![\"IREN\u2019s](\"https:\/\/www.newsbeep.com\/us\/wp-content\/uploads\/2026\/05\/imageforentry14-ner.jpg\")
\t<\/p>\n\t\t<\/p>\n
\u00a9 zhongguo \/ Getty Images<\/p>\n
The artificial intelligence boom has kicked off one of the largest infrastructure races since the early internet era. Nvidia (NASDAQ:NVDA<\/a> | NVDA Price Prediction<\/a>) GPUs are selling faster than they can be packaged. High-bandwidth memory has become scarce enough that companies like Micron Technology (NASDAQ:MU<\/a>) and SK hynix are effectively sold out months in advance. Utilities from Virginia to Texas are warning that grid demand is accelerating faster than transmission upgrades can keep pace.<\/p>\n Now comes the next bottleneck \u2014 and it may be the biggest one yet. What happens when AI companies can\u2019t even get power connected<\/a> to new data centers?<\/p>\n That\u2019s the warning Daniel Roberts, CEO of IREN (NASDAQ:IREN<\/a>), delivered during an interview with Bloomberg Tech. His 11-word statement should grab every data center operator by the collar:\u00a0<\/p>\n \u201cIf you wanted to start today and build a gigawatt AI factory, you are looking 2030 before you get the first compute online.\u201d<\/p>\n The AI Arms Race Is Running Into Physical Limits<\/p>\n For the last two years, investors largely viewed AI as a semiconductor story. Nvidia\u2019s revenue climbed from $27 billion in fiscal 2023 to $215.9 billion in fiscal 2026, but GPUs were only the first layer of the stack.<\/p>\n AI infrastructure requires enormous amounts of electricity, cooling, networking equipment, fiber connectivity, and land positioned near major transmission lines. That combination is becoming harder to secure by the quarter.<\/p>\n Roberts explained that before operators even begin construction, utilities may take 18 to 24 months simply to determine whether a site has available power capacity<\/a>. That is the starting point \u2014 not the finish line.<\/p>\n In plain English, the AI boom is colliding with the realities of the electrical grid.<\/p>\n Consider the scale involved:<\/p>\n Company Microsoft (NASDAQ:MSFT<\/a>) Amazon (NASDAQ:AMZN<\/a>) Meta Platforms (NASDAQ:META<\/a>) IREN According to the International Energy Agency, global electricity demand from data centers could more than double by 2030. AI facilities are the largest contributor.<\/p>\n That means the companies that already control power access may hold the real advantage \u2014 not necessarily the ones still shopping for land.<\/p>\n Why Existing Infrastructure May Become the Most Valuable Asset<\/p>\n Surprisingly, Roberts\u2019 comments suggest the market may still underestimate the value of \u201cready-to-power\u201d infrastructure.<\/p>\n For years, IREN was viewed primarily as a Bitcoin mining operator. But the company spent years securing land, power agreements, cooling systems, and transmission connectivity in regions with scalable energy access. In today\u2019s environment, those assets may be harder to replicate than investors realize.<\/p>\n That matters because a 1-gigawatt AI factory is enormous. For comparison, 1GW roughly equals the electricity demand of hundreds of thousands of homes. Utilities cannot simply flip a switch and accommodate that level of consumption.<\/p>\n Granted, hyperscalers still possess the financial firepower to spend aggressively. Alphabet (NASDAQ:GOOG<\/a>), Microsoft, Amazon, and Meta collectively generated more than $800 billion in operating cash flow over the past year. But cash alone cannot accelerate utility approvals or build transmission lines overnight.<\/p>\n
\nPlanned AI\/Data Center Capacity
\nKey Constraint<\/p>\n
\nTens of billions in AI infrastructure spending in fiscal 2026
\nPower availability<\/p>\n
\nExpanding AWS AI regions globally
\nTransmission bottlenecks<\/p>\n
\nMulti-gigawatt AI cluster ambitions
\nGrid interconnection delays<\/p>\n
\nExisting powered campuses and expansion pipeline
\nScaling advantage from early infrastructure ownership<\/p>\n