IREN이 $2.8B 규모의 AI 클라우드 계약을 체결하고 2026년 런레이트 목표를 $4B 이상으로 상향 조정, 주요 글로벌 기업들로부터의 가속화되는 약정을 시사합니다.

IREN Limited (NASDAQ: IREN) disclosed on July 20 that it has signed $2.8 billion in new multi-year AI cloud contracts with leading AI developers. The agreements raise the company’s year-end 2026 annualized run-rate revenue (ARR) target from $3.7 billion to over $4 billion, with approximately 85% of the revised target now under contract [1].

The neocloud operator’s customer roster spans both bare metal and managed cloud services and now includes Microsoft, NVIDIA, Perplexity, Figure AI, Together AI, Fluidstack, Fireworks AI, Fal AI, Hume AI, and one additional unnamed leading AI developer [1]. These contracts carry a weighted average duration of roughly four years, with customers prepaying approximately 45% of the associated GPU capital expenditure [1]. This prepayment structure significantly reduces IREN’s balance sheet risk on hardware procurement, while the multi-year terms provide revenue visibility through 2030, albeit amid ongoing execution challenges in scaling deployments rapidly enough to meet commitments [1].

As of June 30, 2026, IREN held $7.6 billion in cash and cash equivalents, alongside an additional $1.7 billion in restricted cash earmarked for GPU financing tied to its Microsoft contract [3]. To support its expansion, the company raised $2.6 billion through a convertible notes offering in May 2026 at 1.00% interest, maturing in 2033 [3].

IREN’s infrastructure buildout has been among the most aggressive in the neocloud sector. Over the past 12 months, the company expanded from approximately 3MW of self-built AI cloud capacity to 480MW scheduled for delivery in 2026, with 1.2GW targeted for 2027 [1]. It has secured a total power portfolio of 2.91GW of grid-connected capacity [3]. Parallel to this physical expansion, IREN has entered purchase agreements to scale its GPU fleet to 150,000 units, comprising NVIDIA H100s, H200s, B200s, B300s, and GB300s, as well as AMD MI350Xs [5]. Deployments will be distributed across existing air-cooled data centers in Mackenzie, British Columbia, and Childress, Texas [5].

The breadth of IREN’s customer base marks a strategic shift away from concentration risk—a vulnerability that has historically plagued peers in the sector. By adding AI-native companies such as Perplexity and Figure AI alongside hyperscaler anchors Microsoft and NVIDIA, the company has diversified its revenue streams. IREN’s existing anchor contracts include a $9.7 billion five-year agreement with Microsoft initiated in late 2025 and a $3.4 billion five-year agreement with NVIDIA [3]. The newly announced $2.8 billion in contracts layers additional recurring revenue on top of these foundational relationships. Following the announcement, Goldman Sachs noted that the disclosure “shows customer diversification” and “reinforces tight supply and favorable pricing conditions” [4].

The market reaction was immediate, with IREN shares surging roughly 20% on the news and trading above $40. The rally underscores the commercial pivot from the company’s Bitcoin mining origins, particularly given the stock’s 52-week low of $14.72 [2] [3]. Goldman Sachs maintained a Neutral rating on IREN with a $50 price target, acknowledging the strategic value of the diversified customer mix while flagging execution risks inherent in the rapid scale-up [4]. Broader analyst sentiment remains bullish, with a consensus Buy rating supported by 10 buy ratings, 3 holds, and 1 sell, and a median price target of $84.50 [3].

The announcement arrives amid a broader surge in neocloud commercial activity. Hut 8 recently disclosed a $9.8 billion 15-year lease in Texas, while TeraWulf announced a $19 billion 20-year lease with Anthropic, signaling that AI developers are increasingly willing to commit substantial long-term capital to secure GPU compute capacity outside the traditional hyperscaler ecosystem [2].

Despite the strong commercial traction, IREN faces notable operational headwinds as it scales toward its 1.2GW target, including grid interconnection delays, transformer shortages, and regional energy constraints [3]. While the company currently reports gross margins of approximately 90%, sustaining those levels while deploying infrastructure across multiple geographies and diverse GPU architectures will test management’s execution capabilities [3].