People don't realize how significant this is for $IREN.
IREN just signed the transmission agreement for its first Australian data center campus: 800MW in Bundey, South Australia. One of the largest projects announced anywhere in APAC. The market hasn't priced in what this actually means.
1/ Land nobody else has.
Australia is enormous and mostly empty. You can build at hyperscale without sitting on top of a population center. In Europe and the US, data centers keep getting delayed or killed outright, mostly because communities protest and don't want them nearby. In the outback, that resistance doesn't exist.
2/ Power is the real bottleneck, and they solved it
The site sits in a hot, dry region similar to Texas, which the industry already favors. But the energy story is the kicker. South Australia is targeting 100% net renewable energy by 2027. Abundant clean power is exactly what AI compute runs on, and most operators are fighting for it everywhere else.
3/ This is a renewable energy story too
A project generating this much demand for clean power pulls forward serious investment into wind, solar, and the grid itself. The data center and the renewable buildout reinforce each other. Both sides win, and the whole region benefits.
4/ Stability is underrated
Australia is one of the safest, most stable countries on earth. No conflict, no political risk hanging over the asset. When you're building infrastructure meant to run for decades, that security is worth a premium.
5/ Wired into Asia
The campus has submarine fiber running straight into Singapore, Indonesia, South Korea, and Japan. So it isn't isolated in the desert. It's connected directly into the fastest-growing AI demand centers in the world.
6/ The government is behind it
South Australia's Premier, Peter Malinauskas, came out directly backing the project, pointing to the state's renewable leadership, record investment in higher education, and the appointment of the nation's first dedicated Minister for AI. You cannot build at this scale without the state working with you instead of against you. They want it.
Bottom line:
> Land.
> Clean power.
> Stability.
> Connectivity into Asia.
> Full government support.
Australia is shaping up to be one of the best places on the planet to build AI infrastructure, and IREN is planting its flag there first.
43.35% APY won’t last forever. ⌛
Rates fluctuate, and staying on top of every change is a lot of work.
Let me handle it for you. I’ll keep your deposits earning the best rate available on @Base, with automatic compounding in the background.
43.35% APY won’t last forever. ⌛
Rates fluctuate, and staying on top of every change is a lot of work.
Let me handle it for you. I’ll keep your deposits earning the best rate available on @Base, with automatic compounding in the background.
The Market’s Current Obsession and $IREN’s “Bottleneck-First” Principle
Today, IREN announced that it has signed a transmission connection agreement to support the development of an 800MW data center campus at Bundey in South Australia. The project has secured a high-voltage transmission connection to a utility substation and is expected to begin energization in 2028. Through submarine cable connectivity, it will be able to serve major demand centers across the Asia-Pacific region. The power supply will be primarily supported by renewable energy sources.
Under the agreement, IREN has secured access rights to four 330kV feeder bays at the utility substation, providing support for up to 800MW of load capacity without requiring grid upgrades. Subject to regulatory approvals and the satisfaction of conditions associated with the transmission connection agreement, IREN intends to commence early-stage construction activities and equipment procurement.
Several key points stand out from this announcement:
No grid upgrades required means the project can be deployed as quickly as possible, making this aspect particularly valuable.
South Australia’s high penetration of renewable energy provides long-term access to green power.
The 800MW scale creates significant economies of scale—and this is only the beginning.
Australia serves as a launch point for reaching the broader Asia-Pacific market.
Overall assessment: considering scale, expandability, environmental sustainability, grid integration, geographic reach, population served, and economic significance of the surrounding region, this is an A+ class scarce asset.
IREN’s current portfolio of 5.8GW power sites broadly meets the criteria of A+ scarcity assets. AI data centers are increasingly becoming an important financial asset class and will gradually emerge as a key valuation anchor. At today’s relatively early and unsophisticated stage, the market has not yet fully recognized this reality. The prevailing attitude is still “as long as there is power available, that’s enough.” However, sovereign AI initiatives are already demanding much more than that, and the value of these attributes is likely to become increasingly important as the sovereign AI market develops.
Sovereign AI is not limited to governments. It revolves around ownership and control of data. Industries such as finance, healthcare, energy, and defense cannot tolerate data leakage or dependence on opaque cloud infrastructure, and are increasingly moving toward sovereign computing environments. This migration shifts AI data centers with sovereign-AI characteristics away from low-margin customers and toward strategic customers with larger budgets and stricter compliance requirements, creating substantial premium opportunities.
