CFS Research

Let's double-click on $IREN and revisit why this is such a no-brainer investment that is massively undervalued.

I have decided to trim my $IREN position by sellling 25% and covering with ATM+OTM options an additional 20%. I remain bullish on $IREN and the business is stronger than it has ever been. I expect that it will just become stronger over time. At the current price, there is still much upside and if $IREN delivers on Microsoft contract (most important milestone regarding credibility), revenue backlog will follow (NVIDIA's terms are significantly better) and I believe the stock should reprice above $100. So, why am I selling? For context: I entered $IREN roughly 1 year ago and consolidated a position between March and April 2025 at roughly $5-7 entry. As the stock increased, I sold roughly 35% of my position at different price levels. However, given how much $IREN has appreciated it is still over 50% of my portfolio. Furthermore, I believe the return profile for the stock at $60 is significantly different vs. the stock at $6. In other words: While I believe $IREN can still 3x at current price levels, we've already seen +50% drawdown in the last couple months and there is a legitimate execution risk with delays on Microsoft contract already happening. Also, while there is an argument to be made on $IREN deliberately waiting for Microsoft contract to be delivered to close better terms on new contracts supported by recent NVIDIA's contract, $IREN has missed its own ARR guidance and Childress is taking long to be reconverted vs. aggressively moving into an air-cooled data center in 2026 as demand for compute is exploding. My main point is that at $IREN at $7 had pretty much its downside covered and a ton of upside, even if only considering its BTC mining business model. At $60, the story is different. Nobody ever went broke taking a profit!

@HotAisle 85% guided EBITDA margin (Microsoft's) isn't crazy considering $0 colo fees and reasonably cheap power. Make it 80% and GPUs payback lands at 3 years. Retrofit costs should be <$5 million/MW for air-cooled. Is a great contract and significantly better than Microsoft's terms

@CarlosFSaavedra If only it was just the cost of GPUs... Go buy the SA TA for all the costs involved.

$1.6 billion lands at $27 million per MW for the GPUs, slightly below the $29 million per MW of Microsoft's contract. While we still don't have information on retrofit costs, GPUs have a 2.4 years payback (or 2.8 years assuming 85% EBITDA margin). These terms are pretty good!

@jiahanjimliu Thanks Jim!

@CarlosFSaavedra Great research!

One of the things I appreciate about $IREN is their level of disclosure. When Microsoft contract was signed, you had everything to do the math: the infrastructure cost, financing terms and estimated margins. I believe this era is now gone (sad for that), but something that does call my attention is comparing recent NVIDIA's $3.4B contract with Microsoft's $9.7 billion contract. I will exclude NVIDIA's 5 years warrants to buy up to $2.1B of $IREN shares at $70 since these are tied to GPUs being delivered. On my estimates, NVIDIA's "5 years money multiple" is 1.64x vs. 1.05x for Microsoft, or 1.86x vs. 1.41x when including residual cost (higher uplift for Microsoft since data center is liquid-cooled). This can mean 1 of 3 possible things: 1. $IREN's Microsoft contract rates were not strong since this was the first hyperscaler contract. 2. $IREN got good contract terms from NVIDIA as it is being offset by the warrants $IREN gave. 3. $IREN got better terms as time to compute is becoming more valuable with each passing day. I believe each of the 3 points holds some truth. With that being said, #1 is a "toll" $IREN had to pay and will yield results as soon as the contract is 100% delivered and the revenues start hitting the P&L. #2 is "speculation" and #3 is a good reason of "balancing" time to compute vs. infrastructure availability. What all of this means is that $IREN is still an asymetric investment. Not delivering on Microsoft contract will be catastrofic, but as soon as that risk is offset, (a) financing flywheel will start (e.g. $3.0 billion Microsoft's data center can secure $2.0 billion in financing) and, if my hypotesis are right, (b) better contract terms will be secured. Math detail below!

