D2C3 🚴

Good morning paradise Epicview 🏌️‍♂️Golf & BTC vibes! Keep swinging and HODLing

$IREN Acquires Mirantis Previously, I have said that leading enterprises will move towards the Netflix model where they in-house their infrastructure software: "For AI software, the foundation software layer will come from $NVDA, and the AI Native and Leading Enterprises of 2030 will own most of their infrastructure software" (1). Instead of waiting for the market to move, Mirantis is IREN's acceleration into this market. From Mirantis' website: "Mirantis helps organizations achieve digital self determination by giving them complete control over their strategic infrastructure (2)". In practice, Mirantis essentially allows leading enterprises to own their own cloud stack on bare metal. Current customers of Mirantis' are Adobe, DocuSign, Inmarsat, Paypal, Societe Generale, S&P Global (2). These types of companies are not Netflix grade but strong enough to in-house significant SaaS services with . Enable Leading Enterprises Mirantis operates the orchestration layer software and provides CAPI (Clusters API). API's are commands which encapsulate software logic. Underneath the hood, CAPI manages the logic needed to deploy and run on AWS, Azure, GCP bare metal, VMware managed on-premise bare metal, or on any Intel/Nvidia/AMD bare metal servers. This makes Mirantis a perfect solution for hybrid HS Cloud and on-prem deployments. While Adobe, DocuSign, and Paypal use Mirantis' to tailors performance optimizations and greatly improve margins, many other long tail customers user Mirantis' for data sovereignty. Mirantis' lists many of it's long customers anonymously as "asian bank", leading aerospace company, "leading asian telecom operator", "defense agency cybersecurity office", "european medican center", "top 20 pharma company" (3). With into the rise of Sovereign AI, this is highly strategic. With the developments of the Iran War, it's important for certain organizations in other countries to have full control of their compute and not in datacenters next Washington D.C. (Virgina is the largest datacenter hub in the world), operated by software controlled by American Hyeprscalers. Compute sovereignty is not only having datacenters located in your country but probably more importantly, controlling the software that runs it. This is where all the cybersecurity breaches are. Nebius is positioned well to serve the Israel and EU sovereign AI markets markets, but I'm sure there are many other countries that would prefer their operations to not be EU/Israeli-tied. In fact even countries like Australia want full control over their sovereign compute. Grading Mirantis Previously, I said that in 2-3 years, AI software infrastructure will face a period of consolidation and acquisitions can be had at good bargains. Mirantis, however, does not come from a position of weakness as it's cashflow positive and has over 100m in revenue. There is no need for it to seek a buyout. To top it off, the "transaction consideration will be paid in IREN ordinary shares" (5) or all IREN stock. This signifies alignment as Mirantis owners only derive value if they believed in the success of IREN+Mirantis. From everything above, it's clear that 625m is a great price for a cashflow positive company, strategic alignment with IREN, and at a fair 6.25 P/S. For comparison, Tavily had a ~27 P/S and did not bring on 1500 customers. $NBIS did much better on a it's Eigen AI acquisition which allows it to compete closely with FireworksAI. Mirantis allows $IREN to accelerate where leading enterprises are going: self-owned AI infrastructure. Mirantis Technical Measurements - Mirantis a primary maintainer for Docker, #7 in commits (6). - Mirantis is #20 in commits to Kubernetes (6). - Mirantis is #5 commits in OpenStack (6). In other words, Mirantis is a key contributor to core open source software. This is likely as a by-product that it needs to fix bugs and improve open source as it enables deployments for customers.

🔥🫡

@LukasFrije1re @SMASIMHO Managed K8s is the easiestlol okay. Try building enterprise grade orchestration that actually works for big AI training clusters, keeps Adobe/PayPal-level customers happy, and handles all the GPU scheduling + observability headaches.Meanwhile Nebius is stuck fighting Vineland

@SMASIMHO Yea managed k8s is the easiest lol

For comparison, $IREN just bought this layer So hold your horses, you guys still have a long way to go

@LukasFrije1re better software today, sure. But IREN owns the hard stuff secured power that’s nearlyimpossible to copy .good luck with 5-10 year delays Nebius is already dealing with exactly that constant humming complaintsBTMpermit problem .software gets acquired fast.Real power moats don’t😂

@MktMavPro $IREN $NBIS we all want real confidence, not “feels.”NBIS Vineland NJ -NJDEP is probing if the 32+ Bergen gas engines data center are a single stationary source.If ruled single source: analysts say 12-24 months extra delay.Testing capped at 27 engines.Real constraints hiteveryone

$IREN not a fan of the word “feel”… I want direct confidence in what’s going on in a foggy climate….

