$IREN (-0,04 %)

The question is, is the imagination still there?

Is imagination really a factor? And does profitability no longer matter?

$IREN (-0,04 %)

–Customer prepayments cover 45–55% of GPU CapEx in the most recent contracts. Roberts (CEO) views this as the strongest signal of demand from the earnings call: Customers are not only securing capacity; they are pre-financing the construction.

–Prices: Three-year contracts are about 125% above November, while five-year contracts are about 70% higher. Recent deals amount to $20 million per megawatt of IT load; ongoing negotiations at around $25 million – Payback on the computing investment in about two years.

–FY27 CapEx guidance: $25 billion to $30 billion. This is offset by approximately $14 billion in cash, committed GPU financing, and prepayments; an additional $8 billion is targeted. All data centers are unencumbered and therefore have not yet been used as collateral.

–$4 billion in recurring annual revenue has been contracted for the 2026 capacity, of which $1 billion is currently generating revenue following Microsoft’s $MSFT acquisition of Horizon 1. The $700 million from the NVIDIA cloud contract is not yet included in this figure.

-New to the platform is a multi-year contract with a leading frontier AI lab, whose name IREN has not yet disclosed. The AI developer, which remained unnamed in July, is Prometheus. Together AI and Fireworks AI have renewed and expanded their contracts.

$IREN (-0,04 %) has released its financial results for fiscal year 2026. Revenue was slightly above analysts’ estimates, but the loss is massive—and almost entirely non-cash.

The figures (Q4, quarter ending June 30):

• Revenue: $137.2 million – about 4% above the estimate of ~$132 million, following $144.8 million in the previous quarter
• AI Cloud: $70.5 million, following $33.6 million (+110%) – exceeding the mining business for the first time
• Bitcoin mining: $66.7 million after $111.2 million (−40%)
• Net loss: $684.0 million – of which $450.4 million was write-downs on decommissioned mining hardware
• Adjusted EBITDA: $19.2 million, down from $59.5 million

The loss appears dramatic, but is essentially a accounting consequence of the restructuring: The mining computers are being written off while the facilities are being converted to GPUs. Operationally, cash flowed into the company during the quarter—$1.81 billion from ongoing operations, driven by $1.72 billion in customer prepayments.

According to its own figures, IREN has a total of $14 billion in cash on hand, committed GPU financing, and customer prepayments—with a market capitalization of around $16 billion, this means that nearly the entire company’s value is already fully funded.

In addition, IREN currently has capacity with an ARR of $1 billion already in operation.

More revealing than the net loss is the interest rate differential in the financing:

The $3.6 billion for the contract with Microsoft $MSFT (+0,06 %) costs 6.0%, while the new $2.4 billion funding round from Blue Owl $OWL and PIMCO for clients without an investment-grade rating 9.0%. In my view, a three-percentage-point premium is the true cost of client diversification.

Source: https://iren.com/investor/events-and-presentations

$NVDA (-0,02 %) Statements from the earnings call:

– First annual forecast in the company’s history: approximately 70% revenue growth in fiscal year 2028. Customer forecasts indicate a doubling —the 70% figure represents the maximum the supply chain can handle. The bottleneck is expected to persist at least until the end of fiscal year 2028.

– Revenue per gigawatt increases with each generation: $18 billion with Hopper, $25 billion with Grace Blackwell, $40 billion with Vera Rubin. Vera Rubin is expected to account for about 20% of data center revenue in the third quarter and to be the fastest product ramp-up in the company’s history.

– Half of the business comes from outside the hyperscalers: The ACIE segment generated $40 billion, up 138% compared to the previous year. Neocloud partners are expected to close the year with 8 GW of installed capacity, up from approximately 3 GW at the end of 2025.

– $AMZN (+0,05 %) is expanding the partnership: an additional 2 million GPUs starting this quarter through the second quarter of fiscal year 2029. The five largest hyperscalers are expected to reach just under $800 billion in 2026, and around $1.3 trillion.

– Memory prices are squeezing margins: from 75% to 74% in the third quarter; the low point is expected to be 71–72% in the fourth quarter. Kress (CFO) describes the price trend as extreme and well above the company’s own expectations—citing the AI expansion itself as the cause.

– Kress (CFO) on circle financing: just under $50 billion has been invested in Frontier Labs, in addition to financing platforms involving six investors totaling more than $500 billion in third-party capital. “We do not issue loans.” These labs are expected to generate account for about a quarter of total revenue.

– China is left out: Less than 1% of data center revenue came from China in the second quarter; the outlook projects zero revenue from China.

$NBIS (+0,49 %)$CRWV (-0,02 %)$IREN (-0,04 %)

$NVDA (-0,02 %)

The server systems of Spitzenchips are affected Vera Rubin and Grace Blackwell, which will be delivered . The amount varies depending on the chip generation and memory configuration. The reason cited is the extremely high prices for RAM ($MU (-0,16 %)$SKHY (-3,17 %)$005930). As a result, part of the value creation in the AI stack is shifting downward—away from the computing chip and toward memory. Just how much of this NVIDIA itself is absorbing is likely to be reflected in the gross margin in the quarterly report on August 26 .

Overall, , this affects hyperscalers like Neoclouds and AI labs, which consequently face even higher costs —because Nvidia chips represent the largest cost factor in AI data centers.

Consequently, GPU rental prices will also continue to rise, which will have a neutral effect for these companies.