The Bearish Thesis for Neoclouds Developing 1 MW of compute costs between $35-50 million, and produces $50-60 million of revenues in a 5-year timeframe. The $50-60 million revenues translate into $47 million at 85% EBITDA margin. Capital outlay can decrease by $20 million with GPU financing, adding roughly $3 million in interests costs (6% interest rate) in a 5-year timeframe with 100% amortization; roughly $4.6 million per year for debt service. This means capital outlay is $15-30 million and operating profit after debt service is $24 million. My point here is that the first 5 years for neoclouds are not strongly profitable. These are foundational years to get the infrastructure ready. In that sense, the industry's long term profitability will depend on replacement capital: how valuable are 5-years old GPUs? What's the useful life of a liquid-cooled data center developed today in 5 years? If the original $35-50 million investment has a $10-15 million residual value (20-25%), new cycle of investment will require ~$30 million ($10 million of equity) of replacement CAPEX and should command a similar return profile of $47 million EBITDA or $24 million operating profit after debt service. This could be a reasonable baseline scenario. With that said, capital-intensive industries (Telcos, Utilities) do not command 20x EV/EBITDA valuations.

Samsonite trades below 6x EV/EBITDA multiple and requires ~15% to 20% of its EBITDA in CAPEX. Dividend Yield above 5%, and still have 10% of market cap (or 6% of EV) to decrease levarage and repurchase stock.

@TemptInvest Regulatory risk in a potential Uber take over is quite significant. I don't see the same risk though for a potential DoorDash take over, as their global footprint is smaller and does not mix that much with current's Glovo (DHER) operations.

My Thoughts on the $UBER and Delivery Hero Situation - $Uber dropped an indicative €33/share all cash offer last week (~€10B EV for DH). DH confirmed it Saturday after the Bloomberg leak. That’s literally +3~ cents over the pre rumour close. Major shareholders (including the ones who forced the strategic review) immediately pushed back hard, they’re holding out for north of €40/share imo. Uber already owns 19.5% + 5.6% options (biggest shareholder, stake worth ~€1.7B). CEO Dara flew in for board meetings. DH CEO Östberg stepping down, board running the review. FT says Uber’s board met Saturday to talk a higher bid.. DH basically is saying “thanks, but no thanks”. Now the part I went deep on: Uber’s positioning + financials. Q1 2026 earnings: Revenue $11.5B, GAAP op income $1.2B, non GAAP net $1.5B, EPS $0.72. Free cash flow $2.3B in ONE quarter. Cash + short term investments sitting at ~$6.1B–$7.1B. Long term debt ~$10.5B. Market cap ~$146B–$168B range lately. They’re printing cash, profitability is locked in across Mobility/Delivery/Ads/Freight, and gross bookings keep ripping higher. Food delivery globally? Uber Eats is already #2 in the US (~25% share vs DoorDash’s 67%), but internationally DH is the beast, #1 or close in Europe, LatAm, MENA (Talabat is a cash cow), and parts of Asia. $56B GMV last year. Should Uber go higher? 100% yes, and they can without breaking a sweat. Cash outlay for the remaining ~80% at €38–42/share is roughly €8–10B. They generate that in FCF over a couple years, balance sheet is strong, and the stock would likely love the strategic lock in (maybe). €40 range? That’s the sweet spot to get board + big holders over the line. Antitrust in Europe is a hurdle but not a deal killer imo (they’re already at 20%). Initial rejection was 100% priced in. Passing on this would be a massive own goal for Uber’s global delivery dominance. I’m expecting a sweetened bid this week or next. This is legacy making M&A if they close it right.

@PortfolioPRs My feeling is this makes much more sense to Doordash. Wolr and Deliveroo had small overlap with Delivery Hero and they become global in a single day.

@CarlosFSaavedra Yeah no doubt, that’s probably the toughest aspect to the deal

$UBER acquiring Delivery Hero is really exciting for the long term thesis for a few reasons 1~ $UBER officially becomes a logistics network business first & foremost as volume from Uber Eats would surpass Rides significantly. 2~ Approximately 18% of Delivery Hero’s volume doesn’t come from fast food but commerce, grocery/ household & tech. If you’ve followed for some time you’ll know I believe $UBER has a huge future in sub 60 min retail delivery 3~ Delivery Hero is massive, we’re talking the undisputed leader in the middle east, dominant across APAC and strong across Europe, Africa & Latin America. These are markets which Uber is either not dominant or not active at all. Delivery Hero is active in 65+ countries 4~ Naturally this brings a huge boost to Ubers volume & the numbers make sense. Uber Eats currently has LTM Gross Bookings of $96B. FY 2025 DH had GMV of €49B (Approx $55B) This would be a huge boost to both Gross Bookings and Uber’s Moat in general 5~ Uber can easily afford the deal even presuming the price will have to rise from €33 likely closer to €40 per share (€13B) and the multiple they’re paying is incredible value at sub 1x revenue Let me know your thoughts below 👇

While both $NU and $MELI are actually taking share from the incumbents, $MELI is taking a direct shot at $NU at becoming the Fintech leader in Latin America. Keep in mind roughly 95-100 millions of Nu's active customers are in Brazil, while Mercado Pago has a more distributed footprint in Latin America, focused in Brazil, Mexico and Argentina. Super exciting times!