Through segmented compute zones and physical isolation, operators can transform closed environments into certainty and trust, providing localized, high-value operating environments for financial institutions, governments, healthcare organizations, and advanced foundation-model developers. As a result, data centers can move beyond commoditized competition and become high-barrier sovereign infrastructure.
These assets provide not only large-scale dispatchable green energy but also a form of irreplaceable exclusivity. Once such sites become foundational infrastructure for sovereign AI training, their asset characteristics may eventually support securitization.
Because these assets combine multiple forms of physical scarcity, each one secured reduces the pool of available opportunities. They are fundamentally difficult to replicate. Due to the background and experience of IREN’s management team, they recognized this years ago. Site selection efforts were conducted globally long before the market appreciated these dynamics. Work that began eight years ago has evolved through continuous learning and industry development into a highly refined understanding of what constitutes a truly valuable site.
This is fundamentally different from today’s industry behavior, where many participants simply pursue any available power source without regard for broader considerations. The commercial outcomes of these two approaches may ultimately prove dramatically different.
In a recent article that I refer to as the “IREN Development White Paper,” CEO Daniel Roberts argued that these scarce, high-quality sites have already been largely secured by early movers. New entrants may soon discover that even if they are willing to wait, sites of comparable quality are simply no longer available. Even paying a premium lease rate may not be enough to gain access.
That opportunity window has effectively closed. Over the next two years, additional windows may close as well. The ladder leading to the top tier of infrastructure operators with genuine pricing power is gradually disappearing. While AI still presents many opportunities, much of the remaining space is likely to become increasingly commoditized and competitive.
The market today has not fully interpreted these dynamics. As a result, $IREN’s stock merely finished positive while broader indices declined. The market’s current obsession is straightforward: companies such as $NBIS that continue signing contracts with hyperscale customers are viewed as the winners.
Questions such as:
How much power has actually been secured?
What is the source and quality of that power?
Does the project depend on third-party partners?
Could delays occur?
Is it environmentally sustainable?
Could environmental objections force a redesign?
Will more expensive behind-the-meter solutions eventually be required?
are entirely ignored.
As long as contracts are being signed, early access to systems such as Vera Rubin is secured, and high-profile supporters are involved, investors are eager to buy.
Meanwhile, IREN’s advantages in scarce resource positioning and cost structure receive far less attention. Some NBIS supporters argue that these advantages can simply be purchased later—that NBIS could buy IREN, lease IREN’s facilities, or otherwise replicate these capabilities. As amusing as that may sound, many people genuinely believe it.
Markets naturally focus on what is immediately visible. Looking purely at reported financial results, NBIS currently presents stronger numbers, while IREN has missed expectations. Share-price divergence under those circumstances is understandable. But if investing were only about reading surface-level metrics, the process would be much simpler than it actually is.
Anyone familiar with IREN during the Bitcoin mining era will recognize a recurring pattern. In HPC infrastructure, being first is not always advantageous.
For years, IREN remained a secondary player in mining. It waited until its own data center infrastructure was complete and mining hardware efficiency had improved dramatically. Only then did it move aggressively, eventually leaving competitors behind and becoming the only consistently profitable miner.
The AI cloud industry is significantly more complex, but the principle remains similar.
Consider CoreWeave. It moved fastest, but is its current situation ideal? Its debt has been rated below investment grade, resulting in high borrowing costs. Regardless of how attractive the industry may be, many sophisticated investors avoid businesses that develop credit-related concerns. Operational mistakes can be forgiven; credit problems are far less acceptable.
NBIS currently appears to be performing best. But is it truly operating optimally?
Forget software advantages for a moment and focus on basic business logic. Its own secured power resources are relatively limited, yet it has already committed portions of those valuable resources to customers through bare-metal contracts. While positioning itself as a full-stack cloud platform, it has sold its most important resource at bare-metal economics.
By contrast, IREN is often viewed as the pure bare-metal provider despite the fact that it has not sold a single watt.
At the same time, NBIS relies heavily on third-party power arrangements. Environmental approval issues have already forced alternative plans, increasing costs while still leaving uncertainty regarding successful execution. Similar challenges may continue to emerge in the future.
Forget about how ‘brilliant’ this team claims its tech is — just look at these two decisions. Most corner‑shop owners would’ve shown better judgment. Running a business doesn’t magically change just because you sprinkle some AI on it.