The extremely expensive financing, which continues to be weighed down by macroeconomic uncertainty. This price increase further adds fuel to the fire. 🔥

$GOOGL (+0,27 %)$AMD (-0,46 %)$INTC (-0,25 %)$MSFT (+0,06 %)$META (+0,45 %)$IREN (-0,04 %)$CRWV (-0,02 %)$NBIS (+0,49 %)$AVGO (+0,67 %)

Source:

www.bloomberg.com/news/articles/2026-08-22/nvidia-customers-notified-about-ai-related-price-hikes-above-15

Good morning! Here's my latest earnings calendar from my vacation in Sardinia—my last two are coming up. Have a great Sunday, everyone!

$PDD (-0,42 %)
$XPEV (-1,2 %)
$EH (+1,55 %)
$BOX (-0,22 %)
$ZM (+0,51 %)
$LI (-0,72 %)
$JKS (-2,04 %)
$KSS (-0,09 %)
$NVDA (-0,02 %)
$OKTA (+1,61 %)
$CRM (+0,89 %)
$SNPS (+0,49 %)
$CRWD (+1,46 %)
$HPQ (-0,47 %)
$DRO (-3,08 %)
$PNG (+1,65 %)
$WDAY (+0,72 %)
$ULTA (-0,12 %)
$MRVL (-0,18 %)
$ADSK (+0,19 %)
$AFRM (-0,4 %)
$S (+1,35 %)
$IREN (-0,04 %)

$NVDA (-0,02 %) has secured, together with SB Energy (a subsidiary of SoftBank $SFTBY (+0,37 %)), secured the land, power connection, and building shell for a new data center project at the PORTS-Pike Campus in Pike County, Ohio.

The roles are clearly defined: SB Energy will build, own, and operate the data center and lease it to OpenAI for 20 years. NVIDIA is providing credit support for the expansion, investing $1.5 billion in SB Energy, and supplying the GPU chips.

The first phase includes 4.25 IT-GW of computing power, with the option to expand to a total of 8 GW.

This makes NVIDIA itself a “neocloud”:

The company is installing its chips in the leased building shells to lease the computing capacity exclusively to OpenAI.

What happens if OpenAI doesn’t use this facility?

NVIDIA says that, alternatively, the capacity can simply be subleased, making the compute a “profitable and financeable” asset.

This is literally the financial thesis behind Neoclouds: $CRWV (-0,02 %), $IREN (-0,04 %) and $NBIS (+0,49 %) they raise billions in debt, secured against GPU and data center assets—based precisely on the bet that this capacity will retain a residual value and remain available for subleasing.

For the Neocloud industry, this deal has two sides:

First: With these exclusive locations, NVIDIA becomes a direct competitor—the manufacturer secures the best property, the tenant, and the financing for itself, rather than leaving them to a Neocloud provider.

Second: At the same time, the chip manufacturer—of all companies—is publicly endorsing the residual value logic on which the entire model is based—and is even guaranteeing the financing.

In my view, the concern is justified:

If NVIDIA itself acts as a Neocloud provider, potential partners will become competitors. Nevertheless, the deal can also be seen as a positive sign—if the manufacturer backs the interchangeability and residual value thesis with its own capital, that is the strongest conceivable seal of approval for the very business model on which CoreWeave, IREN and Nebius are built on.

https://nvidianews.nvidia.com/

$IREN (-0,04 %)

New institutional purchases from recent SEC filings (as of June 30, 2026):

Goldman Sachs $GS (-0,35 %)

33,639,454 shares (approx. $1.5 billion) / 9.4%

→ Filed as a passive major shareholder (Schedule 13G)

Norges Bank (Norway’s sovereign wealth fund)

5,197,774 shares (approx. $238 million) / ~1.4%

→ Entirely new position

JP Morgan $JPM (+0,06 %)

2,699,785 shares (approx. $124 million) / ~0.75%

→ +45.8% from the previous quarter (+approx. 850,000 shares)

BIT Capital

approx. 7,000,000 shares / ~1.9%

→ IREN remains the largest portfolio position

Geode Capital

approx. 2,500,000 shares (approx. $114 million) / ~0.7%

Unfortunately, the exact position sizes from Q1 cannot be determined; nevertheless, it is noteworthy how heavily institutions $IREN (-0,04 %) have been buying shares.

$IREN (-0,04 %) has received $NVDA (-0,02 %) received “Exemplar Cloud” status for its deployment of the GB300 NVL72—a seal of quality that NVIDIA awards only after conducting its own review of the infrastructure. This was announced at the same time as the handover of Horizon 1 to Microsoft.

Following certification for the HGX B300 (training workloads) at the end of June (June 29, 2026), IREN has now also received the certification for the GB300 chips.

What this means: NVIDIA has tested and certified IREN’s GB300-NVL72 setup (the current Grace Blackwell Ultra rack system) at the Horizon 1 data center. This certification confirms that the environment can handle demanding AI workloads with the required performance, reliability, and scalability. It serves as a seal of approval from the chip manufacturer for the operator.

The second part of the announcement: Horizon 1 is the first of four 50 MW (IT load) direct liquid-cooled AI cloud tiers that IREN will deliver to Microsoft in 2026 at its Childress (Texas)—as part of the five-year, $9.7 billion contract announced in November 2025. Horizon 2–4 are set to follow later this year.

Overall, these announcements primarily build confidence and underscore the quality of IREN’s infrastructure.

www.iren.com