All articles tagged with #data centres
Soumaya Keynes argues the AI funding binge is real, with AI firms taking a growing share of VC and debt markets and data‑centre demand rising, but evidence of crowding out in the wider economy is inconclusive: BIS data shows AI‑ready economies are investing more, Goldman Sachs finds only a small drag on non‑AI borrowers, and any crowding-out risk remains modest for now, though longer‑term debt and energy constraints could reshape capital allocation.
A $14bn Meta-BlackRock data centre in El Paso is insured only partially—covering limited amounts for property, rent abatement and terrorism—leaving lenders exposed to potential losses in the billions beyond policy caps as full coverage remains unaffordable for AI-scale campuses.
Banks led by Morgan Stanley plan a multi-bond refinancing of about $15 billion tied to a Google-backed data center in Hubbard, Texas that Anthropic will lease, aiming to free lending capacity as AI infrastructure finance heats up. The debt, which features a delay-draw structure and may be split across bonds and leveraged loans, would still carry speculative-grade risk despite Google’s backstop. Nexus Data Centers remains the developer, and the project will house Google TPU chips with separate financing for the chips and a dedicated natural-gas power plant to curb delays and costs. Lenders are also seeking to shed AI-related exposure amid a crowded, high‑capital market for big data-center projects.
Investors are wary of Big Tech’s AI-driven capex as credit-default swaps for Oracle, Nvidia, Alphabet, Amazon, Meta, Broadcom and others hit fresh highs; Oracle’s five-year CDS is around 215 bps and Nvidia about 79 bps, with Meta’s Texas data-centre borrowing costs moving toward junk; Oracle was downgraded to BBB- by S&P amid questions about profitability, while OpenAI/Nvidia discussions for a 10GW Ohio data-centre project highlight the scale of planned spend. Analysts warn AI financing outpaces cash flow, making CDS hedges more common and signaling rising credit risk in the sector.
The author argues that data centres are a strategic test of US industrial resolve, highlighting how China’s dominance in critical minerals and downstream tech was shaped by past American missteps. While local opposition and proposed moratoria raise concerns about energy, water, and permitting, data centres spur investment in domestic manufacturing of advanced electrical equipment and create jobs, aligning demand with rebuilding resilient supply chains. By channeling investment—rather than halting projects—into US workers and clean-energy grid upgrades, the US can strengthen its industrial base and compete with Beijing, avoiding a repeat of the rare-earth era’s missed opportunities.
South Korea unveiled a multi-year, over-$1-trillion plan to lead AI and chip development, with Samsung and SK Hynix committing hundreds of billions to build new chip fabs, large-scale data-centre projects, and regional clusters, aiming to boost capacity by 2035 while balancing regional growth amid political scrutiny.
AI’s data-centre expansion is turning labor into the bottleneck, with a current trades shortfall around 350,000 expected to exceed 1 million by 2030, driving big tech to fund training programs and offer incentives as they invest hundreds of billions in infrastructure.
Scientists have created memory crystals by laser writing nanostructures inside fused silica glass, enabling five-dimensional data storage that could pack hundreds of terabytes on a 5-inch platter and last essentially forever. The read-out uses an optical microscope, with current write/read speeds around 30 MB/s (targeting 500 MB/s in 3–5 years). If scalable, this could dramatically cut data-centre energy use since data centres consume about 1.5% of global electricity and rising, replacing energy-hungry disks and tapes. DNA storage and borosilicate glass are other long-term contenders, but widespread adoption faces compatibility and cost hurdles; in the near term, improving hardware efficiency and smarter data management remain essential.
Moody's warns that current US accounting rules may hide tens of billions in data-centre liabilities for Big Tech by allowing SPV-backed structures and renewal guarantees to stay off balance sheets; lease renewals and residual value guarantees may not be fully recorded unless deemed probable or reasonably certain. Meta and Oracle are among users; Meta's Hyperion facility via Beignet Investor involves a four-year lease with renewal options up to 20 years and RVG guarantees up to $28bn not recorded as of end-2025. Moody's would adjust debt to reflect likely renewal periods or RVG triggers, which could affect credit ratings and access to funding for AI data centres.
Five US tech giants are set to spend about $700 billion on data-centre capex to power AI, a figure larger than last year’s oil-and-gas investment, signaling the early onset of AI financialisation as investors back the infrastructure and anticipate a new wave of AI-linked securities, hedges and collateral.
India's 2026-27 budget prioritizes infrastructure-led growth and a manufacturing push in seven strategic sectors—including rare earths, semiconductors and data centres—with a new tax holiday for foreign cloud investments, no direct personal tax cuts, and a shift to lower debt-to-GDP by 2030-31; markets reacted negatively to a higher securities transaction tax amid global uncertainty.
Despite ongoing expansion of large data centres, experts and industry leaders are debating a future where AI runs on devices or tiny edge centres near users, reducing latency and energy use, while some propose repurposing buildings or exploring space-based data hubs; the shift faces cost, security and reliability hurdles but reflects a move toward more bespoke, local AI processing.
Colder weather boosted heating fuels and a surge in electricity demand from data centres and crypto mining lifted US greenhouse gas emissions by 2.4% last year, with coal up 13% despite solar growth; higher natural-gas prices spurred fuel switching, while transport emissions stayed roughly flat thanks to more hybrids and EVs, and Trump-era policies are viewed as having little immediate impact.
Data centres are increasingly turning to aircraft engines and fossil fuel generators to power AI operations due to delays in grid connection and supply chain issues, leading to a surge in demand for aeroderivative turbines and diesel generators, despite concerns over emissions and higher costs.
Tech companies are shifting $120 billion of AI data centre debt off their balance sheets, indicating a strategic move in financial management within the tech industry.