IREN — which is now a “neocloud,” renting third-party chips to others — reported its fiscal fourth-quarter results at the end of August, at which time it disclosed it had achieved the Nvidia (NVDA -1.10%) “preferred partner” status and rose on that news. But September saw a increase in projections for AI computing demand over at least the next few years, on the heels of extremely bullish long-term forecasts from several AI companies, especially Oracle (ORCL 2.01%), OpenAI, and Nvidia.
IREN also helped its own cause in September when it announced it had doubled its supply of graphics processing units (GPUs), spurring the company to double its already bullish projections for its end-of-year annualized revenue run rate.
On September 8, IREN gave its monthly revenue update. Most of IREN’s revenue still currently comes from Bitcoin mining, and it saw only a slight increase in its fledgling AI cloud data center business in August, recording $2.4 million in revenue relative to $2.3 million in July.
However, on its August earnings call, IREN said it expects its AI cloud revenue to increase by 8 to 10 times just by December. Then, late in September, it hiked that forecast once again.
Early in the month, database and cloud giant Oracle announced an absolutely massive increase in its future remaining performance obligations for its cloud unit, to $455 billion. It turns out ChatGPT parent OpenAI accounts for about $300 billion of that backlog, which spans 2027 through 2031. Then, later in the month, Nvidia announced it would invest up to $100 billion in OpenAI to fund another 10 gigawatts of AI data centers.
These massive deals show incredible demand for AI compute and potential supply tightness regarding access to power and AI chips. That’s why IREN, with an existing 2.91 GW of power and another 750 megawatts contracted, should see high demand for its existing capacity.
Then, on Sept. 22, IREN also disclosed it had secured 7,100 Nvidia B300s, 4,200 Nvidia B200s, and 1,100 Advanced Micro Devices (AMD 23.61%) MI350Xs, for a total of $674 million. The 12,400-chip increase had the effect of doubling IREN’s fleet of GPUs. Along with the announcement, IREN raised its AI cloud annualized revenue run-rate guidance to $500 million by January. On its late-August earnings release, it had previously forecast $250 million in AI cloud ARR by December.
After its September surge, IREN’s market cap has rocketed to $15.7 billion, which is quite striking for a company with a tad over $500 million in revenue over the past 12 months. IREN is slightly profitable, thanks to its Bitcoin mining operations. However, the stock is now pricing in a strong runway of profitable AI cloud growth, which entails buying Nvidia and AMD GPUs and renting them out.
If the AI boom goes the way OpenAI and the optimists believe — if the technology transforms everything rapidly and there is endless demand for computing — the price movements in these neoclouds could be justified. However, as the dot-com bubble has taught us, if this story is interrupted by an economic downturn, or if AI turns out to be less transformative less quickly, these soaring stocks could be in for a rude awakening.
