Summary IREN Ltd. is downgraded to 'Sell' after a sharp Q2 2026 double miss on EPS and revenue. IREN's AI cloud revenue grew 137% QoQ but remains under 10% of total, while Bitcoin mining revenue slumped 28% QoQ. Execution risk is high as IREN must scale AI revenues ~68x to meet ambitious ARR targets, with zero margin for error. Valuation appears excessive at ~4.78x FY2027 sales; fair value implies 38-39% downside if consensus is cut. IREN is Falling Sharply on Q2 2026 Earnings I haven't covered IREN Ltd. ( IREN ) stock as a standalone ticker - only in terms of comparison to Cipher Mining ( CIFR ). In late November 2025, I picked CIFR over IREN because I liked CIFR's landlord-style model with its long-term contracts from Google ( GOOG ) and Amazon ( AMZN ). Right now, IREN is falling by over 3% following its double miss (on GAAP-based EPS and revenues), and the market seemingly didn't like the Bitcoin-driven revenue slump amid the insufficient speed of gaining AI cloud services revenues. Seeking Alpha, IREN At some point, the post-earnings drop was over 18%, based on TrendSpider's data. While it might look like an interesting opportunity to load up on IREN at a lower price, I'm calling for investors' cautiousness, as I think that IREN is lagging behind on its AI pivot compared to the aggressive growth consensus expectations currently baked into its valuation. I'm waiting for some sharp cuts to the medium-term revenue and EPS projections, and if I'm right about this assumption, IREN can eventually start to test its previously strong price levels (in the low 20s): TrendSpider Software, IREN, notes added Reviewing IREN's Q2 Results IREN's Bitcoin mining revenues amounted to $167.4 million for the quarter, and it was a 28% QoQ slump because of the general weakness in the crypto market. At the same time, the firm's AI cloud services revenue grew to $17.3 million (up from $7.3 million, i.e., by 137% QoQ), which looks good on the surface, but the AI sales accounted for less than 10% of consolidated revenues. As a result, the market turned out to be wrong regarding IREN's top line by over 18%, which is the heaviest miss since Q2 2024, according to Seeking Alpha: Seeking Alpha, IREN's revenue beats/misses Another thing that spooked investors was the bottom-line miss (on a GAAP basis). IREN recorded ~$219.4 million in non-cash and non-recurring items: Unrealized losses on prepaid forwards and capped calls associated with convertible notes, one-time debt conversion, and a $31.8 million impairment on Bitcoin mining hardware. The income tax benefit of ~$182.5 million couldn't save the day, and even on an adjusted EBITDA basis, IREN saw an 18% drop on a QoQ basis. IREN's IR materials, Q2 2026 The narrative around the "bullish HPC infrastructure play" that many investors and analysts have been pricing in is now changing because the market suddenly realized that the whole transition from a pure Bitcoin miner into an AI play is easier said than done. After the Q2 misses, all 9 analysts who cover the stock have decided to downgrade their estimates for the upcoming Q3 release. Seeking Alpha We know that as of Q2, IREN had ~$2.3 billion of ARR under contract, including about $400 million at its Prince George site, and the management reiterated their target of getting to 140,000 GPUs by the end of 2026 (that should let IREN achieve an ARR of ~$3.4 billion). But logically, to achieve this scale, IREN's AI segment should grow by ~68x from its current levels in just a few quarters, and this aggressiveness in ramping leaves zero margin for error, in my opinion. The risk of failed execution is probably one of the primary reasons stocks like IREN fall when the market loses conviction. I believe that as IREN shifts its attention and investments from Bitcoin mining to AI services, the firm will not see material contributions to the bottom line for the next couple of years because it will have to service the debt it has accumulated to finance the ramp-up. Another risk comes from the fact that IREN is obligated to build new capacity right in the middle of the memory chips' supercycle. More expensive memory chips make building data centers even more expensive, so IREN may not be able to cope with its current budget. Additionally, there's a risk that their AI services will face a pricing war as more neoclouds expand their capacity. If my concerns are valid, we'll likely see a cut to the current EPS estimates, which still project about 149% YoY growth in FY2028. Seeking Alpha, IREN's EPS revisions In addition to all the above, I still see the "bare-metal server provider" risk that many analysts have already noted before. A comparison with Nebius Group ( NBIS ) makes IREN an outsider on that front - without a software layer similar to what NBIS has in place, IREN's services will look low-margin and commoditized. As a result, the stock will not have a justification for its premium valuation. Now, talking about IREN's valuation, I think that without a high-quality software layer, the stock's FY2027 price-to-sales ratio of ~4.78x looks excessive. Capital-intensive business models usually trade closer to their book values. In terms of sales, they usually don't exceed 2–3x, so IREN is quite expensive if we adjust its sales projections based on the lags from the Q2 data. Seeking Alpha Assuming a possible FY2027 revenue consensus cut of 10% and putting a P/S multiple of 3x on it, I get IREN's fair equity value of ~$7.371 billion. The current market cap is ~$12 billion, so there's a possibility of another 38-39% even after the post-Q2 drop. Conclusion IREN's expansion plans look ambitious, but the fresh Q2 2026 data has shown us that there's no need to hurry and buy IREN at the current price level. We should always keep in mind the existing execution risk and the high odds of seeing the forward EPS and sales consensus being cut significantly from today's figures. I have decided to downgrade IREN to "Sell" from my previous "Hold" rating because I didn't like the Q2 print. Good luck with your investments!