IREN: Priced for an AI Company That Does Not Exist Yet

The Setup

IREN Limited, the company formerly known as Iris Energy, sits at the intersection of the two most exciting narratives in the market: Bitcoin and artificial intelligence infrastructure. Its shares have climbed from single digits in early 2025 to a 52-week high of $76.87, and the stock now changes hands near $66. Wall Street is overwhelmingly positive. The consensus rating is Buy, and price targets reach as high as $99 at Cantor Fitzgerald, with a published high estimate around $136.

This is not a piece about whether artificial intelligence infrastructure matters. It does. Nor is it an argument that IREN is a fraud or that its assets are worthless. They are not. This is a narrower and more uncomfortable argument: the price already assumes IREN has become something it has not yet become, and the gap between the story being sold and the financials being reported is large enough to matter.

The market is paying for a fully realized AI infrastructure operator on the scale of an Equinix or a Digital Realty. The most recent quarterly filing describes a company whose revenue is shrinking, whose losses are widening, and whose AI business is still a rounding error against its legacy Bitcoin mining operation. Both of those things are true at the same time. The question is which one the price reflects.

The Pivot Story

The bull case is genuinely compelling, and it deserves a fair hearing before the numbers complicate it. IREN built a portfolio of grid-connected land and power in renewable-rich regions, the exact resource that has become the binding constraint on artificial intelligence. As IREN co-CEO Daniel Roberts put it on the most recent earnings call, the world is structurally short compute, and the bottleneck is delivered data center and GPU capacity. IREN has the power. That is not nothing.

The marquee validation arrived in late 2025: a five-year, $9.7 billion AI cloud agreement with Microsoft (MSFT) to deploy NVIDIA (NVDA) GB300 GPUs at IREN's Childress, Texas campus, worth roughly $1.94 billion in annual revenue. NVIDIA itself took a position, receiving a five-year right to purchase up to 30 million shares at $70, a stake worth as much as $2.1 billion if exercised. IREN is building toward a 5 gigawatt global data center pipeline, energizing its Sweetwater facility with 1.4 gigawatts of capacity and targeting 140,000 GPUs deployed by the end of 2026.

If you stop reading there, IREN looks like one of the cleanest ways to own the AI buildout. A power-rich operator, an anchor contract with one of the largest software companies on earth, and an endorsement from the chipmaker at the center of the entire boom. That is the story the price is built on. The rest of this article is what happens when you open the financials.

What the Revenue Actually Shows

IREN reported results for its fiscal third quarter, the three months ended March 31, 2026, on May 7. The headline numbers did not match the narrative. Total revenue came in at $144.8 million, down 22% from $184.7 million the prior quarter, and missing the analyst estimate of roughly $220 million by 34%. The net loss widened to $247.8 million from $155.4 million, and the loss of $0.30 per share missed expectations by nearly 39%.

The composition of that revenue is the part that does not get repeated in the headlines. AI Cloud services, the entire reason the stock is valued the way it is, generated $33.6 million in the quarter. That figure grew an impressive 94% from the prior quarter, and the growth rate is the number every bullish writeup leads with. But $33.6 million is only about 23% of total revenue. The other 77%, roughly $111 million, still came from Bitcoin mining, and that business is going the wrong way: mining revenue fell from $167.4 million as the company decommissioned older hardware.

So the overall revenue decline is not an accident or a one-time charge. It is the mechanical result of the pivot itself. The shrinking business (mining) is still far larger than the growing business (AI cloud), so the company gets smaller in aggregate even as the part everyone cares about expands. That can absolutely reverse if the AI cloud ramp delivers on schedule. But as of the last reported quarter, the company being valued as an AI infrastructure leader was, by revenue, still three-quarters a declining crypto miner.

The $3.7 Billion Question

The single most important number in IREN's investor materials is not on the income statement. It is the $3.7 billion in annualized recurring revenue the company projects, a figure that turns a business currently doing about $580 million in trailing revenue into something that could plausibly justify a $24 billion valuation. That number is doing almost all of the work in the bull case. It is worth reading exactly how the company itself describes it.

Per IREN's own filing, the $3.7 billion ARR figure combines the expected $1.9 billion average annual revenue from the Microsoft contract with an estimated $1.8 billion from a roughly 74,000-GPU deployment at its British Columbia and Childress sites, and is based on internal company assumptions regarding GPU models, utilization, and pricing. The company states plainly that the figure is not fully contracted, that there can be no assurance it will be achieved, that actual revenue may differ materially, and that it assumes on-time delivery and commissioning of GPUs.

That is not a criticism of IREN's disclosure. The company is being appropriately transparent. The criticism is of how the market is treating the number. A self-reported projection that depends on assumed pricing, assumed utilization, and flawless hardware delivery is being priced as though it were contracted, de-risked revenue already flowing through the door. It is not. The contracted piece is the Microsoft deal. The larger, more speculative piece rests on demand and execution that have not yet happened.

Funding the Dream

Building 5 gigawatts of data center capacity and deploying 140,000 GPUs is extraordinarily expensive, and the cash to do it is not coming from operations. In the March quarter alone, IREN spent roughly $540 million on property, plant, and equipment. The company reported $2.6 billion in cash at the end of April, a real and meaningful cushion. But the more telling story is how quickly that cushion is being topped up from the capital markets.

In May, IREN priced and closed $3.0 billion in convertible senior notes (including the greenshoe), carrying a 1.00% coupon and a conversion price around $73.07, netting roughly $2.96 billion. In June, it closed a separate $3.65 billion investment-grade GPU financing facility tied to the Microsoft deployment. It also agreed to buy NVIDIA Blackwell systems from Dell (DELL) for about $1.6 billion. In the span of a few weeks, IREN arranged well over $6 billion in new debt and financing to fund a buildout that the income statement is not yet paying for.

