Morgan Stanley projects hyperscaler compute capacity to nearly quadruple from 31 to 117 gigawatts by 2028, requiring up to $8 trillion in capital spending. <\/p>\n<\/li>\n
Nebius posted 684% year-over-year revenue growth and targets between $7 billion and $9 billion in annualized revenue by year-end, with a long-term goal of $51 billion by 2030. <\/p>\n<\/li>\n
Nebius’s new asset-light partnership model lets third-party owners fund data centers while Nebius supplies its platform and earns recurring fees, removing the self-financing burden. <\/p>\n<\/li>\n
Act now: the analyst who called NVIDIA in 2010 just named his top 10 AI stocks \u2014 and Nebius Group didn’t make the cut. Grab the names FREE today<\/a>. <\/p>\n<\/li>\n<\/ul>\n Artificial intelligence stocks have spent much of this year whipsawing between optimism and doubt. Investors embraced companies building AI infrastructure, then abruptly questioned whether too much computing capacity was coming online too quickly.\u00a0 <\/p>\n That shift hit neocloud providers like CoreWeave<\/strong> (NASDAQ:CRWV<\/a>), Nebius Group<\/strong> (NASDAQ:NBIS<\/a>), and IREN<\/strong> (NASDAQ:IREN<\/a>) especially hard because their businesses depend on renting cutting-edge GPU clusters to AI developers. If hyperscalers ended up with excess capacity, the conventional wisdom believed specialized providers would lose their competitive edge.\u00a0 <\/p>\n But a new bottom-up model from Morgan Stanley suggests the market focused on the wrong risk. Instead of oversupply, the next several years may be defined by one of the largest infrastructure buildouts the technology industry has ever attempted. <\/p>\n The AI Buildout Is Only Accelerating <\/p>\n The investment bank estimates total compute capacity across the five major hyperscalers will climb from 30.5 gigawatts in 2025 to 116.6 gigawatts by 2028. That is nearly a fourfold increase in only three years. Every one of those additional 80 gigawatts must be designed, financed, and constructed from scratch, translating into an estimated $4 trillion to $8 trillion of capital spending between now and 2028. <\/p>\n The Morgan Stanley analysis illustrates just how ambitious those plans are. <\/p>\n Act now:<\/strong> the analyst who called NVIDIA in 2010 just named his top 10 AI stocks \u2014 and Nebius Group didn’t make the cut. Grab the names FREE today<\/a>. <\/p>\n Company<\/strong> <\/p>\n<\/td>\n 2025 Compute Capacity<\/strong> <\/p>\n<\/td>\n 2028 Compute Capacity<\/strong> <\/p>\n<\/td>\n % Increase<\/strong> <\/p>\n<\/td>\n<\/tr>\n Amazon<\/strong> (NASDAQ:AMZN<\/a>) <\/p>\n<\/td>\n 13.8 GW <\/p>\n<\/td>\n 35.8 GW <\/p>\n<\/td>\n 159% <\/p>\n<\/td>\n<\/tr>\n Google <\/p>\n<\/td>\n 5.0 GW <\/p>\n<\/td>\n 31.6 GW <\/p>\n<\/td>\n 532% <\/p>\n<\/td>\n<\/tr>\n Microsoft<\/strong> (NASDAQ:MSFT<\/a>) <\/p>\n<\/td>\n 7.5 GW <\/p>\n<\/td>\n 20.3 GW <\/p>\n<\/td>\n 171% <\/p>\n<\/td>\n<\/tr>\n Meta Platforms<\/strong> (NASDAQ:META<\/a>) <\/p>\n<\/td>\n 3.5 GW <\/p>\n<\/td>\n 21.2 GW <\/p>\n<\/td>\n 506% <\/p>\n<\/td>\n<\/tr>\n SpaceX<\/strong> (NASDAQ:SPCX<\/a>) <\/p>\n<\/td>\n 0.7 GW <\/p>\n<\/td>\n 7.8 GW <\/p>\n<\/td>\n 1,014% <\/p>\n<\/td>\n<\/tr>\n Hyperscaler Compute Capacity<\/strong> <\/p>\n<\/td>\n 30.5 GW <\/p>\n<\/td>\n 116.6 GW <\/p>\n<\/td>\n 282% <\/p>\n<\/td>\n<\/tr>\n Source: Morgan Stanley bottom-up hyperscaler compute model. <\/p>\n Story Continues <\/p>\n Those numbers destroy the excess compute bear narrative that crushed AI infrastructure stocks earlier this year. Demand doesn’t expand fourfold over three years if the industry is drowning in unused capacity. <\/p>\n 24\/7 Wall St. <\/p>\n Fears of an AI oversupply were wrong. Now, an $8 trillion infrastructure race is officially accelerating. \u00a9 24\/7 Wall St. <\/p>\n Why Nebius Still Has Room To Grow <\/p>\n That demand outlook helps explain why companies like Nebius occupy an important niche. <\/p>\n First-quarter revenue expanded 684% year over year, while management reiterated expectations for a $7 billion to $9 billion annualized revenue run rate by the end of the year. Longer term, the company has outlined a path toward $51 billion in annual revenue by 2030 based solely on the infrastructure it owns. <\/p>\n Today, that growth story became even more compelling. Nebius announced a new asset-light infrastructure partnership model in which third-party data center owners finance new facilities while Nebius contributes its AI cloud platform, customer relationships, and operating expertise.\u00a0 <\/p>\n The arrangement allows Nebius to expand into new regions without tying up billions of dollars on its balance sheet. Instead of funding every new data center itself, the company can generate recurring licensing fees and revenue-sharing income while partners supply the capital. That dramatically increases the pace at which Nebius can scale and gives management another growth lever beyond simply adding company-owned infrastructure. <\/p>\n Key Takeaway <\/p>\n In short, the bear case against neocloud providers rested on two assumptions: AI demand would cool, and companies like Nebius would struggle to finance enough infrastructure to keep growing. Morgan Stanley’s compute model challenges the first assumption by projecting hyperscaler capacity to jump from 30.5 gigawatts to 116.6 gigawatts by 2028. Nebius’s new partnership model addresses the second by removing much of the capital burden associated with expansion. <\/p>\n Granted, execution risk remains, and Nebius still needs to deliver on its ambitious growth targets. But if Morgan Stanley’s projections prove accurate, the market’s 35% sell-off of Nebius stock increasingly looks like a reaction to a problem that never existed. For patient investors looking beyond the next quarter, that disconnect may be where the opportunity lies. <\/p>\n Act now:<\/strong> the analyst who called NVIDIA in 2010 just named his top 10 AI stocks \u2014 and Nebius Group didn’t make the cut. Grab the names FREE today<\/a>. <\/p>\n<\/a> 24\/7 Wall Street <\/p>\n
\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n