Summary After Microsoft’s $9.7B and Nvidia’s $3.4B deals, I see IREN Limited’s perception shifting from an ordinary bitcoin miner to an emerging neocloud platform. In my view, power is the core asset, with secured grid-connected capacity rising to 5GW globally, including 4.5GW in North America. IREN targets $3.7B ARR by end-2026, supported by $3.1B under-contract ARR at Q3 FY26 exit. On the risk side, the current fundamentals remain weak, with Q3 revenue down sequentially to $144.8M and Bitcoin mining still driving most revenue. That said, I think the market may be willing to look beyond the current story, and if execution goes well, revenue should move higher from here. I own IREN stock, and I think it's a Buy. I am revisiting my bull case on IREN Limited ( IREN ) after the company reported Q3 FY26 earnings that confirm most of the arguments from the bears. The company is still relying on dilution to fund the AI buildout, revenue is still mostly coming from bitcoin mining, and the Nvidia deal looks circular. Despite that, I think the bears are dead wrong. In fact, the market reaction after Q3 FY 2026 earnings makes sense to me. In my view, what changed after the print was perception. In my view, IREN is no longer perceived as a bitcoin miner trying to attach itself to AI because the market is rewarding anything with GPU or power exposure. After the Microsoft (MSFT) and Nvidia (NVDA) deals, I think the company is being treated much more like an emerging neocloud infrastructure platform, alongside names like Nebius (NBIS) and CoreWeave (CRWV). This is the shift in perspective that I want to see in a stock whose current fundamentals look anything but exciting. Q3 FY2026 total revenue fell to $144.8m from $184.7m in Q2, with AI cloud revenue. On top of that, the net loss of $247.8m included $140.4m of non-cash impairment charges, largely tied to decommissioning mining hardware. This proves that the pivot is painful if one is looking at the current fundamentals. In my view, the market favors stories. The story of Iren is strong enough for this stock to be in my portfolio, although I still favor Nebius as a cleaner exposure to the neoclouds. That's my exposure to IREN is in the low single digit % of my portfolio's NVL. As for CoreWeave, I think it’s a similar case to Nvidia: as a leader in the space, most of the upside is priced in, and therefore, any small disappointment (like the Q2 guide of $2.45B and $2.6B, below the $2.7B consensus) may lead to a selloff. In any case, I am looking at Iren in this piece, and why I think it’s still a Buy. Why Is Iren Up After Q3 FY 2026 Earnings? In just a few words, I think it’s because perception has shifted from an ordinary bitcoin miner pivoting to AI “because AI is hot” to a player in the neocloud space, with multibillion-dollar contracts from Microsoft ($9.7B) and now Nvidia ($3.4B) , joining the likes of Nebius and CoreWeave. The big news in the Q3 earnings release is the five-year, $3.4bn AI infrastructure cloud services contract with Nvidia. According to the terms, IREN will provide managed GPU cloud services for Nvidia internal AI and research workloads using air-cooled Blackwell systems at Childress, Texas. I was quite surprised (maybe, not positively) that the deal involves Blackwell GPUs and not the next generation of AI accelerators: the Rubin. Looking at the financials, the company expects an average of $0.7B of annual revenue, which appears to be included in the company’s $3.1bn under contract ARR bridge at the end of last quarter. For reference, the company is targeting a $3.7bn ARR by the end of calendar 2026. That said, there is a second part that raised my eyebrow from the Nvidia deal: the strategic partnership . It seems that Nvidia will support IREN to deploy up to 5GW of Nvidia DSX-aligned AI infrastructure across IREN’s global pipeline, with Sweetwater expected to be the flagship site. How? Vendor financing. I am sure the bears will have a lot to discuss about circular deals and will likely reference charts like the one below (which is way outdated, as IREN is not depicted), along with their arguments: Bloomberg Regardless, Nvidia received a 5-year right to purchase up to 30 million IREN shares at $70, which could generate up to $2.1bn of proceeds if fully exercised. Did you also notice the fact that $70 is 16% above IREN’s closing price on May 7 ($60 after the pop post Q3 earnings)? The fact that Nvidia’s Jensen is willing to pay a premium speaks volumes about the valuation of IREN. More on this in a later section. Moving on to the November 2025 $9.7B deal with Microsoft, in the Q3 update, IREN said that roughly $1.9bn of the Q3-exit ARR was due to this deal. You remember that the Q3 exit ARR was $3.1B, right? Well, if I add up Microsoft’s $1.9B + Nvidia’s $0.7B, I get $2.6B. So, there is another half a billion in under-contract ARR that the company is