Hello London! ✈️🇬🇧 Juicy steak + ice cold Fuller's London Pride = perfect start !

The NYSE welcomes @twentyone to celebrate its listing! $XXI | @jackmallers

@BoHines XXI 🎉

I love Bitcoin… Twenty One.

😂🐕

@LiebermanAustin Bought 15000 shares of Iren at 4.5 because of him lol 😂

This guy is not your friend. He pumps stocks when they’re going up, creating hype and FOMO Then he criticizes inexperienced people who follow him in. $ASST has 750k in TTM revenue and a $750M market cap. Thats a P/S of around 1,000. Even worse, they claim to be some Bitcoin ETF company but their revenues come from discord servers and social media subscription services. I’ll be surprised if they’re still around in 2 years

@Agrippa_Inv @aleabitoreddit 👀🚨

Ouh man 🤦🏻‍♂️... I rarely read such "smart sounding" bad take. What you are failing to realize (or deliberately avoiding to point out) is that there are 2 different "orchestrations": Type 1 is the customer-facing stack: things like Kubernetes/Slurm/Ray, model serving, MLOps, etc. Type 2 is the provider control plane: think of fleet provisioning, isolation, firmware/driver lifecycle, metering/billing, observability, quotas/SLAs, and incident handling. $IREN explicitly targets large enterprises that bring Type 1 themselves and usually demand bare-metal so they retain full control and portability. And ironically, even in the $MSFT × $NBIS deal, Microsoft almost certainly brought its own Type-1 orchestration stack... nobody serious believes MSFT outsourced core orchestration. So does in-house Type-1 help? Sure, for acquiring and upselling smaller customers. Is it a moat with whales? No. Those customers insist on owning their stack. Not having Type-1 isn’t a meaningful disadvantage when you sell dedicated capacity to that sort of clientele. How much does Type-2 matter if parts aren’t in-house? Nowhere near your 20–30% scare number... Take a single B200 at ~$5/hr and 90% utilization: 720 × 0.9 × $5 = $3,240/month revenue. A licensed control-plane fee of $0.10–$0.20 per GPU-hour is $72–$144/month (≈2.2–4.4% of revenue). Add incremental SRE/support equivalent of ~1–3% and you’re at ~3–8% all-in. 3-8% ≠ 20-30% 🙃 That’s what “partially outsourcing Type-2” actually looks like when you keep the customer relationship. And we literally have a way to prove me right. Since last quarter, $IREN now breaks out COGS for its AI cloud segment ... and guess what the gross margin was? --> 92% 😂 What was Nebius' gross margin again last Q? ~70%? Lol. Your flawed 20–30% figure comes from marketplaces (Fluidstack/Poolside) where the platform owns demand, billing, and support – i.e., SMB/on-demand retail, not multi-MW enterprise leases. I would suggest to not rely on your "deep research" AI tool that much.... it fed you bad info mate. You’re also underestimating the infrastructure reality. $NBIS does NOT own its 300 MW NJ campus. It’s a build-to-suit facility leased from DataOne, with lease economics not disclosed (in typical $NBIS fashion, very transparent). In the current power supply crunch, the leases on that facility won’t be cheap – and if tightness worsens, future sites will certainly get even pricier. On the other hand, $IREN controls a massive power pipeline (3GW confirmed + ~5GW undisclosed) and is an excellent developer that knows how to build high-density data centers incredibly cheap – at scale. $NBIS, just like $CRWV, remains at the mercy of landlords and third-party entities for rack density upgrades across their facilities. As GPU generations become more power demanding, facilities will have to be upgraded from top to bottom to handle the increased load & necessary cooling. ... Once again, not a problem for $IREN, but a meaningful cost-driver for entities that are NOT vertically integrated on the physical infrastructure layer – i.e. neoclouds like $NBIS or $CRWV. You’re also underestimating $IREN's track record in Bitcoin mining. $IREN isn't just another $BTC miner... don’t even put it in the same bucket as $CIFR. It has been the fastest-growing and, more importantly, the most profitable large-scale $BTC miner in the history of this industry. The team runs extremely efficient ops and a very lean corporate structure – cost discipline that carries straight into its AI-focused verticals. Bottom line is that large customers don’t require your Type-1 orchestration / software layer – so your margin thesis collapses. Yet, everyone needs power and physical infrastructure, including $NBIS. $NBIS will either search for scraps in a starved power market or try to lease behind-the-meter sites with on-site generation that are VERY expensive to develop and maintain … there goes your margin 👋

