IREN Limited: A Deep-Dive Long-Term Value Investment Study

Hunting ten-year five-baggers among great growth stocks — pressing the upside question: "Can it get much bigger?"

• QHow high is its market ceiling? Is it enlarging a slice of an existing pie, or creating an entirely new market?5/10

  A

  The ceiling is genuinely high, but IREN is grabbing a slice of supply-side share within a "big pie" dominated by a few giants, rather than pioneering an entirely new market. This distinction is key to Baillie Gifford's "5x in ten years" judgment: the denominator (TAM) is large enough, but IREN's position in the numerator is that of a new entrant that does not hold pricing power on the demand side.

  First, how big is the pie. The report cites the International Energy Agency's assessment that global data-center electricity consumption will rise to about 945 TWh by 2030, roughly double the 2024 level, with AI as the core driver. This magnitude is consistent with the IEA "Energy and AI" report, whose official estimate is that global data-center electricity demand doubles to about 945 TWh by 2030. So the long-term ceiling of the "AI infrastructure + power" track is genuine, structurally rising demand rather than short-term thematic hype.

  But what IREN does is "enlarge a slice of an existing pie," not create a new market, for three reasons:

  First, what it sells is GPU cloud compute and power/data-center capacity, a market already defined by Microsoft, AWS, Google, CoreWeave, Oracle, and others; IREN is a latecomer and a supplier. The report itself admits its customers "care more about on-time delivery, compute pricing, SLAs, and power and cooling reliability" rather than having stickiness to the IREN brand. This is precisely the hallmark of a commoditized supply market, not that of a new-market pioneer.

  Second, the money on the demand side is actually the giants' money. As of May 7, 2026, IREN disclosed "signed ARR of about $3.1 billion," made up of about $1.9 billion from Microsoft, about $0.7 billion from Nvidia, and about $0.5 billion from the Prince George GPU deployment. In other words, IREN's "ceiling" depends heavily on wholesaling capacity to a few mega-customers; it is taking orders within the giants' capex flood rather than opening up an independent demand pool of its own.

  Third, the ceiling of the other leg (bitcoin mining) is instead moving downward. The report discloses that the company actively shrank installed hash rate from about 50 EH/s at the end of June 2025 to about 38 EH/s at the end of March 2026 to make way for AI; and the net electricity cost per bitcoin has risen from $18,127 in FY24 to $25,642 in FY25. Mining is being structurally squeezed out rather than "enlarging the pie."

  Conclusion: the denominator (AI power/compute TAM) has an extremely high and real ceiling, which is the premise for IREN's story to hold; but IREN's role is "to enlarge its own slice of supply share within an already-formed large market by grabbing power and delivery windows," not to create an entirely new market. In Baillie Gifford's terms, this means upside is not lacking; what is lacking is exclusivity evidence for "why IREN, rather than players with more capital and customer relationships, will capture this increment." And on this point the evidence is currently insufficient.

  Jun 10, 2026
• QOver the next five years, can its revenue at least double? Is growth driven mainly by volume, price, or new business?6/10

  A

  It can almost certainly double, and far more than double; but the "certainty" of this growth is far lower than the growth rate itself, and it is driven mainly by new business (AI Cloud starting from near zero) rather than by simultaneous volume-and-price gains in a mature business. This is the side of IREN that most resembles a Baillie Gifford growth stock, and also the side that most needs a discount.

  First, the scale of the growth. The report discloses revenue of $75.5 million in FY23, $187.2 million in FY24, $501.0 million in FY25, and $569.8 million in FY26's first nine months; that is, nearly 7x growth in two years. Looking only at official quarterly data: in Q1 2026 (ended March 31), total revenue was $144.8 million, of which AI Cloud revenue was $33.6 million, nearly doubling from $17.3 million in the prior quarter. AI Cloud nearly doubling quarter-over-quarter is real acceleration.

  More critically, the answer to "doubling over the next five years" is hidden in already-signed contracts. The company disclosed signed ARR of about $3.1 billion (Microsoft about $1.9 billion, Nvidia about $0.7 billion, Prince George about $0.5 billion), with an internal-assumption target pointing to a $3.7 billion ARR goal (based on about 74,000 GPUs deployed across BC and Childress). Against FY25 full-year revenue of $501 million: even realizing only the signed $3.1 billion ARR would put revenue at more than 6x the current level. So "doubling in five years" for IREN is a question of whether it will be discounted by execution/delivery, rather than a question of whether it happens at all.

