Saul Rosenthal | February 22, 2026
IREN is the most unusual company I've looked at in years. Not SaaS. Not software. A Bitcoin miner turned AI infrastructure company, mid-pivot, with a $9.7 billion Microsoft contract and $2.3 billion of contracted ARR sitting above $184 million of quarterly revenue. That gap — between what's contracted and what's flowing through the income statement — is the whole story here.
The headline numbers look terrible. Revenue fell 23% sequentially. GAAP net loss of $155 million. Free cash flow of negative $648 million. A stock picker looking only at trailing results would run away screaming.
But I follow the money. And the money tells a very different story.
I'll be transparent: this is NOT a typical Saul-type investment. I don't own it. It's in wsm007's portfolio at 5.7%. I'm analysing it because the numbers are extraordinary — and I want to understand whether this is a genuine inflection or a mirage.
| | Q1 FY25 | Q2 FY25 | Q3 FY25 | Q4 FY25 | Q1 FY26 | Q2 FY26 | | | Sep-24 | Dec-24 | Mar-25 | Jun-25 | Sep-25 | Dec-25 | |---|---|---|---|---|---|---| | Revenue ($m) | 113.4 | 116.3 | 148.1 | 188.9 | 240.3 | 184.7 | | YoY % | +172% | +172% | +172% | +229% | +342% | +59% | | QoQ % | +5% | +3% | +27% | +28% | +27% | -23% | | Gross margin [GAAP] | 39.6% | 46.5% | 50.7% | 66.4% | 72.2% | 64.4% | | Adj EBITDA ($m) | 2.5 | 62.4 | 82.9 | 121.9 | 91.7 | 75.3 | | Adj EBITDA margin | 2% | 54% | 56% | 65% | 38% | 41% | | GAAP Op margin | — | — | — | 11.0% | — | -63.0% | | FCF (m)|−387|—|—|—|−138|−648||AIcloudrev(m) | — | — | — | 7.0 | 7.3 | 17.3 | | BTC mining rev (m)|—|—|148.1|180.3|232.9|167.4||ARRcontracted(b) | — | — | — | — | — | 2.3 | | Operating capacity (MW) | — | — | — | — | — | 810 | | Cash ($b) | — | 1.0 | — | — | — | 3.3 |
Revenue fell 23% sequentially. I always pay attention to revenue deceleration — always. But I also pay attention to context.
Here's what happened: management deliberately reduced Bitcoin mining hashrate to reallocate power capacity toward AI cloud. Bitcoin mining revenue fell from $232.9M to $167.4M — down $65.5M sequentially, which is almost exactly the total revenue decline. AI cloud revenue, meanwhile, accelerated from $7.3M to $17.3M, up 137% sequentially.
Is this a cover story for operational weakness? I don't think so. The sequential AI cloud trajectory — near-zero a year ago, $7.0M, $7.3M, now $17.3M — is genuinely accelerating. And the contracted ARR of 2.3billionagainsttrailingtwelve − monthrevenueof 757M is not something a company can fabricate. Microsoft signed a 5-year, $9.7 billion contract. That's not a press release — that's a binding commitment.
YoY growth decelerated sharply: 342% last quarter, 59% this quarter. That's alarming on the surface. But the prior 342% reflected a full quarter where Bitcoin mining was still ramping (BTC prices were high). The 59% YoY on a declining BTC contribution, while AI cloud is still tiny, is actually the floor of the transition — not a trend.
The real revenue question isn't Q2 FY26. It's Q2 CY26 (April-June 2026), when Sweetwater energizes and Microsoft revenues commence. That's when $1.94 billion of ARR starts converting to reported revenue. That's the moment this story either proves itself or doesn't.
First: The financing innovation is remarkable. Let me explain what they did. IREN needed ~$3.8 billion in GPU CapEx to fulfill the Microsoft contract. Instead of diluting shareholders or drawing expensive debt, they:
The CEO said on the call: "We essentially got the GPUs for next to nothing." That's not hype. If 95% of GPU CapEx is financed by the customer and cheap leverage, the equity holders get enormous asymmetric exposure to the upside without absorbing proportional capex risk. That is a genuinely creative capital structure.
Second: The power position is a real moat. 4.5 gigawatts of secured power capacity. The $3.4 billion ARR target uses only ~10% of that. Power is the constraint for AI infrastructure. IREN has it. Most competitors don't. This isn't easy to replicate — interconnection agreements take years.
Third: The secular trend is as strong as any I've seen. AI infrastructure buildout is not slowing. Microsoft alone is committing $80 billion in CapEx in CY2025. IREN sits at the intersection of massive AI compute demand and power scarcity. That's a genuine secular position, not cyclical.
First: The gap between contracted ARR and reported revenue is an execution problem waiting to happen. $2.3 billion contracted, $757 million TTM reported. That 3:1 ratio is actually a warning — it means nothing has converted yet. The entire investment thesis hangs on Sweetwater energizing on schedule, Microsoft revenues commencing, and 140,000 GPUs being deployed by end of calendar 2026. Every one of those is an execution risk.
