changed now. It's not about getting land, power, or GPUs anymore. The first clear signal came in Q2 FY 26 . IREN's revenue from the AI Cloud Services jumped to $17.3 million from merely $2.7 million last year. That's stellar growth, but it also shows that things are still very early. All that revenue was from Canada. The much larger Microsoft cluster in Childress has yet to be a significant revenue source, and that's important. I'm maintaining a 'Buy' rating on IREN. Financing and an infrastructure base are in place, and Microsoft's prepayments support visibility. Furthermore, financing of GPUs at less than 6% reduces pressure, and supply alignment with Dell ( DELL ) and NVIDIA ( NVDA ) also minimizes risk. Hence, it's now a matter of conversion. Usage, timing, and cost of capital will determine revenues and cash flow. Q2 FY26: Early AI Monetization, Bitcoin Still Core Q2 FY 26 was pretty interesting, to say the least. Total revenue and costs (IREN Q2 deck) Revenue from its AI Cloud Services surged more than 6x to $17.3 million. This means that on the surface, growth is accelerating, but the real value is in the costs. The direct AI Cloud cost of revenue was just $2.4 million, so that gives a segment gross contribution margin of roughly 86%. It's this number that's really important, as the efficiency with which it's converting is notable. Yes, AI is still relatively small, less than 10% of total group revenue this quarter. That said, it already accounts for more than 12% of the segment's gross profit. So even at this point, it's already kind of delivering above average. Meanwhile, Bitcoin mining is driving the business. The revenue from mining was $167.4 million compared to $17.3 million for AI. However, the economics are very different. The company is selling all of its Bitcoin's daily production, so this part of the business is sensitive to Bitcoin prices, difficulty changes, and power costs. On the same direct cost basis, the mining margin is around 62%, compared to AI's 86%. This gap matters because even a gradual change in mix will start to alter the overall profit profile. So, what we are actually seeing is just the first wave of monetization. The real catalyst is yet to come, particularly as the Microsoft deal converts into revenue down the line. That's a whopping $9.7 billion five-year deal with a 20% prepayment, and it is structured across four tranches of deployment in 2026 connected to 200 MW of load. As of now, none of these tranches have been delivered and accepted, so future revenue is not in the current RPO yet. So yes, earnings look a bit confusing, but it's just timing, build-out, acceptance, and ramp. Capital is Solved, Deployment is Not IREN CEO Daniel Roberts (Livewire Markets) IREN can't be considered as a single source of capacity anymore. Let's break it down into what's already live versus what's only contracted or under construction. IREN had either 99,900 GPUs installed or on order by the end of 2025. Moreover, it announced another substantial tranche of more than 50,000 Nvidia B300 GPUs last month. So, the planned capacity reaches closer to 150,000 GPUs, with management's internal forecast referring to an AI Cloud run rate of over $3.7 billion by the end of 2026. Well, that's where the difference comes in. Planned capacity is not really deployed capacity, and it's actually deployed capacity that generates revenue. It's the same for power. The 4.5GW of secured capacity is a good start, but not all of it is near-term monetizable. Oklahoma provides longer-term optionality, but it only starts from 2028. So, for 2026 and 2027, the earnings bridge is still anchored in British Columbia, Childress, and the next energization phases at Sweetwater. Sweetwater 1 substation energization for Q2 2026 is a critical milestone because it is the start of commercial readiness. From a capital perspective, the structure is very sound. On one hand we have $3.6 billion of GPU financing (at less than 6% rates), and on the other we got a $1.9 billion Microsoft prepayment. This essentially covers around 95% of the GPU-related capex for the Microsoft build and takes strain away from the balance sheet. By early February, total funding was over $9.2 billion. Then the March GPU order was on post-shipment payment terms, backed up by another $3.5 billion of expected H2 2026 funding. In simple words, funding is mainly pre-arranged. Under this setup, Dell ships Microsoft orders across four tranches in 2026 for due payment within 30 days after shipment. But revenue is only finalized upon acceptance. This clearly causes a timing issue. Depreciation starts before the revenues are seen if products are shipped faster than clusters are scheduled and filled. If deployments and utilization are in sync, operating leverage will grow very fast. Q3 FY26: What to Expect vs. What Actually Matters IREN will report its Q3 results next month. I'm anticipating a marginal sequential increase in revenue in the $190-$220 million range. It would be largely powered by Bitcoin mining. Additionally, revenues from the AI Cloud should keep growing. However, in dollar terms, it is likely to be modest, probably around $25-40 million. On the bottom line, it will continue to be under pressure, probably with an EPS in the range of -0.20 to -$0.30 as depreciation and finance expenses outpace revenue. Well, the topline does not fully reflect the