  What drives the growth? Breaking down the three:

  • New business (AI Cloud) is the absolute main engine: from $3.1 million in FY24 and $16.4 million in FY25 to $58.3 million in FY26's first nine months, and then to several billion in ARR on a contract basis, almost all the increment comes from this new curve.
  • Volume (capacity coming online) is the core variable: the company targets expansion to 480MW in 2026 and 1,210MW under construction in 2027, ramping toward 5GW over the longer term; growth is essentially the process of "gradually installing, powering up, and billing the already-secured 4.5GW+ of power."
  • Price (mining) is a negative contributor: bitcoin-mining revenue fell from $167.4 million in the prior quarter to $111.2 million this quarter, dragged down by a declining average BTC price and the active retirement of mining rigs; it is a speed bump for revenue rather than an accelerator.

  But we must honestly assess "the quality of the doubling": while revenue multiplied several times, Q1 2026 still recorded a $247.8 million net loss (including $140.4 million of impairment), and on the report's basis the FY26 nine-month operating loss widened to $426.3 million. In other words, the revenue doubling is highly real, but it is "scale bought with extremely high capex and continuous financing" rather than "growth compounded by unit economics on its own."

  Conclusion: doubling revenue in five years is almost the floor scenario (the signed contracts alone far exceed a doubling), the driver is the capacity ramp of the new AI Cloud curve (volume-driven mainly), and mining is a negative contributor. By Baillie Gifford standards this is a qualified "growth engine," but the quality of the growth must be discounted: it has not yet proven it can convert revenue into distributable cash without relying on large-scale dilution.

  Jun 10, 2026
• QFive years from now, what will take the baton as the next growth engine? Does this "second curve" exist today?5/10

  A

  The second curve genuinely already exists today, and it is precisely the one drawing the most attention right now: AI Cloud. But the issue is that IREN's "first curve" (bitcoin mining) is being actively cut back by the company itself, so the more accurate statement is: the second curve is replacing the first curve, while the true "third curve" (the next growth pole to take the baton from AI Cloud capacity wholesaling) is still unclear today. This is an important structural difference between IREN and a typical Baillie Gifford growth stock.

  First, confirm the second curve has landed and is not just a slide deck. The report discloses that AI Cloud is already the company's "strategic priority," with evidence in the form of concrete contracts and hardware: a roughly $9.7 billion five-year contract with Microsoft (using NVIDIA GB300, 200MW of critical IT load, delivered in four phases through 2026), and a roughly $3.4 billion five-year AI cloud contract with Nvidia + a 5GW strategic partnership + the right to subscribe for up to 30 million shares at $70/share (about $2.1 billion) within five years. About 150,000 GPUs are installed or in transit. So for the question "does the second curve exist today," the answer is a clear "it exists and is already at scale."

  But to answer "what takes the baton in five years," we must see clearly IREN's current curve structure:

  • The first curve (mining) is being actively shrunk: installed hash rate fell from about 50 EH/s at the end of June 2025 to about 38 EH/s at the end of March 2026, and mining revenue fell from $167.4 million in the prior quarter to $111.2 million this quarter. It is an asset being deliberately retired rather than "a stable cash cow that steps back to second place after handing off the baton."
  • The second curve (AI Cloud capacity wholesaling) is all of the current growth: targeting $3.7 billion ARR, with capacity ramping from 480MW (2026) to 1,210MW (2027) and then to 5GW (2028+).

  The third curve is not clear today, and this is an honest shortfall. Following Baillie Gifford's probe of "who takes the baton in five years," IREN has two possible baton-passing directions, but both are still embryonic: first, moving up from "selling raw compute / capacity wholesaling" to higher-margin "hosting + networking + software platform" services. The company's 2026 acquisitions of Mirantis (cloud-native software) and Nostrum vaguely point in this direction, but AI Cloud's revenue share is still extremely low ($58.3 million in FY26's first nine months vs $569.8 million in total revenue), and whether platformization can take shape is unproven; second, turning the 4.5GW+ of power resources themselves into a repeatedly monetizable "energy + data-center REIT-style" asset platform. Neither of these has yet formed a quantifiable second/third growth pole; they are more like "options" within the report's and management's narrative.

