IREN's H100s Just Re-Let at Double the Original Price — Now Comes the Hard Part

IREN shares jumped 21% on Friday, and the headlines credited better-than-expected quarterly results and fresh customer contracts. But the most telling disclosure of the week concerned hardware the company had already bought and paid for. During the earnings call, Co-CEO Daniel Roberts said IREN had re-let Nvidia H100 graphics processors at twice their original contract price. Chief Commercial Officer Kent Draper added that the higher rate was locked in for several years.

That single detail cuts to the heart of the investment case. IREN's AI business is gaining economic substance — provided older equipment can keep earning attractive rates well past its first lease. Judging the company means watching the second rental of a chip as closely as the first order. That is where the durability of the returns on installed hardware actually gets decided.

IREN disclosed neither a dollar figure nor the number of GPUs involved, nor where they sit. The claim therefore speaks to the usability of aging hardware only to the extent those terms can be replicated across a meaningful slice of the fleet. A single favorable re-lease carries less weight than a repeatable pattern, so the yardstick should be the breadth of such follow-on contracts. The more legacy systems that find long-term takers, the more convincing the economics underpinning the buildout.

Depreciation Makes Re-Leasing a Double-Edged Test

The financial statements give the re-let a second meaning. IREN assumes a five-year useful life for its high-performance computing hardware. In the first quarter of fiscal 2027, depreciation and amortization ran to $148 million, against $85 million a year earlier.

Re-leasing can help absorb that rising depreciation charge — if the revenue left after operating costs contributes enough. Doubling the contract price is a positive signal. The missing detail on scale, however, caps how much it says about the company as a whole.

The balance sheet shows what is at stake. GPU hardware at acquisition cost climbed from $2.035 billion on June 30 to $5.819 billion on September 30. The installed base is expanding far faster than individual re-leasing wins can explain. That is the structural challenge of AI infrastructure: capital goes into equipment early, and its economic return has to prove itself over years. IREN becomes more credible when long-term leases keep that growing hardware base consistently utilized. Rapid growth also raises the pressure to secure such contracts at scale.

By quarter-end, graphics processors and related infrastructure carrying a net book value of $4.769 billion stood pledged as collateral. The same assets thus sit at the center of both operations and the securing of financing. Good rental terms are worth more when the equipment also backs the borrowings — a link that favors IREN as long as the hardware stays economically useful and the contracts deliver dependable cash flows.

Microsoft Redraws the Blueprint

Execution has become the binding constraint. Reuters reported on Thursday that IREN's infrastructure deliveries for Microsoft are running late after the customer requested changes to the network design. IREN said the delay will not trigger financial penalties under the contract — a relief for shareholders, since at least that immediate cost falls away, according to the company.

The design change also shows how tightly the buildout is meshed with the customer's requirements. Winning an order convincingly requires delivering on it just as convincingly. The Microsoft relationship still counts in IREN's favor, provided the revised specifications can be translated into a workable delivery schedule.

The same day brought another timetable shift: the operating target of more than US$4 billion in annualized recurring revenue has moved. Instead of year-end, it is now expected in January 2027 — possibly not until February 2027. That date belongs at the center of any valuation. It ties the growth ambition to a concrete operational test, and the upside gains weight only if IREN holds the revised schedule and actually brings the promised capacity online.

The Numbers Behind the Rally

Thursday's results provide a strong commercial starting point. Revenue in the first quarter of fiscal 2027 reached US$228 million, of which US$200 million came from the AI cloud business. That mix matters: AI cloud already defines the quarterly top line, which makes its operating progress a more sensible gauge than the strength of any single trading session.

AI cloud revenue rose 182% from the prior quarter, and adjusted EBITDA came in at US$66 million. Alongside contract extensions and new agreements announced after quarter-end, that supports confidence — contingent on IREN meeting the added obligations reliably.

Customer oversight adds another layer. According to CNBC, Microsoft is reviewing its neocloud partners and could cut their number to prioritize its own data centers. Reuters and Bloomberg reported closer scrutiny of projects including those at IREN. For IREN, that raises the value of resilient customer relationships, where technical capability translates into long-term utilization. A growing hardware fleet needs both: competitive systems and customers who book them across their useful economic life.

The verdict is constructive but calendar-bound. Business momentum favors IREN if it turns into infrastructure delivered on schedule and cloud capacity customers can actually use. The repeatability of the H100 win is the thing to watch: multi-year follow-on leases at attractive terms would strengthen the economics of the AI buildout, if they cover a significant share of the fleet. That is what determines whether older hardware becomes a dependable earnings engine — and whether the next stage of expansion looks more convincing than the last.