IREN's Perplexity Deal Puts Real Workloads Behind the GPUs — November 5 Will Say What They Earn

IREN shares fell 4.5% on Friday, caught in a wider retreat across data-centre and crypto-mining names as investors fretted over the cost of financing AI build-outs and the returns those projects might eventually generate. Reports linked the pressure to that sector mood alongside a pullback in Bitcoin.

The same day, IREN announced something rather more tangible: a multi-year agreement giving Perplexity access to NVIDIA Blackwell Ultra GPUs at one of its data-centre campuses. The capacity is earmarked for the models and agents behind Perplexity Computer, a system IREN describes as coordinating agents and distributing tasks across models from multiple providers.

That distinction matters. A named customer with a specific application offers a firmer read on demand for IREN's AI cloud than broad optimism about the artificial-intelligence market. The multi-year term on its own, though, says nothing yet about what the arrangement will contribute to earnings.

A Relationship That Predates the Headline

Friday's announcement formalises a connection investors had already been shown. In results published on 27 August, IREN named Perplexity as its most recent AI cloud customer. The new disclosure fills in the technical underpinnings and the multi-year horizon.

The segment it touches is no longer a sideline. Quarterly revenue from AI cloud reached $70.5 million, overtaking Bitcoin mining at $66.7 million for the first time. Against that backdrop, the contract is an operational step forward — not a substitute for demonstrating profitable growth.

Two Signals, One Calendar Date

IREN has committed to reporting results for the quarter ended 30 September 2026 on 5 November, with a conference call to follow. Consensus looks for revenue of $221.71 million and a loss of $0.76 per share. Those figures make plain that demand for compute and profitability have to be judged separately.

B. Riley trimmed its price target to $64 from $91 on Friday while keeping its buy rating. The direction of the revision is unmistakably negative; the underlying recommendation is not. A positive rating does not neutralise a $27 cut, and the cut does not amount to abandoning the investment case.

What investors are left with is a pair of contrasting signals: Perplexity strengthens the operational demand base, while the lower analyst target reflects more cautious valuation expectations. The upcoming report is the next hard checkpoint at which the AI cloud business can be measured on revenue and profit.

Financing Is Settled; Returns Are Not

Behind the expansion sits a completed equipment financing arrangement with Wingspire Equipment Finance worth $363 million. It refinances AI computing infrastructure that is already generating revenue. The facility is secured mainly by the financed equipment and supported by multi-year customer contracts — a concrete transaction rather than a statement of intent, and a useful counterweight to the general scepticism surrounding AI funding.

The caveat is that the financing answers a funding question, not the question of whether the whole strategy pays off. It provides room to manoeuvre; it does not certify success.

The scale of the ambition is visible in Badajoz, announced on 21 October, where IREN plans a 300 MW AI data-centre campus with initial operations slated for 2028. The creation of IREN Europe has given the continental push an organisational home as well. For shareholders that is an opportunity and an added execution burden in equal measure.

Between an announced site and a profitably operating one lie a series of valuation questions. The build-out should be treated neither as a finished contribution to value nor written off simply because the sector is out of favour. The 5 November report — and what management says about commercial progress and delivery — will carry more weight than any attempt to apportion Friday's decline between company-specific and sector factors. A convincing update would need to do more than invoke long-term demand for AI compute.