  Conclusion: the second curve (AI Cloud) conclusively exists today and is already the main engine, which is a plus; but it is replacing rather than layering on top of the first curve, meaning IREN has no stable, mature business to fall back on. The true third curve (what continues to drive a 5x after the AI Cloud capacity ramp) has no clear shape today. By Baillie Gifford's standard of "concentrating firepower on years 3–10," this imaginative space exists but the evidence is insufficient, so it cannot warrant a high score on that basis.

  Jun 10, 2026
• QWhat is its core competitive advantage? Will this moat widen or narrow over the next three to five years?4/10

  A

  IREN's core advantage is "large-scale secured power + land + self-built data centers + delivery capability validated by top-tier customers," a genuinely existing asset-type/entry-type moat. But it is more like "a ticket that gets you onto the track" than "a moat that earns you excess returns over the long run." Over the next three to five years it will most likely widen (capacity and contracts are expanding), but the "type" of moat is hard to upgrade, and stickiness and pricing power remain weak. The report's Moat Strength Score of 2/5 is a low judgment I agree with.

  First, where it is genuinely hard. IREN's most respectable barrier lies in power rather than brand, network effects, or patents. The report discloses that the company's secured grid-connected power rose from 2,910MW at the end of June 2025 (about 95% in Texas) to about 4,510MW at the end of March 2026, and outside parties have corroborated its data-center portfolio with more than 4.5GW of locked-in power resources. At a time when AI compute is constrained by "where the power comes from," in-hand grid-connected power + operational campuses + built-out data centers genuinely take time, approvals, capital, and local grid relationships to replicate. This is a real threshold. Microsoft and Nvidia signing on in turn also proves that external customers treat IREN as "a platform with real delivery capability" rather than a concept stock.

  But this moat has three hard limits that determine why it cannot deliver high-return certainty:

  First, it is an entry barrier rather than an excess-return barrier. Securing power lets you enter and take orders, but it does not guarantee you can maintain high gross margins in competition with Microsoft, AWS, CoreWeave, Oracle, and others. The report clearly points out that customers "do not necessarily have strong stickiness to IREN itself; they care more about on-time delivery, compute pricing, SLAs, and power and cooling reliability." This is a hallmark of commoditized supply.

  Second, both switching costs and network effects are weak. The report's self-assessment: weak brand, no network effects, low-to-moderate switching costs, weak channels, weak data advantage. GPU capacity is inherently substitutable, and when contracts expire customers can switch suppliers or turn to self-building. This is not the same thing as the "the more you use it, the more you depend on it" network/platform moat that Baillie Gifford favors.

  Third, the industry does not lack rivals willing to spend to replicate. AI infrastructure is an arms race of capital and power, and players such as hyperscale cloud providers and Oracle/CoreWeave have stronger capital and customer relationships than IREN. The power advantage will be continuously chased down.

  Will it widen or narrow over the next three to five years? Directionally it will widen: capacity ramps from 480MW (2026) to 1,210MW (2027) and then to 5GW, signed ARR of about $3.1 billion, targeting $3.7 billion, with scale and delivery track record both accumulating, and the deep binding with Microsoft/Nvidia (Nvidia even took the right to subscribe for up to 30 million shares at $70/share) will strengthen its positioning within the NVIDIA ecosystem. But "widening" refers to added depth in the same type of moat (asset-type/scale-type); the moat's type is hard to upgrade into a high-stickiness platform type, unless software layers such as Mirantis genuinely grow independent switching costs, which cannot be seen today.

  Conclusion: the moat is real but low-quality: an asset-type entry barrier of "power + delivery" that widens as capacity expands, yet struggles to convert into pricing power and customer stickiness. By Baillie Gifford's high standard for "a durable moat that can convert into excess returns," IREN falls short, and the low 2/5 score is honest.

  Jun 10, 2026
• QIf its core business is disrupted, does it have the DNA to reinvent itself? How does it treat mistakes and bad news?5/10

  A

  Yes, and IREN's "self-reinvention DNA" is precisely the most validated item in its entire story: it is itself a successful, drastic act of self-disruption, actively transforming from a pure bitcoin miner into an AI Cloud infrastructure provider. Its attitude toward mistakes and bad news also leans candid (openly disclosing restatements, internal-control weaknesses, tariff and GST disputes, and future impairments). But this DNA is that of "entrepreneurial expansion" rather than "steady harvesting," and what Baillie Gifford must watch for is that its reinvention capability is strong yet comes at the cost of continuous shareholder dilution.