situation. It's the trends that we need to look at. I'd expect AI gross margins to remain healthy, likely in the 80-85% range, which would be a further part of the structural tailwind. If AI revenue is strong quarter over quarter and margins don't suffer, it means that early monetization is on track. Conversely, if costs increase more rapidly than revenue, profits might look ugly before they get better. Overall, I don't think Q3 alters my thesis, but it would test it. I will be focused on any guidance on Childress acceptance because that's where the next wave of revenue is expected to come from. Notably, the Childress deployment is not expected to have a material impact on this quarter's revenue (due to acceptance timing), despite shipments and installations. So, Q3 is really about whether or not IREN is on schedule to convert funded capacity to revenue in the coming quarters. Valuation Discussion Now here's when things get interesting. At the current price, IREN is not valued on short-term earnings. It is valued on a single assumption that infrastructure build-out will immediately translate into ongoing high-margin monetization of AI. IREN consensus estimates (Seeking Alpha) Let's assess the snapshot above. Analysts expect revenue at approximately $1.01 billion and $2.94 billion in FY 26 and FY 27, respectively. This drops the anticipated forward price-to-sales (P/S) multiple from 16.7x to 5.7x. To put it simply, today's output is not being priced by the market. It is pricing the ability to turn secured power and GPU capacity into future cash flows. The short-term outlook is different for earnings per share. Consensus estimates are around $0.38 in 2026, followed by a swing to negative $0.45 in 2027. This shows that the company is still in the buildout phase. Depreciation, funding costs, and growth are impinging on profitability. At this point, the value is not based on earnings. It is anchored to the execution and capacity to generate revenues. Let's see how the market is pricing this. IREN is commanding a huge premium to the sector, even on forward numbers. The forward EV/Sales is 17.7x , versus about 3x sector median, and the EV/EBITDA is 37-38x versus ~14x median. That's indeed a statement, and it's not accidental either. The market seems ready to pay a premium for what it perceives to be scarce, scaled AI infrastructure. It's especially true when it is backed by guaranteed power, rapid data center build-out, and direct access to GPU deployment. IREN moving averages (Seeking Alpha) Additionally, the stock has been consolidating on the technical side after its earlier run. It is currently in the low to mid $50s and is above all the major moving averages. The structure looks like a longer-term uptrend to me, although the momentum has slowed down a little. The recent move back above the shorter-term averages also looks like strength is re-emerging after consolidation, rather than breaking down. What really ties the whole thing together is pretty simple. Market is not pricing IREN on how profitable it is now. It is valuing the speed at which it can grow from $1 billion to $3 billion in revenues while still enjoying the infrastructure advantage. This is why the multiples are so high. In terms of price, I expect the base case scenario to be near the $60-$80 range if execution is in line with consensus and multiples begin to normalize as scale increases. Meanwhile, if execution outperforms, with a clean deployment and ramp-up of utilization, the upside moves up. In that case, the bull scenario goes to $85-$105. This is supported by ongoing scarcity in AI infrastructure and the faster monetization of the capacity deployed. Risks to My Thesis and Takeaway For starters, timing of execution remains the biggest risk. IREN is already pointing to several areas where delays can be seen. It includes the availability of GPUs, the buildout of data centers, liquid cooling deployment, grid connections, and the wide interconnection process at scale. Notably, the Microsoft deal has specific delivery schedules and even termination rights for late delivery, though with cure periods. So, the investment case is dependent on whether funded and deployed capacity actually transforms into billable usage on time. Any delay here not only impacts revenue but also sentiment. The second layer of risk is more systemic. Customer concentration is significant here, with Microsoft being the key anchor contract. This helps with visibility but also increases reliance on a single counterplay. There is also ongoing dilution risk. With convertibles, equity offerings, and ATM programs already being part of the funding mix, scaling has been achieved, but there is still equity overhang. Last but not least, we have margin pressure to think about. Competition is increasing in the AI infrastructure sector. CoreWeave is growing rapidly, and Nebius is also pushing ARR targets. If the capacity supply outpaces the power to raise prices, the return on incremental capex will return to normal faster than expected. All in all, demand is clearly there and isn't an issue anymore. Now the focus is fully on execution. The question is how well IREN translates GPUs, contracted capacity, and capex into sustained utilization before the costs catch up. Even so, the setup remains favorable. Demand is contracted, capital is secured largely, and the revenue inflection is coming up.