To settle the Neocloud debate: $NBIS > $IREN + others. Based on the $ORCL report, NBIS whitepaper, $CRWV acquisitions, and other factors, I decided to consolidate millions into Nebius and sell off miners like $CIFR. 📏 Gross Margins > GW capacity Here's the math + why: We've had a huge speculative run across the board on Neoclouds backed by forward revenue (eg. 19B $MSFT deal with $NBIS). However, crypto miners have recently gone up 500%+ due to raw GW capacity like $IREN and cheap energy. However, what people miss out on is capacity leads to much higher revenue but that revenue means nothing if it's not profitable (eg. $ORCL 14% gross margins). Now that more information has come out regarding $ORCL's buildout failure, $NBIS's whitepaper, and others, we know power and capacity mean little if the economics don’t translate. And $NBIS has everything. 1. The Margin Gap: Full Stack vs. Middleware Miners like $IREN and $CIFR must rely on orchestration partners (Fluidstack, Poolside, etc.) to monetize their GPUs. That means giving up 20–30% of revenue (~private estimates from deep research), plus absorbing GPU depreciation, power, and O&M. The result? At $3–4/hr GPU pricing and ~80% utilization, IREN/CIFR margin = ~44–52% At $5–6/hr and 90% utilization, margin could reach ~55–60%, but that’s the ceiling Meanwhile, Nebius earns 70–75% today by owning the orchestration layer and amortizing over 4 years (71.2% from previous Q, likely 60-70%, possibly higher from recent whitepaper claiming higher GPU utilizations). This gap compounds as scale and utilization rise. Miners give away a piece of every dollar they earn while Nebius INCREASES gross margins over time. 2. Software Is the Moat and Oracle Proved It Even $ORCL, an $800B hyperscaler, failed at building GPU orchestration profitably, reporting ~14% gross margins on their AI rental platform and losing ~$100M in the process. If Oracle can’t execute, expecting $100M marketcap small miners like $SLNH to build hyperscaler-grade orchestration from scratch is wishful thinking. Platforms like Fluidstack are essential bridges, but they come at a cost: margin compression, revenue leakage, and platform dependency. Nebius? It already did the hard part. Its in-house orchestration software, GPU utilization, with no middlemen taking a cut. 3. Power vs. Margin Calculations A single H100 uses ~0.7–0.84 kW. Even at $0.10/kWh, power is just $0.07–0.08 per GPU-hour. When GPUs rent at $4–6/hr, that’s 1–2% of revenue. What actually moves the needle? Utilization: 50% -> 85% uptime = +70% revenue per GPU That’s a multi-hundred bps margin swing, far outweighing any "cheap power" comparison with miners. Software utilization/orchestration is a moat and matters ~30–70 TIMES more than cheap power. And it's not like $NBIS doesn't have cheap power either lol. 4. The CRWV Exception to Miners Everyone points to $CRWV as miners pivot but Coreweave literally spent YEARS to do this, along with billions in software acquisitions. And even now, reports suggest their utilization trails Nebius. If miners think they’ll replicate CRWV, good luck. I'd expect in NBIS to strong outperform in next year's earnings reports when it comes time for execution vs. speculation. 🧩 The Asymmetry Nebius GM: 70–75% (full-stack, 4yr depreciation) IREN/CIFR realistic GM: 40–60% (middleware, 2–3yr depreciation) Gross margin delta: 15–30 points Execution risk: Nebius = 0 (already doing it); IREN/CIFR = high Nebius isn’t a power play. It’s a software margins play, with hardware upside. With $IREN and others, you're guessing if it can pull off a $CRWV. People keep saying "software it not a moat, and look at $CRWV for $IREN future HPC margins", but if $ORCL one of the largest hyperscalers is stuck at 14% gross margins. Then how do we expect these small crypto miners to pull off a Coreweave. In terms of ASYMMETRICAL UPDATE, $NBIS is the clear favorite out of anything. In terms of raw potential upside if $IREN manages to pull a $CRWV out a hat and beats out $ORCL in this, sure I'd concede. Software full stack is a HUGE moat that people vastly, vastly underestimate. With $NBIS, they can just let it plug and play scale up over time with the highest gross margins in industry. $NBIS is what true asymmetric return looks like.