  First, the "DNA evidence" for self-reinvention. Baillie Gifford asks this question to learn whether a company can find a new life when its core business is disrupted. IREN is almost a textbook positive case: instead of clinging to the structurally deteriorating mining business (net electricity cost per bitcoin rose from $18,127 in FY24 to $25,642 in FY25), it actively cut installed hash rate from about 50 EH/s to about 38 EH/s, retired mining rigs, and freed up power and data-center space to install GPUs, winning contracts of about $9.7 billion with Microsoft and about $3.4 billion with Nvidia within two years. Being able to actively "disarm itself" and pivot to a better track while its own cash cow is still generating money is itself the strongest proof of reinvention DNA. Many miners cannot do this.

  Next, how it treats mistakes and bad news: leaning positive but requiring a discount:

  • Candid disclosure: the report notes that the company openly disclosed historical financial-statement restatements, material weaknesses in internal control, a roughly $100 million U.S. import-tariff dispute, a roughly $27.7 million Canadian GST dispute, and roughly $520 million in possible future additional impairment. Not hiding landmines is a good governance signal.
  • Genuine remediation: the report cites the 2025 10-K that the historical material weaknesses have been remediated and the auditor has confirmed internal control is effective. Mistakes were corrected rather than concealed.
  • Bad news recognized honestly: in Q1 2026 it directly recognized $140.4 million of impairment (mainly retired mining rigs), without whitewashing the book cost of the transition.

  But we must honestly point out the other side of this DNA. IREN's "reinvention" is driven by extremely aggressive capital operations: ATM issuance, convertible notes, equipment financing, prepayments, and M&A all together. The report discloses that common shares ballooned from 64.75 million at the end of June 2023 all the way to 341 million at the end of March 2026, and after the quarter it issued another 15.88 million shares to raise $683 million; in May 2026 it further issued $3.0 billion of 1.00% convertible notes due 2033, netting about $2.96 billion. So its self-reinvention capability is strong, but the cost is that old shareholders are repeatedly diluted. This is the entrepreneurial DNA of "daring to bet and daring to pivot," rather than the capital-discipline DNA of "cherishing per-share value."

  Conclusion: the self-reinvention DNA is clearly a "yes," already validated by one successful transformation, and its attitude toward mistakes and bad news is candid. These two points are genuine pluses for IREN and traits Baillie Gifford would appreciate. But the demerit is that the base color of this DNA is the aggressiveness of "trading dilution for expansion," rather than the restraint of putting per-share shareholder returns first; it has proven that IREN can survive and can switch tracks, yet it has not yet proven it will protect the value of every share for long-term shareholders.

  Jun 10, 2026
• QDoes management (especially the founders) have a long-term vision and interests deeply aligned with the company? Are they willing to sacrifice current profit for five to ten years out?6/10

  A

  Both long-term vision and interest alignment are a "yes," but the form of alignment leans toward "control" rather than "sharing per-share value with external shareholders": the two co-founders firmly hold control through the dual-class share structure and are clearly willing to sacrifice current profit for a platform position five to ten years out, but they look more like entrepreneurs willing to burn cash lavishly on expansion than capital allocators who protect intrinsic value per share for minority shareholders. On this question I give "neutral-to-positive."

  First, interest alignment. The report discloses that the two co-founders and co-CEOs (Daniel Roberts and Will Roberts) together controlled 35.4% of the voting power as of August 15, 2025 through Class B shares and their holdings; each Class B share carries 15 votes, and the control structure lasts until at most November 17, 2033. This means they are by no means "without skin in the game": holding control is itself a form of deep alignment, and the dual-class structure plus the arrangement lasting until 2033 objectively provides institutional room for "making long-cycle decisions without being hijacked by short-term market sentiment," which is the long-termist soil Baillie Gifford values.

  Next, "whether they are willing to sacrifice current profit for five to ten years out": the answer is a very clear "yes, even overly willing." The evidence is that the company sacrifices current profit for future capacity without hesitation: actively retiring the mining cash cow, directly booking $140.4 million of impairment and recording a $247.8 million net loss in Q1 2026, with the FY26 nine-month operating loss widening to $426.3 million on the report's basis. At the same time, to lock in Microsoft/Nvidia capacity it pushed capital commitments up to the roughly $11.9 billion the report discloses. This is a management team placing heavy bets on the long term and willing to sacrifice short-term book earnings; long-term vision is not lacking.