IREN has returned to a position of patience. It has spent more than six months without signing major contracts. During that time, The unit price for GPU compute leasing has surged by roughly 30% compared with the levels when CRWV and NBIS signed their contracts, while demand remains extremely strong.
Yet IREN continues to wait.
Its approach is different. I call it the “bottleneck-first” principle: solve the hardest problems first, and you gain both optionality and pricing power.
Bottleneck #1: Power Resources
IREN began working on this challenge years ago and focused on securing high-quality assets rather than simply accumulating capacity. Those efforts are now beginning to produce visible results.
Over the past six months, investors have watched IREN nearly double its portfolio of premium power assets across major economic regions worldwide. I believe we will continue to see additional sites secured.
Bottleneck #2: High-Performance Data Center Technology
This encompasses rack density, liquid cooling, orchestration, networking, topology optimization, operational management, and overall performance at gigawatt scale.
IREN has now advanced to the point of partnering with NVIDIA to build the DSX AI Factory, which is expected to become the flagship deployment site of its kind.
Bottleneck #3: Large-Scale Compute Deployment
IREN now controls approximately 5.8GW of power capacity. Much of this infrastructure can be energized before 2028. Leaving such resources unused would represent an enormous waste.
For IREN, the pressure to activate power capacity is as real as Anthropic’s accelerating demand for compute. The challenge is not securing symbolic advantages such as being among the first suppliers to receive Vera Rubin systems. The challenge is deploying GPUs safely and efficiently at massive scale.
That is why IREN is working with BE to use digital-twin technology to evaluate the performance characteristics of a 50,000-GPU Blackwell deployment before large-scale implementation.
These issues are far more important than near-term contract announcements or short-term AI revenue growth.
The relationship between markets and Companys is often like the relationship between a person and a dog on a leash. The key question is whether the dog is dragging the person, or the person is leading the dog.
Those things that please the market in the short term are often nothing more than a trap.
@tomyoungjr "(Bitcoin has an) almost insurmountable edge over other alternatives"
I dunno. Maybe. We do know it is flawed, can be forked and probably improved upon in ways no human has thought of yet. Good luck.
@tomyoungjr Quantum - why risk it at all when there is (possibly) quantum resistant crypto available now or in the near future?
Transactions - How about we skip the 'Bitcoin settlement layer' completely and remove any counter party risk? One ledger for all transactions.
@tomyoungjr Perhaps it starts with the bitcoin protocol and improves it? E.g. quantum resistance, fast(er) tx finality, does not compete for electrons (as much).
My point is Bitcoin is not preordained to 'win' AI.
IREN — The Overlooked Massive Significance of DSX Air: A Supplementary Explanation
@TrentBuysValue
I like the thought process. But not sure if IREN will have anywhere to operate VRs right now. If we are the test bed for VR we’d be testing them now as production is ramping right? Didn’t Jensen say VR is in full production. With press release “ramping into full production”. If that’s the case these will be running in DCs before IREN has SW up and running.
Thank you for your detailed reply and sharp questions — they are completely fair and give me a great opportunity to lay out the full picture.
Why not just test with Vera Rubin directly?
The core logic of Digital Twin simulation is this: a virtual model isn’t “making up a Rubin from nothing.” It must be calibrated against real physical hardware to be truly reliable. That’s exactly why IREN is currently running a 50,000+ Blackwell Ultra digital twin instead of jumping straight to a Vera Rubin simulation.
Here’s why:
1. There is currently no large-scale Vera Rubin hardware deployed anywhere in the world
Two days ago (May 31, 2026 at GTC Taipei), NVIDIA officially announced that Vera Rubin has entered full production / ramping into full production. But “ramping production” does not equal “mass delivery to customers.”
The official wording is clear: volume shipments only begin in H2 2026 (this fall). Right now only the supply chain (TSMC, server OEMs, etc.) is manufacturing at scale.
No hyperscaler or AI cloud provider has a 50,000+ physical Vera Rubin cluster up and running yet. Early samples may exist, but they are nowhere near the scale needed to generate meaningful real-world operational data.
Therefore, trying to build a high-fidelity digital twin of Vera Rubin without large-scale real hardware would result in lower simulation accuracy.
