IREN Drops 9.3% as Meta's Compute Selloff Tests the Neocloud Investment Case

Meta's announcement to sell excess GPU compute to outside customers triggered a sector-wide selloff that compounds a brutal week for IREN, whose $22 billion in hyperscaler commitments now faces scrutiny from a powerful new competitive entrant.

IREN closed at $39.81 on July 8, down 9.3% on the session, as the market digested a July 7 announcement by Meta that it would begin selling surplus compute capacity to third-party customers. The move rattled the entire specialist GPU-rental sector: Nebius reportedly shed roughly $12 billion in market capitalization in a single session, while CoreWeave and IREN fell in sympathy. The selloff compounded an already difficult stretch — a combination of the Meta competition narrative and an executive compensation controversy had already pushed IREN down sharply in the preceding days. The market's logic is direct and punishing: if one of the world's largest AI spenders is transitioning from a pure customer into a competitor, the addressable market for independent neoclouds contracts, and the premium multiples assigned to contracted backlogs come under pressure.

That logic collides with a narrative IREN has spent two years assembling. The Sydney-based company pivoted away from Bitcoin mining and began signing major capacity agreements; a Microsoft deal disclosed via SEC 8-K in November 2025 was reported at $9.7 billion, and 24/7 Wall St. reported last week that IREN has accumulated approximately $22 billion in committed hyperscale capacity, positioning itself — in analysts' framing — as a strategic alternative to CoreWeave and Lambda. Against that backdrop, B. Riley raised its price target to $96 on July 5, more than double the current price, citing the planned 800-megawatt data center at Bundey in South Australia. Jefferies had earlier argued the stock could be as much as 37% undervalued. IREN's inclusion in the Russell 1000 index in late June formalized its transition from speculative miner to institutional-grade infrastructure provider. The gap between those analyst targets and the current market price reflects how sharply sentiment has shifted in a matter of days.

The capital commitment underlying that ambition is substantial and accelerating at a pace that has no real precedent among neocloud peers. SEC filings show IREN spent $573 million on property, plant, and equipment in fiscal year 2025 ending June 2025, against $501 million in revenue — a capital-intensity ratio exceeding 100%. Spending then surged: the nine-month cumulative figure in the 10-Q for the period ended March 31, 2026 totaled $1.67 billion, nearly three times the prior full-year total. The company has also broadened its platform through acquisitions: it bought OpenStack specialist Mirantis in May 2026 to add cloud orchestration capability, completed the purchase of Nostrum Group in June 2026 to extend into Europe, and deployed Nvidia DSX Air interconnect fabric alongside partner BE Networks for large-scale AI factory builds. A June 2026 disclosure cited Microsoft and Nvidia partnerships targeting $4.4 billion in annual recurring revenue, and analysts at the time estimated the total contract backlog at $13 billion across AI data center and HPC commitments.

The Meta shock arrived on top of a pre-existing company-specific concern. Earlier this week IREN disclosed an executive compensation package that reports variously described as $700 million to $800 million in equity awards — a figure that attracted pointed scrutiny given the company's current revenue scale. Multiple 8-K filings submitted to the SEC on July 7 suggest the governance restructuring is substantive; the precise terms require a reading of the underlying documents, but the volume of regulatory filings in a single day signals that the changes are material. The combination of sector-level competitive threat and company-level governance questions has widened the gap between IREN's contracted backlog narrative and investor willingness to assign it a premium multiple.

The structural investment case has not been invalidated, but it has become harder to bracket. GPU compute demand from frontier AI labs and enterprises continues to expand, and IREN's vertically integrated model — combining owned power, connectivity, and software orchestration through Mirantis and the broader stack — offers cost and coordination advantages that pure-play colocation providers lack. The contracted backlog, if delivered, supports a credible path to the $4.4 billion ARR target management has cited. Yet three risks now demand sustained attention. First, Meta's entry into external compute sales is a genuine structural shift; if other hyperscalers follow, independent neoclouds face a pricing and supply overhang that is difficult to model. Second, no neocloud has sustained capital deployment at the pace IREN's SEC filings now reflect — $1.67 billion across nine months against a sub-$600 million annual revenue base — and execution risk at that scale is material. Third, an executive pay package reportedly in the $700 million to $800 million range will face sustained institutional scrutiny at the next proxy cycle. Three signals worth watching: the build-out pace and customer terms at the Bundey campus, whether the July 7 8-K filings reveal any modification to the compensation structure, and whether existing hyperscale customers publicly reaffirm or expand capacity commitments in the weeks following Meta's announcement.