  But there are two demerits that must be honestly pointed out, precisely what Baillie Gifford and external minority shareholders must watch for:

  First, alignment ≠ sharing the dilution of per-share value. The founders are aligned through control, but much of the expansion money comes from issuing shares to external shareholders. The report discloses that common shares ballooned from 64.75 million at the end of June 2023 to 341 million at the end of March 2026, with a further 15.88 million shares issued after the quarter, and another $3.0 billion of convertible notes issued in May 2026 (netting about $2.96 billion). The founders keep their voting power undiluted through Class B shares, while external shareholders' economic interest is repeatedly diluted; the cost of the long-term vision falls squarely on minority shareholders.

  Second, concentrated control is a double-edged sword. 35.4% of the voting power + 15x super-voting rights mean external shareholders have almost no ability to provide checks and balances. The report fairly notes that this "may bring both a long-term orientation and a control discount." Once the founders' expansion judgment goes wrong, minority shareholders lack means to correct course.

  Conclusion: the founders have a strong long-term vision, are deeply aligned through control, and are clearly willing to sacrifice current profit for the long term. These three points all fit Baillie Gifford's preference for quality management. But the form of alignment is "control alignment" rather than "sharing dilution with minority shareholders," and under super-voting rights external checks are absent. So they are "entrepreneurs who can be trusted to place long-term bets" rather than "capital-allocation masters who will cherish every share for you"; the report's characterization of "neutral-to-positive / worth watching but not worth blind faith" is apt.

  Jun 10, 2026
• QIf it disappeared tomorrow, how much would customers miss it? Is its growth model sustainable and not reliant on harming society or regulation?4/10

  A

  If IREN disappeared tomorrow, a few large customers would miss it in the short term on the matter of "the capacity-delivery window," but not to the point of "unforgettable and irreplaceable": its compute capacity can be replaced by other suppliers. On the positive side: its growth model is highly sustainable and does not rely on harming society or crossing regulatory red lines; instead it steps on the real social need of AI power shortage. So on this question, "indispensability" is weak, but "social/regulatory sustainability" is strong. What Baillie Gifford wants on this question is meeting the dual bar of "indispensable + sustainable," and IREN is strong on one and weak on the other.

  First, "how much would it be missed" (indispensability): weak. The measure is whether customers can easily find a substitute. What IREN sells is GPU cloud capacity and power/data centers, inherently substitutable supply. The report clearly points out that customers "do not necessarily have strong stickiness to IREN itself; they care more about on-time delivery, compute pricing, and SLAs." By customer structure, signed ARR of about $3.1 billion is highly concentrated in Microsoft (about $1.9 billion) and Nvidia (about $0.7 billion). This concentration shows that within the window of "AI capacity being extremely scarce right now, and whoever can power up on time is needed," IREN is needed; but this "missing" comes from the capacity value at a scarce moment rather than from irreplaceable technology or platform. Once industry capacity loosens, or Microsoft/Nvidia self-build and find alternative supply, the "missing" will decay quickly.

  Still, a fair plus is warranted: Nvidia not only signed a $3.4 billion contract but also took the right to subscribe for up to 30 million shares at $70/share (about $2.1 billion) within five years, and reached a 5GW strategic partnership. Nvidia's willingness to bind deeply through equity shows that within its ecosystem IREN is currently "an important enough capacity partner worth locking in." This is stronger than the stickiness of pure spot-leasing customers, but it is still "important" rather than "indispensable."

  Next, "whether growth is sustainable and does not harm society or regulation": strong, and this is IREN's clean side.

  • Social dimension positive: what IREN satisfies is the real power and infrastructure demand behind AI compute. The report cites the IEA's assessment that global data-center electricity consumption doubles to about 945 TWh by 2030, consistent with the IEA "Energy and AI" report. It grows by filling a real supply gap rather than by regulatory arbitrage or harming users.
  • Regulatory/compliance dimension basically clean: the legal matters the report discloses are a roughly $100 million U.S. import-tariff dispute and a roughly $27.7 million Canadian GST dispute, which are routine tax/tariff disputes of cross-border operation, not fundamental regulatory risks such as fraud, monopoly, or data abuse that harm society. The historical financial restatements and internal-control weaknesses have also been disclosed as remediated.
  • One honest caveat: the bitcoin-mining leg has long faced social controversy over energy consumption and the environment, and highly power-hungry AI data centers will also continue to face regulatory scrutiny over grid and community electricity use; but IREN is actively retreating from mining and expanding on a base of secured grid-connected power, all advancing within a compliance framework.

