Tree (@Tree75769) header
Tree (@Tree75769) profile picture

Investor & CRE Developer | Early $IREN & Ripple | also hold $GOOGL $MSFT and more

New York, USA
Joined March 2020
And this 🤯:
1
2
78
Replying to @_Sgr_A_Star
Well then I’ll just leave this here 😏
2
2
104
My last $IREN post looked at the economics of building a new AI data center. A post from @_Sgr_A_Star got me thinking about the mining halls IREN already owns at Childress. From a developer’s perspective, that’s a different deal. How much cash could I get back out of a retrofit while I still own the property? I ran it on the same 10 IT MW example as before. IREN disclosed a five-year deal at about $22M of annual revenue per IT MW, with the customer prepaying 90% of the estimated GPU cost. IREN didn’t name the site. I wanted to see what those terms could do on a retrofit. I used $400M for GPUs and $80M to convert the halls. That $8M per IT MW is @_Sgr_A_Star ‘s well reasoned assumption, at the high end of the $5M–$8M conversion range Core Scientific showed for its own projects. The customer advances $360M, leaving $40M on the GPU bill. Say IREN puts up the $80M for the retrofit, then borrows $100M against the GPUs and $80M against the finished property. After paying the remaining GPU bill and $21.8M in fees and reserves, $118.2M comes back to IREN. That’s the entire $80M retrofit investment back, plus $38.2M of borrowed cash for the next build. The first site keeps earning. At my assumed 75% operating margin, I get another ~$220M over five years after customer credits, upkeep, debt payments and estimated taxes. I’ve assumed no sale. IREN still owns the site and GPUs, with ~$55M of building debt left. Here’s why I think it matters. Of the 250 gross MW IREN plans to retrofit at Childress, ~60 MW is allocated to NVIDIA. That leaves ~190 gross MW outside the NVIDIA deal. At my assumed 1.4 gross MW per IT MW, that’s about 135 IT MW, or roughly 13½ times this example. If that whole footprint earns similar terms and financing, I get ~$1.6B back as the loans fund in phases, including $520M above the new retrofit investment, plus **$3B of cash over five years of service**. What I like as a developer is that the same property could help pay for the next build while it’s still earning on the first contract. Of course some of that capacity may be spoken for under deals already closed, but the illustration is important. One day the market will understand the flywheel $IREN has at its disposal. This is not a troubled company. It’s priced like it is. I recommend acting accordingly! (NFA of course 😉)
6
5
38
2,362
How much more stock does $IREN actually need to sell to cover its $6B funding gap? I’ve lost count of how many times I’ve seen this question on X. The dilution concern is fair. IREN has already sold a lot of shares. What I keep coming back to is how much of the remaining buildout they could fund with the assets they own and the contracts they sign. First, the $6B was a September 30 number. After quarter-end, IREN raised about $630M through its ATM at an implied average of roughly $36 per share before fees. Giving them credit for that raise, I get about $5.4B left to fund against the original FY27 capex plan. My last post looked at how IREN finances one AI deal. This time I put together four ways it could cover that $5.4B without another ATM sale or convertible. The last row is the one I keep coming back to. IREN reported $6B of unencumbered data center assets, including land and projects under construction. If it can borrow $3B against Horizons and other eligible sites, another $2.4B in new GPU loans would cover the balance. That case assumes zero new customer advances. Sweetwater is the obvious one to watch for a big prepayment, but a few smaller contracts at Childress or in Canada could help too. IREN already signed a five-year deal with a prepayment covering 90% of the associated GPU cost. Even a lower percentage on the next few deals could make a real dent. That’s why I’d pay as much attention to the payment terms as the headline contract value. A smaller deal that puts cash in IREN’s hands before the GPU bills come due could do more for FY27 funding than a much bigger deal that pays in July. It would be a mistake to think that management is not close to addressing this very concern. I think we get news in the next month before the holidays. Buy the steep discount if you can!