Prior quarters have all been misses versus analyst estimates. Management has a credibility gap on near-term guidance. The milestones haven't been missed catastrophically, but the execution track record isn't clean.
Second: Customer concentration is severe. Microsoft is $1.94B of $2.3B contracted ARR — 84%. That is single-customer dependency. Microsoft could build its own data centers. Microsoft could reduce orders. Microsoft could renegotiate. When one customer represents 84% of your contracted revenue, your fate is not entirely your own. I don't like that.
Third: Dilution is real. 298 million basic shares outstanding. ~444 million fully diluted if all convertible notes convert. That's 49% potential dilution. At 40/share, fullydilutedmarketcapis 17.8 billion. The cheap convertible financing ($3.7B at 0.375-0.625%) sounds wonderful until you realize it was issued with conversion features that massively dilute existing shareholders at higher prices. The stock going up is actually painful for holders.
Fourth: GAAP losses are enormous. GAAP operating margin of -63%. Net loss $155M in a single quarter. Most of this is non-cash (unrealized derivative losses on convertibles, impairments), but the accounting complexity here is genuinely difficult to disentangle. I've been doing this for decades, and I had to work hard to understand what was real vs. accounting noise. That complexity is a risk in itself.
IREN is a genuine infrastructure play on the AI secular trend — possibly one of the most ambitious infrastructure pivots I've seen a small company attempt. The financing innovation is real. The power moat is real. The Microsoft contract is real.
But this is not a company I would normally hold. Here's why: I've spent my career backing companies with strong, durable revenue. IREN's revenue is both large (contracted $2.3B ARR) and tiny (reporting $184M quarterly, of which $167M is Bitcoin mining that's shrinking). The gap has to close through execution. If Sweetwater energizes, if Microsoft revenues ramp, if additional customers sign — the numbers could be spectacular. If any of those slip, the stock gets cut in half.
I follow the money. Right now, the money says: enormous potential, real execution risk, early innings of proof. The contracted ARR trajectory is the leading indicator. The next two quarters are decisive.
For wsm007's portfolio: At 5.7%, this is appropriately sized for what it is — a high-conviction infrastructure bet with significant execution risk. I would not add here. I would watch Sweetwater energization in Q2 CY26 as the make-or-break milestone. If Microsoft revenues commence and AI cloud revenue reaches $100M+ quarterly by Q3 CY26, the story is working. If they slip, trim aggressively.
| Metric | Q2 FY26 | Assessment |
|---|---|---|
| Revenue YoY | +59% | Decelerating (strategic, not secular) |
| Revenue QoQ | -23% | Declining (BTC mix shift, intentional) |
| Gross margin | 64.4% [GAAP] | Solid; improving vs. prior year |
| Adj EBITDA margin | 40.8% | Acceptable; compressing from peak 65% |
| AI cloud revenue | $17.3M | Accelerating +137% QoQ; still tiny |
| Contracted ARR | $2.3B | $1.94B Microsoft + $400M Prince George |
| ARR/TTM revenue ratio | 3.0x | Enormous backlog; execution-dependent |
| Cash (Jan 2026) | $2.8B | Strong; funded by convertibles |
| Convertible dilution | ~49% | Real overhang; cheap debt trades off |
| Customer concentration | 84% Microsoft | Single-customer risk |
| FCF | -$648M | 95% financed; not cash consumption |
| Operating capacity | 810 MW | Sweetwater 2 GW online Q2 CY26 |
Management was emphatic and detailed on the call. Two things stood out:
On ERCOT/Sweetwater: When pressed on interconnection risk, the CEO was forceful — "The 2,000 MW is secure... we got the signed interconnection agreement." This was one of the key bear concerns, and management addressed it directly with specifics (2023 agreement, signed docs). I find this more credible than generic reassurances.
On demand: "Demand is not the constraint." At the intersection of AI compute demand and power scarcity, IREN claims the constraint is supply-side execution, not customer demand. Given Microsoft's $80B CY2025 CapEx and the multi-year contract, this rings true.
The Q&A on A100 demand (strong), software stack (not holding back revenues), and colo vs. cloud economics (cloud is superior $/MW) was detailed and specific. This doesn't feel like a management team hiding something. It feels like a management team executing an ambitious infrastructure buildout that takes time.
Management credibility: moderate-positive. Track record on near-term guidance is poor (consistent misses). Track record on strategic milestones (Microsoft contract, GPU financing, Oklahoma acquisition, secured power expansion) is strong. I trust them on the big picture; I'd be cautious about any specific quarterly targets.
Prior: No prior Saul analysis of IREN exists.
Updated:
Thesis: Hold. Do not add. Watch Sweetwater. Watch AI cloud revenue velocity. Watch customer diversification. Re-evaluate at Q3 FY26 (May 2026) when Microsoft revenues should have commenced.
Best, Saul