  Conclusion: the degree to which customers would "miss" it is moderate-to-weak: needed by a few large customers within the capacity-shortage window and locked in by Nvidia through equity, but capacity is substitutable and it lacks indispensability. Meanwhile, the growth model is highly sustainable and does not rely on harming society or crossing regulatory red lines, clearly meeting the bar on this dimension. By Baillie Gifford's dual standard of "indispensable + sustainable," IREN passes on sustainability but fails on indispensability.

  Jun 10, 2026
• QWhat are this business's unit economics (gross margin, incremental returns)? Do they improve or worsen as it scales? Where does the money it earns go?3/10

  A

  At the reported level, gross margin is very pretty (about 67–68%), and the adjusted EBITDA margin of about 41% is not bad either; but this is an accounting-basis "good," while the owner cash return that truly determines returns (owner earnings, incremental return on invested capital) is currently very likely still near breakeven or even negative. As it scales, unit economics "should theoretically improve" (AI Cloud margins are higher than mining), but this has not yet been validated today. The money earned is almost entirely, and far from enough, plowed back into capital expenditure. This is the core battleground of IREN's valuation debate, and the link Baillie Gifford should most interrogate.

  First, the accounting-level unit economics: superficially excellent:

  • The report discloses gross margin rising from 47.8% in FY23 to 68.3% in FY25 and 67.3% in FY26's first nine months, because the share of high-margin AI Cloud rose while the mining share fell.
  • On the company's official basis, in Q1 2026 adjusted EBITDA was $59.5 million, with the margin holding at 41%.
  • The direction is right: AI Cloud's unit economics (contract-based recurring revenue + high gross margin) are indeed superior to the commoditized mining of selling off bitcoin mined each day.

  But we must see through to the owner-return level, where it is no longer optimistic:

  First, EBITDA is not owner earnings; this business's depreciation is real capital consumption. The report discloses that FY26 nine-month depreciation and amortization reached $305.6 million, and GPU/mining-rig technology iterates extremely fast; the 41% EBITDA margin shrinks substantially after deducting real equipment depreciation and maintenance capex. In Q1 2026 the company still recorded a $247.8 million net loss (including $140.4 million of impairment) on $144.8 million of revenue, with the FY26 nine-month operating loss at $426.3 million on the report's basis.

  Second, operating cash flow looks good, but free cash flow is very poor. The report discloses FY25 operating cash flow of $245.9 million and FY26 nine-month of $289.3 million; but investing cash flow over the same periods had net outflows of $1.380 billion and $2.609 billion. Translated into owner's language: the company "needs more external capital the more it grows" rather than "generating more cash the more it grows." On this basis the report's conservative owner-earnings range is about -$50 million to +$40 million; that is, the current distributable-cash-flow capacity is near breakeven or even negative.

  Third, incremental return on invested capital (incremental ROIC) is not yet proven. Capital commitments were pushed up to the roughly $11.9 billion the report discloses, but whether these investments can produce a mid-to-high double-digit owner-earnings margin over the long run is currently only an assumption. The report is clear: if the mature owner-earnings margin is below 15%, the current share price is expensive; only above 20% does it barely have room for discussion.

  Better or worse as it scales? Directionally it should improve: depreciation/power dilution from the capacity ramp, a rising AI Cloud share, and the scale effect of contract-based revenue should theoretically improve unit economics. But there is a reverse risk: the report warns that "if AI compute-leasing prices fall faster than equipment costs," IREN could be caught in a squeeze of "heavy assets on the supply side, weak pricing on the demand side," in which case the larger the scale, the faster the losses amplify.

  Where does the money go? The answer is very clear: almost all of it is plowed back into expansion, and internal cash is far from enough, so it relies on external financing to fill the gap. The report discloses the company has no dividend and no meaningful buyback; common shares ballooned from 64.75 million (2023.6) to 341 million (2026.3), and in May 2026 it further issued $3.0 billion of convertible notes (netting about $2.96 billion). This is the path of "first scale up the platform, then hope unit economics deliver."

  Conclusion: accounting gross margin/EBITDA margin are eye-catching, but the true owner return (owner earnings, incremental ROIC) is currently near breakeven or negative, a classic case of "good on paper, not enough cash." After scaling, unit economics have a direction of improvement, but this is unvalidated and carries the risk of being bitten back by declining compute prices; the money earned, and far more than what is earned, is all reinvested in expansion. By Baillie Gifford's standard of "high and sustainable incremental returns," IREN currently falls short. This is its most concrete weakness.