5
3
36
3,711
I work in real estate development, so when I saw the prepayment terms on $IREN’s newer AI deals, my first question was simple. How much of my own cash would I actually have tied up in one of these? I ran the numbers on 10 MW. IREN reported about $22M in annual revenue per IT MW on recent five-year contracts. It also disclosed a five-year deal with a customer prepayment covering roughly 90% of estimated GPU cost. I used those figures for this example. The financing, margin and exit are my assumptions. Using IREN’s illustrative costs, 10 MW takes $400M of GPUs and a $200M data center. At $22M per MW per year, that’s $1.1B of revenue over five years. The customer pays $360M up front, leaving $740M to collect over the contract. I assume IREN puts in $200M of peak equity, then borrows $100M against the GPUs and $116M against the facility. $200M equity + $360M prepayment + $216M debt − $600M build − $22M fees and reserves = $154M back to IREN. That financing returns 77% of the original equity. IREN still owns the project and has $46M of its original cash tied up. I used a 75% operating margin. For context, IREN’s own example leaves 85% of revenue after direct operating costs. My assumption puts five years of operating costs at $275M. Add $10M for upkeep, and $455M of the $740M still to collect is left. After principal, interest and modeled taxes, operations return roughly $206M. So before selling anything, IREN gets about $360M back on $200M invested. That’s $154M from financing and $206M from operations. It still owns the site and GPUs, with about $79M of debt left. I’d look at renewing the compute contract before selling. I modeled a year-five sale so I could compare the return with a real estate deal. For the property, I assumed a future colocation tenant could support $14.1M of annual net operating income. At an 8% cap rate, less conversion and lease-up costs, I get about $157M for the property. I valued the GPUs separately. Including the sale, I get about $469M back on $200M invested. That’s $2.35 back per dollar over five years, or roughly $269M of project-level profit (~38% modeled equity IRR). Not bad if you ask me! As @_Sgr_A_Star says, word. The number that interests me most is the $46M. That’s how much of IREN’s original $200M is still tied up after financing. The other $154M can go toward the next build while the first contract is still running.
23
40
5
423
45,046
Replying to @BitcoinAIGuy
Me to the haters
1
2
107
Before you give up on $IREN, take a look at what the next two years could bring. Annual AI Cloud revenue: FY26 actual: $129M FY27 base model scenario: ~$3B FY28 base model scenario: ~$9B Roughly +2,230%, then +200%—nearly 70× from a small starting base. My illustrative execution case from my model based on reasonable assumptiond, not company guidance. The scale of this opportunity is why I’m still looking beyond the selloff and bad macro.
8
1
48
3,295
$IREN Just added more shares - I couldn’t help myself. Sentiment feels different now and the company is at its inflection point. The market will catch on soon and I am not missing out!
5
1
1
75
4,770
$IREN 2026 hires look a lot more interesting when you put them all on one page. People with experience at Oracle, AWS, Google, NVIDIA, CVC and JPMorgan. Across operations, engineering, product, security and finance. The more I look at the team coming together, the more confidence I have in what IREN is building. Nicholas Sandberg → Head of US Operations. Joining from Oracle, with earlier AWS and IBM experience. Kambiz Aghili → Chief Product Officer. Previously VP of Products at Oracle Cloud Infrastructure, leading its native multicloud platform strategy. Now responsible for IREN’s AI Cloud product roadmap. Michael Nudelman → Chief Development Officer. Senior experience at Google, CyrusOne and Beale Infrastructure. Leading global data-center development, including site planning and permitting. Eric Hammersley → Chief Information Security Officer. From Nutanix, where he was VP Engineering and Chief Product Security Officer. Earlier security work at NVIDIA. John Gross → Chief Innovation Officer. Liquid-cooling and data-center engineering specialist. Already advised IREN for years; now brought in-house to lead engineering standards and commissioning. Chris Parker → Chief Marketing Officer. IREN acquired Awaken and brought its founder and senior team inside the company. They’ve also added Boris Dangubic from CVC, Ryan Boers from Quadrant, and Bob Chen from JPMorgan into investor relations. AFR has separately reported the recruitment of KKR’s Aaron Tan. Then there are the teams brought in through Mirantis and Nostrum, adding software, customer support and European development and operating experience. What I like is how closely these additions match the work ahead. You need people who can get sites developed and complex GPU clusters running. You also need people who can turn that compute into products enterprises will trust with important workloads. The product, security and software additions are especially telling. They support IREN’s ambition to serve customers who need managed services and support alongside the hardware. That could broaden both the customer base and what IREN earns from each relationship. And experienced people are choosing to join while the existing core leadership remains in place. I find that encouraging: continuity at the top, with more depth underneath it. Management is showing a willingness to go outside the company for expertise, bring specialists in-house and acquire established teams. That’s the kind of commitment I want behind an expansion this ambitious. Of course, compensation and acquisitions have to earn their return. A résumé doesn’t get a cluster accepted. But the hiring is a concrete response to execution risk. IREN is assembling people with experience in the very areas that will determine whether this buildout succeeds. That strengthens my conviction. Better execution means more of the opportunity can become operating revenue, repeat business and, ultimately, value per share. The talent coming in is becoming a meaningful part of my bull case.