@KrownCryptoCave Congratulations 🎉🎊

the ultimate HODL

@RHouseResearch your $CIFR/$IREN take is total garbage. Miners data center skills and cheap power are perfect for AI …Sit down, watch and learn loser

$CIFR and $IREN investors seem to believe the “AI cloud provider” business model is “one of the best” businesses to be in. This, of course, is absurd. Running an AI cloud is extremely capital intensive, requiring huge outlays for GPUs (and data center CapEx if self-building). It’s also a very labor intensive business, necessitating deep pools of talent across hardware engineering, network engineering, networking, enterprise sales, etc. In addition, a significant portion of the potential customer base is start ups, who are likely to go bankrupt or get acquired. For these reasons, and others, running an AI cloud is very difficult to scale. $CIFR and $IREN investors seem to think the opposite, and that they’re investing in SaaS businesses that should be valued on ARR multiples. Probably most importantly, it’s an extremely intense and competitive environment. There are dozens upon dozens of neoclouds, in addition to the traditional hyperscalers (Amazon, Microsoft, Google) plus the emerging hyperscalers (CoreWeave, Oracle). I struggle to see how smaller upstarts intend to compete with CoreWeave, Oracle, Amazon, Microsoft, and Google over the longer term when all 5 have massive balance sheets and thousands of the world’s best engineers. $CIFR and $IREN investors seem to believe that because CoreWeave successfully “pivoted” from Ethereum mining, they can successfully pivot from Bitcoin mining. In my view, this is completely irrational for a variety of reasons. For starters, CoreWeave actually bet big on AI in 2021 before any of the traditional hyperscalers and had always intended on building a cloud offering. CoreWeave ordered $100MM+ of GPUs right before the launch of ChatGPT, and obtained data center colocation capacity in advance of the subsequent demand wave. As a scrappy start-up, CoreWeave had no choice but to automate every key internal workflow around the Neocloud business model which they pioneered. CoreWeave then leveraged their first mover advantage into long-term, 5-year contracts, which provide for attractive unit economics. The on-demand and short (one to two) year contract revenue model seen with other neoclouds leaves these businesses exposed to spot rental prices as their contracts roll-off. I suspect many neoclouds will likely never recoup their GPU investments. Coreweave, on the other hand, locks-in profitable economics over 4-5 year contracts and is thus not exposed to spot pricing. It’s difficult to say when, but I expect the neoclouds who are now just trying to scale up from a few thousand GPUs to be in for a rude awakening when investors actually look for them to deliver on the revenue and profits they’ve guided towards. Right now, investors seem to be rewarding them for promises alone. By all accounts, the technological and physical demands of smoothly and successfully operating larger clusters are incomparable to running a few thousand GPUs - these new neoclouds likely have no idea what they’ve signed up for.

Thank you to @DavidSacks, @chamath, @friedberg, @Jason, and the entire @theallinpod @allinsummit team for an incredible event! It was a spectacular three days connecting with business and technology leaders, learning about the exciting AI innovations being built, and how $IREN Cloud™ can help power the next big thing!