  Jun 10, 2026
• QWhat conditions must simultaneously hold for it to 5x in ten years? Are these conditions realistic? What expectations does today's share price imply?2/10

  A

  For IREN to 5x in ten years, a long string of high-difficulty conditions must hold "simultaneously"; each condition alone is not far-fetched, but for all of them to be realized without continued large-scale dilution is very difficult in practice. And today's share price has already priced in "fairly successful execution"; what it implies is an optimistic price of "AI will very likely deliver at high quality," rather than a discounted price of "will AI deliver."

  First, calibrate today's share price (be sure to use current values rather than the report's snapshot). At the time of writing (May 20, 2026) the share price was about $47.74 and the market cap about $15.9 billion. But the price has since risen markedly and swung sharply: as of the June 9, 2026 close of about $54.02 (down 8.73% that day, having closed at $59.19 the prior trading day), the market cap was about $19.3 billion, shares outstanding about 357.4 million, and reported PE about 99x. In other words, the current price is a notch higher still than the report's "clearly overvalued range" (the report sets above $50 as clearly leaning toward optimistic pricing); the report's conclusion of "insufficient margin of safety" only holds more firmly at the higher current price.

  What conditions must hold simultaneously for a 5x in ten years? Starting from about $54, a 5x means the market cap must go from about $19.3 billion to about $96.5 billion within ten years. To support this scale, the following conditions are almost all indispensable:

  1. Microsoft, Nvidia, and follow-on contracts powered up, installed, and billed on schedule, truly turning the signed about $3.1 billion, targeted $3.7 billion of ARR into revenue, and continuing to expand to an even larger scale on top of that;
  2. The mature owner-earnings margin stably reaching the mid-to-high double digits (the report requires ≥15%, ideally 20%+), rather than staying at today's near-breakeven level;
  3. No major delivery delays, cost overruns, or another round of large impairment during the capacity ramp from current levels toward 5GW (the company has already flagged roughly $520 million of possible future impairment);
  4. AI compute-leasing prices not being driven down by supply expansion, avoiding the squeeze of "heavy assets on the supply side, weak pricing on the demand side";
  5. The critical one: no longer relying on large-scale dilutive financing. This is the hardest: the report discloses common shares have ballooned from 64.75 million (2023.6) to 341 million (2026.3), with further issuance after the quarter, and another $3.0 billion of convertible notes issued in May 2026 (netting about $2.96 billion), plus Nvidia's right to subscribe for up to 30 million shares at $70/share. If the market cap is to 5x while the share base keeps ballooning at this pace, the difficulty of a 5x per share is raised further.

  Are these conditions realistic? Each is possible individually, but "all holding simultaneously without relying on continuous dilution" is a low-probability stack. The broad direction of AI power demand (the IEA estimates data-center electricity doubles to about 945 TWh by 2030, see the IEA report) provides the soil for upside, but for IREN to exclusively capture this increment amid encircling giants and convert it into high-quality per-share cash returns, the chain of evidence is currently still incomplete.

  What expectations does today's price imply? Reported PE about 99x, the report's P/S about 21x, P/B about 6x; this is the valuation language of a "high-certainty AI infrastructure platform" rather than a miner's valuation. In other words, the market has already assumed most of the optimistic conditions above will be realized. The report's three-tier DCF puts the optimistic-scenario intrinsic value ceiling at about $36–50 per share, while the current price of about $54 already stands above the top of the optimistic scenario. This means: even if IREN executes fairly well, what the current price can offer is only mediocre returns; the real 5x upside would either require execution far beyond the optimistic assumptions, or first require a round of valuation digestion and re-buying at a lower price (the report considers the more reasonable long-term entry point to be $18–25).

  Conclusion: the checklist for a 5x in ten years is long, and the hardest condition, "not relying on dilution," runs counter to the company's past playbook, so it is overall a low-probability stack; and today's price of about $54 already implies the optimistic expectation that "AI will very likely deliver at high quality," with an insufficient margin of safety. By Baillie Gifford's standard of "emphasizing upside imagination but requiring the price to leave room for error," IREN's upside story holds while the price does not; the current price is a prepayment for success rather than a discount on growth.

