“Neocloud” describes a new class of cloud provider built specifically to rent out GPU compute for artificial-intelligence workloads GPU-as-a-Service (GPUaaS) rather than the general-purpose compute, storage and software services of the traditional hyperscalers. Over the past eighteen months this niche has become one of the most consequential and hotly debated corners of public markets, because it is the purest listed expression of the AI infrastructure build-out. The two bellwethers, CoreWeave and Nebius, both reported blow-out second-quarter 2026 results in mid-August: CoreWeave grew revenue 112% to $2.58 billion with a contracted backlog of $104 billion, while Nebius grew 454% to $582 million and signed four separate contracts each worth more than $1 billion in a single quarter. Both stocks have multiplied over the past year. This report frames the neocloud opportunity through a deliberately macro lens, because the sector cannot be understood in isolation from three forces larger than any single company: the trajectory of AI capital spending (heading toward roughly $690 billion across the hyperscalers in 2026), the cost and availability of capital in a higher-for-longer rate regime, and the physical ceiling imposed by electricity and grid capacity. Our central conclusion is that the neocloud model is genuine and the demand is not (yet) a mirage but the equities are priced for a continuation of near-perfect conditions, and the risk/reward within the group is decided less by growth (which is uniformly spectacular) than by balance-sheet quality, contracted-revenue coverage, and the ability to convert contracted power into live, revenue-generating capacity. For an investor already positioned in the theme, the actionable questions are which business models survive a funding shock and where structural economics actually accrue which, uncomfortably, is disproportionately to Nvidia rather than to the neoclouds themselves.
Macro notes
Short views on rates, FX and sector dynamics that frame the assumptions behind our coverage.
Top Call: This is a hawkish-Fed, higher-for-longer tape . September is live for a hike, not a cut. The single thing to keep front of mind: the Warsh Fed held at 3.50–3.75% on July 29 but voted 9-3, with three members favoring a quarter-point rate hike the first time since September 2016 that three policymakers dissented with a unified view. The central bank hinted at a possible hike at its next meeting in September. With the July jobs report due August 7 and CPI on August 12, we get the two prints that decide whether September is a genuine hike. Position for asymmetry: a hot number reprices the front end and hits duration/growth hard; a soft one only gets you back to "extended hold." We are not in an easing regime trade the tape accordingly.(via CNBC)