Jack Mallers gives the best argument ever on why #Bitcoin is SUPERIOR to Gold against Peter Schiff 🔥

$IREN's Next Step into Multi-Cloud (Many of my post long but all are hand written. This one is my coup de grâce.) Everyone is speculating on whether IREN’s next customer is Google, Meta or some other hyper-scale company. This is not the way to look at it. The first step to understanding what will happen is to understand Multi-Cloud. To explain Multi-Cloud, I’ll use an example none other than the Nebius-Shopify relationship. Shopify specifies exactly what their relationship is with NEbius: “we [Shopify] are partnering with GCP, as our main infra provider, and also engaging with neo-cloud providers such as Nebius to utilize large training clusters” (1). In other words, GCP is Shopify’s main cloud but Shopify is using Nebius for GPU intensive functions due to Nebius' lower GPU cost. Before any Nebius analyst jump up and say it’s not for lower cost, further down the Shopify blog post, Shopify says “Our tooling abstracts away from the underlying cloud provider, allowing us to move between vendors to get access to the latest GPUs as they're made available” (1). Shopify is a Tier 1 Tech Company as an E-commerce company but not a Tier 1 AI Research Lab, when Shopify says large training clusters, they mean training clusters are larger than inference but not the massive scale training cluster xAI, OpenAI, etc are building for their next big run. Now why doesn’t Shopify run everything on Nebius? Shopify needs GCP for user analytics, fraud detection, Spanner DB, customer data protection, Identity Services, Dataproc, etc. Let’s put it this way, hyperscalers are hyperscalers for a reason: the breath and depth of their software platform/services. Now why doesn’t Shopify use GCP for everything? GCP is “out” of GPU compute even though Google has already spent $75B in capex, or as Microsoft CEO Satya Nadella puts it, “I’m good for my $80B” (2). Note that GCP is not really "out" of GPU compute, but it’s got so much demand that GCP GPU time commands ridiculous prices. In other words, nothing is ever “out”, it just depends on the price. Hyperscalers are conscious in their capex in order to maintain their asset-light model and focus on high margins from IP/Software. Why doesn’t Google just build more datacenter faster and take that free business away Nebius? First, Google is max allocated the capex that it’s investors are comfortable with. Second, Google already maxed out its DC expansion capacity and operations teams. While GPU datacenters are not nearly as complex as TSMC fabs, just like TSMC fabs are not copy and paste, GPU datacenter that have 99.8% uptime are not copy and paste. When you guarantee 99.8% uptime, you are guaranteeing the long tail of hardware bugs. You get hardware bugs that show up on one board and not anther. Some bugs only show up above certain hardware temperatures. Some networking bugs only show up above certain bandwidth and/or certain packet types. Uneven levels of hardware wear across heterogenous and homogenous chips. Inverse logic for poorly documented driver code from the antique ages. Tracing voltage fluctuations all the way down the schematic. God Forbid EMI. Google has already assigned all its teams to build and operate its own datacenters and part of the bottleneck is the training of new teams. Colocation gives Google one of the most important component of power but doesn’t solve all the bottlenecks. The building and more importantly operating of GPU datacenters for IaaS once you have power is not copy and paste! @brianfry01 knows! Thirdly and most importantly, Google doesn’t want to get stuck with all the current gen GPUs - multi-cloud distributes the risk! So what’s the value of IaaS? Why is Microsoft signing a deal with Nebius? By extending Azure compute on top of Nebius IaaS, Microsoft gets X PaaS revenue and pays Nebius Y IaaS revenue where X > Y. X-Y is a cashflow machine for Microsoft without the capex burden, GPU obsolecence risk, and Microsoft like Google cannot build GPU datacenters fast enough anyways. The more important question is how can IREN participate in Multi-Cloud? $NBIS's Inference as a Service stack (3) is: DataOne (Power + DC Construction + DC Physical Operations) with Nebius (DC Design + IaaS + PaaS) = Full Inference Stack’s Value Chain. IREN’s official partner is TogetherAI (4) and TogetherAI’s Inference as a Service stack (5) will be: IREN (Power + DC Construction + DC Physical Operations + DC Design + IaaS) and TogetherAI (PaaS). Now why does inference need PaaS? Didn’t this Jim bro tell me bare metal GPU was the best? When you write inference code for an application, you need to optimize the GPU kernels, CUDA capture graph, and batching for that application. If you are T1 AI Lab, you tune this yourself, but if you are most other Software company working on a Software Application, you just want to focus on the Framework, API calls, DB Schema, fine-tuning your training model, but you don’t care how your inference is optimized as long as it’s optimize. The PaaS portion of “Inference as a Service” is a software service that fine-tunes the application’s inference and runs it on top of bare-metal GPUs. T1 AI Lab will tune it themselves and run on bare metal GPUs. In other words, PaaS is literally a service, not what inference runs on top of. TogetherAI+IREN = Nebius+DataOne and can serve as a AI Cloud in the Multi-Cloud ecosystem. Now here’s the banger. It doesn’t matter if IREN works with TogetherAI or AWS or Microsoft, to IREN they all serve the same purpose: the PaaS layer. Now if AWS/Microsoft is uncertain about IREN’s capabilities, IREN will prove it in its partnership with TogetherAI. In fact TogetherAI needs IREN more than AWS/Microsoft because TogetherAI doesn’t have its power, DC operations, etc, so TogetherAI will give better margins to IREN than AWS/Microsoft. But once AWS/Microsoft sees that IREN can be trusted as Y in its X-Y cashflow machine, it will want IREN IaaS to fight for PaaS market share since PaaS is made by deploying code (aka copy and paste). But just as Nvidia keeps Neoclouds as leverage, it may make sense for IREN to partner with TogetherAI for better margins and better leverage in the relationship. IREN+TogetherAI will be in position to serve as Inference/Small Training Cluster Cloud in Multi-Cloud for companies ranging from startups up to large companies like Shopify. As IREN builds credibility, IREN’s IaaS may be able to partner other strong PaaS/SaaS companies like Databricks. I’ll note that $CRWV PaaS > TogetherAI PaaS which might be the factor to push IREN to consider working with hyperscalers or Databricks, etc. Furthermore, IREN can work with model development focused startups like HumeAI; although HumeAI doesn’t have a PaaS, HumeAI can optimize the GPU kernels, CUDA capture graph, and batching for its own application. This is why Tim Delcourt is in SF: to court the TogetherAI and HumeAI and work with the next generation AI companies.