  Jun 10, 2026
• QWhy hasn't the market realized all this yet? Is it that they can't understand it, look down on it, or can't see far enough? What will become the "narrative inflection point"?3/10

  A

  This question must be answered in reverse for IREN: the market is not "yet to realize" its potential; on the contrary, the market has already fully, even excessively, realized the imaginative space of the AI transition, pricing in "very likely success" via about 99x PE and about $19.3 billion market cap. So IREN is a hot stock that is "very well understood, seen far, even seen too optimistically," rather than a dusty stock that is "not understood / looked down on / not seen far enough." The real "narrative inflection point" lies in whether execution can support the high expectations already priced in, rather than in when the market discovers its merits. This is the opposite of the situation for most Baillie Gifford targets, and must be pointed out honestly.

  Why say the market has already realized it rather than failed to? Three pieces of evidence:

  First, the valuation is already the language of an AI platform rather than the language of a miner. Currently reported PE is about 99x and market cap about $19.3 billion (June 9, 2026 close of about $54.02, down 8.73% that day), with the report's P/S about 21x and P/B about 6x. The market is clearly not pricing it as a bitcoin miner.

  Second, its cross-sectional market cap already far exceeds all comparable transitioning miners. The report discloses that IREN's market cap (about $15.9 billion at the time of writing) already significantly exceeds WULF's about $9.03 billion, RIOT's about $7.87 billion, CIFR's about $7.62 billion, CORZ's about $7.40 billion, MARA's about $4.73 billion, and CLSK's about $3.35 billion; the market gave IREN a markedly higher "AI-delivery-probability" premium with real money. This is a display of "understanding and being bullish" rather than "looking down on it."

  Third, the top-tier endorsements are already public and digested by the market. The roughly $9.7 billion Microsoft contract, the roughly $3.4 billion Nvidia contract + 5GW strategic partnership + the right to subscribe for up to 30 million shares at $70/share have all been announced, and the share price has already reacted to this good news (spiking above the report's snapshot). The market does not lack information; what it lacks is proof that these contracts convert from "signed" to "high-quality owner earnings."

  So where might the market "get it wrong / fail to see far"? Honestly, there are two cognitive blind spots pointing in opposite directions, rather than a one-directional underestimation:

  • What might be underestimated: if IREN truly ramps capacity to 5GW, raises its owner-earnings margin to 20%+, and does not rely on heavy dilution, then relative to its long-term cash-generating capacity, today's price might actually not be expensive; but this is a tail scenario that only holds if "execution far exceeds the optimistic assumptions."
  • What might be overestimated (needs more caution): the market may underestimate the erosion of per-share value from continuous dilution, underestimate the risk of declining AI compute prices and further impairment (the company has flagged about $520 million), and may mistake "contract value" for "owner earnings already in hand." The report's optimistic-scenario intrinsic value ceiling of about $36–50 per share is already below the current price, meaning the blind spot in current pricing leans more toward the "overly optimistic" side.

  What will become the "narrative inflection point"? Because the story is already fully priced, the inflection can go either up or down:

  • Upward inflection: the Microsoft/Nvidia projects power up on schedule and begin large-scale billing, AI Cloud revenue scales from $33.6 million in a single quarter (already nearly doubling from $17.3 million the prior quarter, see the Q3 FY26 results) to an annualized several billion, free cash flow and free cash flow per share turn significantly positive for the first time, and there is no more large dilution; this would upgrade the "contract narrative" to a "cash-flow narrative" and support a higher valuation.
  • Downward inflection: large contracts delayed/renegotiated, another round of large issuance or convertible notes, AI projects' actual gross margins persistently below expectations, new material impairment or accounting problems; any one of these would puncture the high expectations already priced in and trigger a sharp drawdown (the report judges the risk of permanent capital loss can reach 50% or more, because the current price is far from static asset value; the current P/B of about 6x also confirms that book net assets cannot hold up the share price). The -8.73% swing that day itself shows this is a stock extremely sensitive to the narrative.

  Conclusion: IREN's situation is the opposite of the typical Baillie Gifford "great growth stock the market has yet to discover"; the market has already understood it, seen far, and granted an optimistic premium, with the perception gap leaning more toward "whether it is overly optimistic" than "whether it is underestimated." The key to the narrative inflection is on the execution side: the delivery progress from "signed ARR" to "distributable owner earnings" reinforces the story if up and punctures it if down. For long-term investors, this means chasing in at the current price is "betting on delivery" rather than "picking up a bargain."

  Jun 10, 2026