2
13
3
60
14,541
Everyone is talking about $NVDA’s $12.9B Hugging Face deal. Almost nobody is asking what NVIDIA is actually building — or why it may be the answer to OpenAI, Google and Meta designing their own chips. I went down that rabbit hole, and it led somewhere I didn’t expect: $IREN The custom-chip threat to NVIDIA is real. The biggest AI companies have enough scale to design silicon around their own workloads. But NVIDIA’s response doesn’t look like “just keep making better GPUs.” Look at what they’re assembling. Hugging Face gives NVIDIA a front door to millions of developers, models and datasets. Run:ai + Slurm schedule workloads. Dynamo and the inference stack help run them efficiently. Lepton connects AI demand to distributed compute. DSX standardizes the AI factories underneath it. And NVIDIA is working with some of the largest pools of capital in the world to finance hundreds of billions of dollars of those factories. What surprised me most is how deliberately open the whole thing remains. Hugging Face stays multi-cloud and multi-accelerator. Slurm stays open source. DSX is modular. NVIDIA has repeatedly made the point that it doesn’t need to own the cloud itself. That makes sense. If AI spreads from a few giant labs into millions of enterprises, startups, developers and sovereigns building on open models, most of them aren’t designing chips or building gigawatt data centers. Someone still has to provide the factory. That’s where $IREN gets interesting. IREN owns the physical layer: Power → land → data centers → GPUs → operations Then it bought Mirantis, adding the software/orchestration layer needed to turn that infrastructure into a real multi-tenant cloud. Mirantis’ k0rdent AI already integrates with NVIDIA DSX. Sweetwater is already designated a flagship NVIDIA DSX deployment. And here’s a detail I think gets overlooked: IREN is an NVIDIA Exemplar Cloud, while Mirantis k0rdent AI is an NVIDIA-Certified Hypervisor for Grace Blackwell. Based on NVIDIA’s current public lists, IREN appears to be the only Exemplar Cloud provider with both under the same corporate roof. NVIDIA is already an IREN customer. Its investment rights in IREN vest as NVIDIA infrastructure gets deployed. And there’s another connection I hadn’t appreciated until digging into Hugging Face: Together AI, Fireworks AI, Fal and Cohere are all part of the Hugging Face inference ecosystem. They’re also $IREN customers. That does NOT mean Hugging Face workloads suddenly get routed to IREN. The broader pattern is more interesting anyway. NVIDIA appears to be building an open, distributed AI ecosystem from the model layer all the way down to the physical factory. And $IREN is increasingly positioned as one of the factories underneath it. I used to think of IREN mostly as a power + GPU infrastructure story. The more I dig into NVIDIA + Mirantis, the more I think there’s a real chance it becomes something closer to an AI platform sitting on top of extremely scarce physical infrastructure. If that happens, the market eventually has a very different question to answer about what $IREN should be worth. Credit to @jiahanjimliu, @Agrippa_Inv, @FransBakker9812 + the OnlyFrans crew — a lot of their research/threads helped send me down these rabbit holes. If you’re new to $IREN, I broke down the numbers behind my higher-execution case in my last post ⬇️
🤖 Made with AI
6
18
3
80
6,591
$IREN I did something with my IREN model this week that I’ve intentionally avoided for a while. I stopped asking: “What can I underwrite with high confidence today?” And asked a different question: What does IREN look like if most of the things already in motion simply work? I expected the answer to require some pretty stupid assumptions. It didn’t. My normal model is intentionally conservative. Unsigned contracts don’t get full credit. Future capacity is discounted. I don’t assume Rubin economics before they’re proven. And I count the debt, dilution and replacement costs. That’s how I get to roughly $50/share today. For this exercise I changed the question, not the discipline. No faster timelines. No $30M/IT-MW assumption. No 6-year GPU life. No NVIDIA financing backstop. No Mirantis software multiple. I mostly let IREN execute. IREN is already targeting 0.8GW IT in CY27. My high-execution branch gets to 1.35GW IT at YE28. That’s another 550MW IT, or roughly 770MW gross. IREN already has 4.4GW gross of announced CY28+ capacity. So I only need ~17.5% of that later capacity to convert by YE28. That’s not where the aggressive assumption is. The real assumption is the economics: 95% monetization × $20M ARR/IT-MW And even that isn’t particularly heroic. IREN is already signing recent 3-year contracts at >$20M/IT-MW, with ~2-year GPU paybacks and 45–55% customer prepays. I’m also giving Rubin zero pricing uplift. That may prove conservative. WhiteFiber’s filed VR200 deal implies ~$7.14/GPU-hour versus ~$4.75 on a separate B300 deal. Not apples-to-apples. But directionally, Rubin pricing coming in above Blackwell would hardly be surprising. And NVIDIA is showing Rubin at up to 30x throughput/MW vs GB300 on a specific agentic workload. I’m obviously not multiplying IREN pricing by 30x. The point is simpler: Customers may get dramatically more economic output from the same scarce MW while IREN’s pricing remains strong. Dell also just exited the quarter with a $95B AI-server backlog. So the demand backdrop isn’t exactly weakening. Now the valuation: 1.35GW × 95% × $20M = $25.65B ARR At 4.25x ARR: $109B operating value $10B remaining pipeline − $34B net debt = $85B equity value ÷ 520M diluted shares = ~$163/share at YE28 Discounted back at 12%: ~$126/share today. And I’m not pretending the growth is free. My normal model uses ~508M shares and ~$24B net debt. This branch uses 520M shares and $34B net debt. At 4.0x ARR: ~$116 today At 4.25x: ~$126 At 4.5x: ~$135 That’s what surprised me. You can get to a very different outcome without needing the crazy stuff. No $30M+/MW portfolio pricing. No NVIDIA backstop. No huge Mirantis software business. No full pipeline conversion. No faster build. No major multiple rerating. This isn’t my base case or a new price target. It’s a high-execution branch. But I’m paying a lot more attention to it now. As always: I own $IREN. NFA.