@bolander72 @BitcoinLayers 🎊🫡

Last year, I sent a cold DM to the guys at @bitcoinlayers. It ended up with them taking a chance on me and bringing me on as their first dev hire. Since then, I am proud of what we accomplished with the layers site: a complete redesign, expanded L2 project and token reviews, data visualizations, and much more. Not only that, we built @lemondropfi, a automation tool for @solana users to stack more sats. We helped users build up their BTC savings - bundled up in a compact, sleek browser extension. And while all things must eventually end, I want to say thank you to @redvelvetzip and @januszg_. It's been a wild ride. Looking forward, I am happy to share with you all now that I've joined the @taprootwizards to work on all things product and software. I am grateful for this opportunity. I’m ready for this next chapter building alongside the Wizards and carrying forward the same spirit of curiosity, craft, and conviction. Here’s to shipping fun, magical, and meaningful Bitcoin products together. 🧙‍♂️

$IREN is pleased to announce an expansion of its AI Cloud business. - Purchase of 1.2k air-cooled NVIDIA B300s & 1.2k liquid-cooled NVIDIA GB300s - Expands AI Cloud to approximately 10.9k NVIDIA GPUs - New liquid-cooling installation at Prince George underway - Additional $96m in GPU financing secured - NVIDIA Preferred Partner status Press Release: irisenergy.gcs-web.com/static-files/a…

$IREN is pleased to provide our monthly update for July 2025: - Record monthly revenue ($86m) and hardware profit ($66m) - AI Cloud expanded with 2.4k NVIDIA B200/B300 GPUs - Initial B200 shipment completed and fully contracted - 50MW Horizon 1 liquid-cooled AI data center on track for Q4’25 - Mackenzie fiber upgrade complete, further enhancing our ability to support continued growth of AI Cloud beyond Prince George - 1,400MW Sweetwater 1 civil and electrical works continuing Press Release: iren.gcs-web.com/static-files/d…

@udiWertheimer @opensea Yo, all you Fudders out there drop to your knees and bow to the Master Wizard! 🤣

smol update regarding @opensea and ordinals