2. IREN’s 50,000+ Blackwell GPUs are “about to go live” — real, contracted hardware
IREN’s current contracts (with Dell and others) call for large-scale Blackwell Ultra deployment first. DSX Air digital twin is precisely the tool being used before the physical data center is built and the GPUs are racked — to fully validate the entire AI factory stack: network topology, automation workflows, storage, orchestration, security, etc.
The official press release (IREN + BE Networks, June 1, 2026) explicitly calls this a “production-representative digital twin” designed specifically for IREN’s upcoming 50,000+ Blackwell Ultra deployment.
3. Mutual calibration between virtual and physical is the real power of DSX Air
NVIDIA DSX (including DSX Air + Omniverse DSX Blueprint) was designed from day one for the Vera Rubin era (the Vera Rubin DSX AI Factory reference design was released in March). The actual workflow is:First use real Blackwell hardware for calibration → run simulations → collect massive telemetry data on GPUs, networks, power, failures, etc. → make the digital twin extremely accurate.
Then layer in NVIDIA’s official Vera Rubin reference design (GPU, Vera CPU, NVLink 6, Spectrum-6, etc.) → instantly upgrade the already-calibrated “factory production line” to Rubin specifications.
Result: When Vera Rubin hardware arrives at scale, IREN’s Sweetwater 2GW flagship site (NVIDIA’s designated DSX architecture flagship) can achieve near plug-and-play deployment with far superior speed and reliability.
This is exactly like building airplanes: you first run full-process digital twin rehearsals with the current engines (calibrating every manufacturing step, automation, and supply chain).
When the new engines (Rubin) arrive, you simply apply the already-mature production line.NVIDIA itself emphasizes: CoreWeave and other partners are using DSX Air to run “operational rehearsals well ahead of physical delivery.”Special note on “operational rehearsals”:
This is the part CoreWeave (CRWV) and Nebius (NBIS) are handling — they are the first to receive and run Vera Rubin hardware in the real world, getting it up and generating revenue early.
That’s one side of the coin. The more strategically important part — the one IREN is using DSX Air to run production-grade digital twin testing on its 50,000+ Blackwell GPUs that are about to go live. IREN is validating the entire end-to-end “installation and operations production line” (design → simulation → deployment → operations) for the Rubin era at massive, replicable scale.
CoreWeave/Nebius getting hardware first and IREN building the production-grade digital twin with real, soon-to-be-live Blackwell hardware are two sides of the same coin — different phases, same goal: scaling Vera Rubin.
IREN has 2GW of already-powered sites ready. That positions it to do the heavier, longer-term, non-replicable work: establishing the standardized, infinitely replicable “AI Factory” template for the entire industry.
This is the real core of NVIDIA’s “AI Factory” moat — not who gets the first batch of new GPUs. Plenty of companies can get early Rubin hardware. But only IREN is doing the next-level, large-scale deployment testing and standardization of the full production line (design-simulation-deployment-operations). That cannot be replicated overnight.
Does this significance far outweigh simply being first to take delivery?
Absolutely. It just requires a deeper level of understanding. The market may take a little time, but once it clicks, the reaction will be the right one.
Here's how the @Mamo Drop works...
Every week, @Mamo passes 100% of the revenue it generates to MAMO depositors. The whole amount, paid in Bitcoin (cbBTC) and MAMO. If you've already got MAMO deposited, you're earning a slice of what the platform brings in.
Mamo's edge is market-making on @AerodromeFi, the exchange with the highest trading activity on @Base. It supplies liquidity across the pairs people trade most, putting the $MAMO token next to assets like $cbBTC, $ETH, $VIRTUAL, $USDC, $EDGE, $KTA, and plenty more. Every trade through one of thesse pairs earns a fee, and over the week they stack up into a payout for MAMO depositors (the Mamo Drop).
Why do some weeks pay more than others?
Revenue moves with the market, so earnings vary. Active weeks mean more trades, more fees, and a larger Mamo Drop. Quieter weeks bring less of all three. I've seen it reach as high as $110K in a single week.
What you earn is a direct share of fees from real trading, paid out week after week. It grows as the network grows, so when @Base and @AerodromeFi get busier, depositors benefit directly.
As Mamo earns, and so do you. gmamo 🌱
@LukeYoungblood@MoonwellDeFi@mamo True, you discussed Mamo. Thanks.
However, you did not talk about any new, coming agentic features. I cannot begin to express how disappointing that is at this point.
Is there anything new, agentic coming soon? When?