None of that is inherently reckless. Financing a contracted, cash-generative asset with debt is exactly what infrastructure operators do. The risk is the sequencing. IREN is taking on billions in obligations and dilution risk today, against revenue that is still largely projected rather than realized. The convertible notes convert into stock above roughly $73, which adds dilution if the thesis works, and they remain debt to be repaid if it does not. The buildout has to land close to plan, close to on time, for the math to hold. The valuation leaves very little room for it not to.

What the Wealth Engine Scores Say

Before the valuation verdict, here is what the Wealth Engine Pro platform's systematic scoring shows for IREN right now.

The platform rates IREN as Weak with a Bearish outlook and a blended fair value of $6.79, a figure that would require the stock to fall roughly 90% from its current level near $66. Put another way, the market is paying close to ten times the platform's estimate of fair value. The low Moat score (4 of 15) and modest Growth score (6 of 15) reflect a capital-intensive business with thin durable advantages and a revenue base that is currently contracting.

These scores are systematic. They evaluate companies based on reported financials, balance sheet quality, moat characteristics, and valuation models (discounted cash flow, peer comparison, earnings power). They measure what a company is today, not what it might become. That is by design: the scoring system is built to keep emotion and forward speculation out of the numbers.

In this case, the editorial thesis and the platform scores point in the same direction. The article's argument is precisely that the price has run far ahead of the realized business, and the systematic data, which cannot price an unproven buildout until the revenue and earnings show up in filings, says the same thing from a different angle. When both the quantitative model and the qualitative analysis flag the same gap between price and substance, that convergence is worth paying attention to. Research IREN yourself on the platform and decide which signal matters more for your situation.

The Valuation Verdict

Strip away the labels and the valuation question is simple. IREN carries a roughly $24 billion market cap on trailing revenue of about $580 million, revenue that declined last quarter, and a widening net loss. The entire bridge from that reality to the valuation is the projected $3.7 billion ARR, a number the company itself labels as uncontracted and assumption-dependent.

There is a respectable bull-case math here, and it is worth stating fairly: if IREN hits $3.7 billion in high-margin recurring revenue, the stock is not expensive at all, and today's price will look like a bargain. But that is the point. The price does not reflect a probability-weighted path to that outcome. It reflects the outcome itself, arriving on schedule, at the assumed pricing and utilization, with the financing fully in place and the GPUs commissioned on time. Every one of those is a real variable, and the company has already missed revenue estimates by a third in the most recent quarter.

That is the disconnect the data keeps surfacing. The buyer at $66 is not paying for what IREN earns. The buyer is paying for what IREN has promised to earn, on the assumption that the promise converts cleanly into cash. When a valuation prices the best case as the base case, the risk is asymmetric in the wrong direction.

What Could Go Wrong

This is an avoid thesis, which means the honest work is steelmanning the case that it is wrong. There is a real version of the bull argument, and it is stronger than most avoid candidates we cover.

The Microsoft contract is real and large. Unlike the speculative portion of the ARR figure, the $9.7 billion Microsoft agreement is signed. If it ramps as planned, it alone provides a substantial, contracted revenue floor that the trailing financials do not yet capture. The market may simply be pricing that forward, which is what markets are supposed to do.

The balance sheet can fund the build. Between $2.6 billion in cash, the $3.65 billion GPU facility, the $3.0 billion in convertible notes, and NVIDIA's backing, IREN has assembled the capital to execute. Plenty of growth stories die from lack of funding. This is not obviously one of them.

Power is a genuine moat. The binding constraint on AI is not chips alone, it is grid-connected power, and IREN's secured capacity across multiple regions is a real asset that is hard and slow to replicate. If the compute shortage persists, that capacity becomes more valuable, not less.

The crypto optionality is free. The legacy mining business, declining as it is, still throws off cash, and a sharp Bitcoin rally would add upside the AI thesis does not even require.

The thesis breaks if IREN executes the buildout close to on time and the AI cloud revenue ramps toward that $3.7 billion figure. In that world, the forward multiple compresses quickly and the stock grows into its price. What the avoid thesis argues is not that this is impossible. It is that the current price already assumes it, leaving little reward for the buyer if everything goes right and significant downside if the timeline slips, the pricing disappoints, or demand softens. That is the same pattern we have examined in The Quantum Computing Bubble and The SaaS Reckoning: a real technology, a real opportunity, and a price that has detached from the part of the story that has actually been proven.

The Bottom Line

IREN is not a bad company. It owns valuable power capacity, it signed a real anchor customer, and it has the financing to chase a large opportunity. If the buildout lands on plan, the bulls will be right and this article will age poorly. That outcome is possible. It is simply not what the current price represents, because the current price represents that outcome as already achieved.

The most recent quarter showed revenue falling 22%, a net loss widening to $247.8 million, and a business that is still three-quarters declining Bitcoin mining. The number carrying the valuation, $3.7 billion in ARR, is described by the company itself as uncontracted and assumption-dependent. The platform's systematic scoring reads Weak, Bearish, and Expensive, with a fair value roughly 90% below the market price. Every independent angle lands in the same place.

That is the entire philosophy here. Wealth Engine Pro evaluates companies on what they are, not on what the most exciting version of the story hopes they will become. The narrative says IREN is an AI infrastructure leader. The data says it is a money-losing miner in the middle of an expensive, unfinished transition, priced as though the transition were already complete. When the gap between those two pictures is this wide, the data is the part worth trusting until the financials prove otherwise. For now, the numbers do not support the price.