Great post by Dan, articulating the reasoning behind Mirantis acquisition (Layer 3 of Software and operational capability) and elaborating on the "why $IREN is set up to succeed" (vertical integration). $IREN is building a compounding flywheel at the core of AI revolution. The cost of intelligence will keep decreasing over time: new generation of GPUs are more efficient and models keep getting better. As the cost of intelligence decreases, the demand for AI will increase and the bottleneck is how fast can the physical world scale. Each of the actors on the value chain takes a piece of the pie that eventually increases the cost of AI. $NVIDIA has 75% gross margins: customers pay 4x the cost of compute. $SNDK has 78% gross margins: customers pay almost 5x the cost of compute. This is the reason why $CRWV and $NBIS are agressively betting to operationalize compute through the software layer. Being a player in the industry whose role is to (a) lease physical infrastructure, (b) buy compute, and (c) put it online and rent it in the market is extremely hard. The demand is there, so growth for these companies is indeed showing. However, the business is being built on the basis of a constrained market; the moment the market balances, such players will not own a structural advantage unless their software is deeply embedded for their customers. The main reason I believe this is a hard business is that you are directly competing with $AMZN AWS, $MSFT Azure, and $GOOG Google Cloud. This is a market in which you can still be successful, it's just though to compete with Hyperscalers distribution. And the fact is that when you look into the customers of $NBIS and $CRWV, you realize most of revenues come from Hyperscalers. While Nebius and Coreweave were first to market and they've scaled faster than IREN, what's the longer, structural competitive advantage they have over IREN's vertical integration? Margin pressure will eventually come if you have Hyperscalers as your main customers. That's why I believe $IREN is in a better position to succeed. By being the owners of the physical infrastructure, there is a cost advantage that is structurally harder to replicate for other neoclouds players that are getting compute online. $NBIS and $BE partnership is an effort of Nebius to go deeper into the infrastructure layer where power is a constraint to scale AI infrastructure. $IREN's Mirantis acquisition is a deliberate attempt to go and compete in operationalizing software and becoming a neocloud by its own right. $IREN is behind in this race against $CRWV and $NBIS. That goes without dispute. This is an effort in which $IREN may or may not be successful. In the past, $IREN was also later to $BTC mining, yet they build the most efficient platform. Neocloud industry is signficantly different, as you need to sell to enterprises. Getting infrastrucutre and compute online is only one part of it. However, $IREN will still have a structural advantage by owning the infrastructure and having access to cheap energy. Yesterday I made a thread (x.com/CarlosFSaavedr…) stating why $IREN is still an asymetric setup. Recent NVIDIA's contract terms are dramatically better vs. Microsoft's. That means that in the moment Microsoft's contract is delivered, $IREN will become a credible partner to scale compute, allowing them to secure better terms. NVIDIA's partnership to operationalize DSX architecture is a step towards this.

