All articles tagged with #data centers
Moody's warns that roughly $662 billion in planned data-center expansion is highly concentrated among five firms, creating material concentration risk for lenders and investors if demand weakens, financing tightens, or supply chains falter; the piece emphasizes potential impacts on project financing, asset valuations, and the broader tech infrastructure buildup.
Nvidia unveiled Vera Rubin, its next AI rack-scale system set to ship in H2 2026, boasting about 10x the performance per watt of the Grace Blackwell platform while consuming roughly twice the total power. The 72 Rubin GPUs and 36 Vera CPUs reside in a fully liquid-cooled, modular rack designed for quick swaps and easier upgrades. Competition from AMD’s upcoming Helios and in-house cloud chips from Amazon and Alphabet adds pressure, but Meta plans to deploy Rubin by 2027. Pricing is reportedly around $3.5–$4 million per rack, signaling strong pricing power as AI data-center demand grows. Analysts show a Strong Buy with a notable upside potential.
Trump’s pledge that tech firms should fund their own power generation for AI data centers would only partially shield Americans from rising electricity costs. Experts say the data-center boom requires extensive grid upgrades and infrastructure, and with hundreds of planned centers and vague details on how costs will be shared, ratepayers could still bear a significant portion of the bill.
President Trump is set to announce agreements with major tech firms to shoulder more of the electricity costs for new data centers, a move framed as protecting consumers from rising power bills as AI infrastructure expands. The unnamed pacts, described as “rate payer protection pledges,” aim to keep electricity costs from surging in areas where data centers are built, but critics argue they amount to backroom deals and could influence energy policy ahead of elections.
Meta and AMD announced a multi-year agreement to power Meta's AI infrastructure with up to 6GW of AMD Instinct GPUs, aligning roadmaps across silicon, systems and software as part of Meta Compute; shipments for the first deployments begin in H2 2026 on the Helios rack-scale architecture, marking a diversification of Meta's AI compute stack and accelerating its personal AI ambitions.
Public sentiment toward AI is turning negative as voters worry about children’s safety, job displacement, surveillance, and the energy costs of data centers; polls show high distrust of AI and skepticism about corporate use, with governors noting a looming political backlash that could shape 2028 elections.
Amazon will invest $12 billion to build multi-site data center campuses in northwest Louisiana (Caddo and Bossier Parishes), creating 540 full-time data-center jobs and supporting about 1,710 additional local roles. The plan includes up to $400 million for local water infrastructure and a $250,000 community fund focused on STEM, sustainability and local projects. Amazon will fund energy infrastructure upgrades with SWEPCO, add up to 200 MW of solar energy, and cool centers using surplus water and outside air for most of the year, cutting grid electricity demand by 25-35% during peak periods and pursuing a water-positive goal by 2030. STACK Infrastructure leads construction, with up to 1,500 construction jobs anticipated, aiming to boost regional economic growth and resilience.
OpenAI CEO Sam Altman told the India AI Impact Summit that worries about data-center water use are unfounded and claimed the energy required to train a human over 20 years outweighs the energy used to answer a single ChatGPT query, urging a fair, post-training comparison. He noted advances in cooling that cut water use, referenced global and US projections for data-center electricity growth, and urged a rapid shift to nuclear, wind, or solar power, while also touching on AI-related job-loss concerns.
Meta’s 2026 capex is set to reach roughly $115–$135 billion, funding Nvidia GPU/Grace CPU deployments that will power its AI infrastructure; Nebius Group, a Nebius cloud partner and Meta customer, is forecast to grow revenue to about $3.4 billion in 2026 from $530 million in 2025, bolstered by a backlog over $20 billion, expansion to 16 data-center sites, and 0.8–1 gigawatt of capacity as demand for cloud AI ramps up.
Hedge fund investor Michael Burry questions when the aggressive AI data-center buildout will end, criticizing hyperscalers such as Oracle, Alphabet (Google), Meta, Microsoft, Amazon and Nvidia for expansive capex and potential cash-flow strain. He warns of possible earnings restatements and depreciation masking costs, and likens current AI hype to past bubbles like the 1920s radio boom and the dot-com era.
OpenAI’s plan to rapidly deploy AI compute through the Stargate joint venture with Oracle and SoftBank has stalled, hampered by leadership gaps and disagreements over who does what, leaving the three partners without staffing or actively developing OpenAI’s data centers and slowing the broader race for AI-powered computing power.
At the India AI Impact Summit, OpenAI CEO Sam Altman dismissed sensational claims about AI’s vast energy use, arguing that data centers do consume energy but that the sector should shift to nuclear and renewables. He contends that training and sustaining human civilization itself required far more energy, food, and evolutionary time, prompting a backlash online and highlighting the industry’s transparency gaps in reporting energy and water usage.
Brookfield Corporation is positioning itself as a leading AI infrastructure play, launching a Brookfield AI Infrastructure Fund to deploy up to $100 billion in AI assets with seed commitments from Brookfield and Nvidia, and launching Radiant—an Nvidia cloud partner to build AI data centers. With a broad footprint in infrastructure, power, and credit, Brookfield expects strong earnings growth over the next five years. At about $46 a share, it trades at a discount to an estimated value of roughly $68 per share, with a long-term bull case near $130 by 2030, which is why it’s cited as the highest-conviction AI stock to buy now.
A POLITICO poll finds data centers are moving to the forefront of politics with public opinion still forming; Republicans show a slight edge in support while Democrats voice concerns about energy and water costs, and campaigns are testing how to regulate AI-related infrastructure as more centers are proposed and the issue rises in importance for elections.
As AI infrastructure spending rises, software stocks face pressure while select hardware plays like Broadcom and Micron Technology stand to benefit. The article argues these two are the new pick-and-shovel AI infrastructure stocks, boosted by data-center capex and memory/networking bottlenecks, making them attractive long-term bets amid the tech sell-off.