5
12
3
119
13,773
Replying to @RealJimChanos
@RealJimChanos says $IREN bulls should stop leaning on ARR, $/MW and prepayments and look at the actual accounting. So I did. It got more bullish. At June 30, IREN disclosed ~$5.1B of ASC 606 remaining performance obligations plus ~$11.4B of contracted arrangements under ASC 842. That’s ~$16.6B of contractual obligations in the financial statements. Then ~$2.8B of new contracts were signed in July. Before assigning anything to the unnamed frontier AI lab announced in August, I can identify >$19.4B of contracted obligations. I’m intentionally not calling all of that “backlog.” Different accounting, different durations. That’s the point: you don’t need the ARR slide to find the demand. The contracts are in the filings. IREN also ended June with $1.84B of deferred revenue. Roughly $1.62B was advance payments on lease-accounted tranches that hadn’t even commenced. Not profit. Customer cash arriving BEFORE the asset starts producing revenue. And this is where I think the dilution argument starts to break. For Microsoft/Horizons: ~$3.6B of investment-grade GPU financing + ~$1.9B of Microsoft prepayments against a ~$5.8B GPU purchase agreement. ~95–96% of associated GPU capex funded. For Mackenzie, the $2.4B financing alone funds ~90% of associated GPU capex. Customer prepayments sit on top, pushing total GPU-linked funding above GPU cost and helping fund data-center capex. Now FY27: $25–30B capex. ~$14B existing cash + committed GPU financing/prepayments. ~$8B additional GPU financing/prepayments targeted. That leaves a $3–8B gross residual BEFORE data-center financing and operating cash flow. Meanwhile management says 100% of its data-center portfolio remains unencumbered — including Horizons 1–4, the 300MW gross / ~200MW IT Childress build supporting a $9.7B Microsoft contract. Think about that structure. The Microsoft compute layer is already ~95–96% funded. The data-center layer is still available to finance. So the first DC financing doesn’t have to solve a $30B problem. It attacks the remaining $3–8B residual. In the favorable end of my funding bridge, DC financing + ex-prepayment operating cash flow can take that residual close to zero before incremental common equity. I’m not making that my base case yet. But it’s a very different setup from “IREN needs to issue $30B of stock.” Customer prepay → GPU debt → operating asset → DC financing/refi → recycle capital. That’s the flywheel. The point isn’t zero dilution. It’s that common equity can become the residual funding rail instead of the primary one. Now Chanos’ ARR/revenue point. How can IREN target >$4B of operational ARR at Dec. 31 while sell-side FY27 revenue is only ~$2.5B? FY27 runs July through June. Dec. 31 is halfway through it. The >$4B is the annualized rate of the operating fleet at that point. The first half is the ramp. Different clocks. No contradiction. His capex comparison has the same issue. FY27 capex funds multiple vintages: finishing CY26, CY27 air-cooled deployments, major DC construction for H2-CY27 Childress/Sweetwater, plus early CY28+ work. The GPUs for those new H2-CY27 liquid-cooled facilities aren’t even in the FY27 guide. A multi-vintage construction budget divided by one earlier vintage of ARR will obviously make returns look bad. The numerator and denominator aren’t the same assets. One Chanos point I’ll give him (cont.)
11
17
2
142
21,687
$IREN is COOKING H1 delivered and $NVDA GB300 exemplar status solidifying them as a top AI compute provider. Also, we have 11 new job listings in last few days with a total of 202 openings. Every week this number droos by 5-10 and adds another 5-10. This is a company with massive ambition. Here are the new jobs:
4
5
1
118
7,704