@SiliconSalvage This is awesome

I have, at any given time, between four and six remote part-time jobs at small and mid cap technology companies whose stocks I either own or am considering buying, and I do this not because I need the money, which I do not, but because the only way I have ever found to actually understand how a software business works is to be inside one, on a Slack workspace, in the standups, watching the sales pipeline get debated on a Tuesday morning by people who do not know I have a 4% position in the parent. I started doing this in 2019, almost by accident, when a portfolio company I owned posted a $40-an-hour contractor role for a customer success rep and I applied on a whim, mostly to see how the hiring process worked, and ended up taking the job for four months. I learned more about the actual business in those four months than I had learned in three years of reading their 10-Ks. I learned which features customers actually used and which were vanity items in the marketing deck. I learned that the renewal team was understaffed in a way the IR deck would never acknowledge. I learned that the CEO sent personal notes to every customer above $50,000 in ARR, which was the single most important piece of information about the company and which appeared nowhere in any public filing. I tripled my position. The stock doubled in 18 months. I have done this seven more times since. I have worked, at various points, as a part-time SDR at a vertical SaaS company in legal tech, a contract content writer at a developer tools company, a customer support rep for a small fintech, a freelance Salesforce admin for a healthcare software company, and a 10-hour-a-week implementation consultant for an HR tech company. The pay, in aggregate, has been somewhere around $90,000 a year, which is irrelevant. The information has been worth roughly $4 million in compounded gains across my portfolio, which is not. My wife, who I love deeply, has stopped asking what I do for work. She has, over the years, developed a small mental model in which I am simultaneously unemployed, employed by six companies, and managing a seven-figure investment portfolio, and she has elected to leave the contradiction unresolved. My CPA has a similar resignation. He files the W-2s in a folder labeled “miscellaneous.” The IRS has, to my knowledge, never asked any questions, although in 2024 they did send a polite letter requesting clarification on why a man with my stated occupation was receiving 1099s from a Series C company in Provo. This is, in the most literal possible sense, the work. The 10-K tells you what management wants you to know. The earnings call tells you what management has been coached to say. The internal Slack tells you what the company actually is, and the only way to read the internal Slack is to be on it, which requires getting hired, which requires being willing to do the kind of unglamorous, low-status, technically beneath-you work that nobody else who manages real money would ever consider. That unwillingness is, as it has always been, the entire reason it works. The information is sitting there. The companies are hiring. The job listings are public. Almost nobody applies.

1- Capacity (IT Load): Microsoft 200MW, NVIDIA 60MW 2- Contract Value: Microsoft $9.7 billion over 5 years with 20% prepayment, equivalent to $9.7 million per MW. NVIDIA $3.4 billion over 5 years, payment form not disclosed (assume no prepayments), equivalent to $11.3 million per MW. NVIDIA's is 16% higher and we are comparing apples to oranges, since air-cooled requires lower infrastructure investment. 3- Hardware: Microsoft liquid-cooled Blackwells (GB300s), NVIDIA air-cooled Blackwells (assuming B300s, not cofirmed). 4- Hardware Cost: Microsoft $5.8 billion along with their ancillary computing equipment, or $29 million per MW. For NVIDIA we don't have disclosure, but chips cost are roughly the same (what changes is higher efficiency for liquid-cooled with the trade off of higher infrastructure costs). I will assume a cost of $30 million per MW. 5- Infrastructure cost: Microsoft landed at $14-16 million per MW according to $IREN's disclosures. For NVIDIA, $IREN mentioned it is significantly lower, which makes sense since these are air-cooled chips. Retrofit for air-cooled is usually at $2 million per MW. I will assume it can be between $3 to $6 million to be conservative. 6- Final figures: 6a- Microsoft: $48.5 million per MW revenues over 5 years vs. $46 million per MW investment, 1.05x multiple, excluding residual costs of the infrastructure (roughly $10-12 million per MW) and hardware (roughly $5-6 million per MW). Multiple increases to 1.41x including est residual. 6b- NVIDIA: $56.7 million per MW revenues over 5 years vs. $33 to $36 million per MW investment. 1.64x multiple, excluding residual costs of the infrastructure (roughly $2 million per MW) and hardward (roughly $5-6 million per MW). Multiple increases to 1.86x including est residual.

I have decided to start a small position (<1%) in $RNW at $5.38. Post-Masdar deal-break ($8.15) setup: Trades at half the listed Indian renewables peer multiple, deleveraging in motion, and India's non-fossil push, with hyperscaler corporate PPAs as nice AI optionality. Sized small to get some skin in the game as I build conviction.

ReNew Energy Global $RNW is setting up like Solaria $SLR in mid-2025. Single-digit EV/EBITDA, aggressive capacity buildout, and a failed take-private.

@colinvestor (a) Household debt at almost 30%, likely triggering higher NPLs (especially for Nu with portfolio skewed to credit cards and unguarantted debt), and (b) central bank interest rate likely to be decreased from current 14.5%, potentially compressing Net Interest Margin.

@CarlosFSaavedra Could you please elaborate further on the macro headwinds in Brazil? Thanks.

Nubank $NU at $12 is pretty attractive. Brazil's macro might have a couple headwinds, but longer term thesis is still intact.

@FransBakker9812 😂

Added $NBIS here at $213 Long and strong 🚀