not disclosing (likely due to confidentiality clauses). The only piece of information they are disclosing is that these $0.5B ARR are coming from the Prince George site. IREN In my view, as long as the hyperscalers are constrained by supply (mainly power, that’s why I’m deep into the power infrastructure theme in industrials, like GEV or VRT ), neoclouds will perform well. As soon as the music stops (for whichever reason) and hyperscalers’ compute supply catches up with demand, it would be the time to head for the exit and even short the same names, in my view. Moving on to the deployment roadmap, the company is targeting a 150,000-GPU fleet, including recent purchase agreements for more than 50,000 Nvidia B300 GPUs (with deployment across Mackenzie and Childress through H2 2026). That said, despite the compelling (at least, it sounds compelling to me) AI story, the P&L shows that the company is still an ordinary bitcoin miner: Iren As seen above, the bitcoin mining revenue in the last quarter was down 34% YOY, as the company decommissioned miners ahead of GPU installation and billing. Excluding D&A effects, the margins of the AI cloud services segment look decent, currently sitting at 86% (of course, this is a non-GAAP figure). Finally, I have to discuss power, which is, in my view, the core asset of the company. Let’s go back in time for a second. As of June 2025, IREN had 2,910MW of secured grid-connected power, including 810MW of operating data center capacity. Well, the secured grid-connected power jumped to 5 GW globally by May 7, 2026, with 4.5 GW in North America and 0.49 GW in Spain after the Nostrum acquisition. The major change versus the older 2,910MW figure is due to the energization of Sweetwater 1 (1.4 GW) earlier in May. As a side note, this is a planned 2GW campus. So, how much of that secured grid-connected power is making money? Well, here is where things get disappointing. Risks Only a small portion of IREN’s 5GW secured power portfolio is currently producing AI services revenue. Without looking at the P&L statement, one can easily infer that, based on management’s 2026 AI Cloud capacity target of only 480MW. IREN As seen above, the Q3 FY26 update says that the 2027 expansion to 1,210MW of AI Cloud capacity is in build. The chart below from the latest earnings slide deck shows the magnitude of the problem: IREN The neocloud growth story of IREN is barely in its pajamas, and I think that may spook some investors with shorter horizons or those that want to wait until the AI services segment kicks in. Seeking Alpha I included above the Street’s estimates for future revenue, which clearly show the ramp later this year, reaching triple digits YOY growth by the September 2026 quarter. Until then, the company is still (from a technical point of view) a mainly bitcoin miner. That said, whether the market is perceiving it as such is a different matter. The other risk factor to consider is, how not, funding. You remember my circular deal chart in the last section. I would expect an updated version of that chart to circulate soon this year, which is likely to deteriorate the sentiment in the AI trade. The balance sheet is anything but encouraging. By the end of last quarter, convertible notes were $3.69bn, and total liabilities were $4.60bn. On top of that, IREN also raised equity during Q3, with $380m of ordinary share issuance. Iren In my view, despite the fact that IREN had secured $9.3bn of funding over the prior eight months from customer prepayments, convertible notes, GPU leasing, and GPU financing, I think dilution is still part of the execution equation. With cash and equivalents sitting at only $2.6bn (April 30), I think dilution is only going one way from here: up. Looking Ahead In my view, the bull case leans on two milestones. The first one is AI cloud ARR conversion. I want to see how the company turns its $3.1bn under-contract ARR and >$3.7bn year-end 2026 ARR target into commissioned GPU capacity and, therefore, revenue. The current P&L looks anything but encouraging, with Bitcoin mining still accounting for a large part of the operating income. Based on the market reaction post earnings, I think the market is willing to look ahead at the upside from the Microsoft and Nvidia deals, and not at the current results. On the deployment side, the 5GW secured power portfolio only creates equity value if grid-connected sites become usable AI data centers. Execution is key, and any delays will likely move the revenue recognition to the right, similar to CoreWeave’s delay, followed by a downgrade in FY25 revenue guidance, back in Q3 last year. Despite being a grotesquely overpriced stock , trading at 192x forward earnings, and the fact that, today, the company is still making most of its revenue from bitcoin mining, I own IREN in my portfolio due to its compute deals from Microsoft and now Nvidia.
