Data Centers

When the people building the infrastructure start selling pieces of it, that's not confidence, that's balance-sheet management. Data-centre developers are reportedly racing to offload billions in stakes to sovereign wealth funds and infrastructure investors, spreading the risk of a capex cycle that assumes AI demand keeps compounding at today's rates. It's a sensible hedge if you're the seller and a bet on someone else's optimism if you're the buyer. Watch which sovereign funds show up on the buy side, because that tells you who's still willing to underwrite the AI infrastructure story at scale.

Political backing means nothing if the National Grid can't deliver the power on time, and that's exactly the wall a Starmer-backed AI data centre has hit. The project is now seeking alternative power arrangements after grid connection delays threatened its timeline, a familiar story in the UK where connection queues can run past 2030 in some regions. This is the practical bottleneck nobody's pricing into UK AI investment pledges: announcing gigawatts of data centre capacity is easy, getting grid connection dates that match is not. Any operator planning UK data centre capacity should be negotiating private wire or on-site generation now, not waiting on the grid queue.

Blackstone's QTS Data Centers has abandoned a planned facility in Virginia after sustained local opposition, and the lesson for data centre developers is not about public relations: it is about site selection risk being underpriced at the deal stage. Northern Virginia already hosts the world's highest concentration of data centre capacity, and communities that once welcomed the tax base are now pushing back on power draw, water consumption, and the visual and acoustic footprint of hyperscale builds. The broader consequence is that US data centre development is being forced into less-contested geographies, extending build timelines and raising capex. For UK investors in data centre REITs or infrastructure funds, Virginia's resistance is a leading indicator: planning and community risk is now a material line item that belongs in underwriting models, not a footnote.

Databricks' former AI chief has launched a startup targeting a 1,000-fold reduction in AI inference energy consumption, a number that sounds like marketing until you price the current trajectory. Hyperscale AI inference is on course to consume more power than several European countries combined by the end of the decade, and every percentage point of efficiency gained is worth billions in avoided data centre capex. The 1,000x figure is almost certainly aspirational for near-term deployment, but a credible 10x improvement in inference efficiency would materially change the economics for any company currently rationing AI workloads due to cloud compute costs. UK operators running AI at scale should be tracking this closely: the British grid simply cannot absorb a proportional build-out of AI infrastructure at current efficiency levels, and a hardware or architecture breakthrough changes the feasibility calculus for domestic deployment.

Microsoft and Chevron are partnering on one of the largest gas-powered data centre projects in the United States, sited in West Texas where Chevron has direct access to Permian Basin supply. The deal is honest in a way most hyperscaler energy announcements are not: rather than routing fossil gas through a renewable energy certificate and calling it clean, this is an explicit acknowledgement that AI inference loads are outpacing what the grid and current renewables capacity can deliver. The business logic is unambiguous. Microsoft gets dedicated, dispatchable power; Chevron monetises stranded gas at industrial scale. The reputational exposure sits with Microsoft's 2030 carbon-negative commitments, which this project does not obviously advance. For UK operators watching the energy-AI nexus, this is the clearest signal yet that the path to AI buildout runs through fossil fuel infrastructure for at least the next five years, whatever the sustainability slides say.

The automaker is developing entirely new sodium-ion chemistry with startup Peak Energy, targeting grid-scale energy storage rather than vehicles. GM expects trial production by 2028 at its new Battery Cell Development Center, positioning itself as the first major non-Chinese automaker to manufacture sodium-ion cells. Sodium-ion offers cheaper, safer storage using abundant materials, though with lower energy density that makes it better suited for stationary applications. With AI data centres driving explosive battery storage demand, GM is betting its $900 million chemistry investment can capture scale beyond automotive applications.

Nvidia posted record quarterly revenue of $81.6 billion, up 85% year-on-year, as data center sales hit $75.2 billion on explosive demand for AI agents and generative AI workloads. The company has cycled past last year's $4.5 billion inventory charge on China-targeted H20 chips, with gross margins recovering to the mid-70% range as Blackwell platform sales scale. Revenue has grown 10x in three years, reshaping semiconductor power dynamics and making Nvidia the de facto standard for AI infrastructure from hyperscalers to enterprises.

NextEra is acquiring Dominion Energy for about $67 billion in the biggest power deal ever, explicitly targeting AI's electricity crunch. The combined entity would control over 130 gigawatts of large-load opportunities, with Dominion serving 450 data centers in Northern Virginia's 'Data Center Alley' where AI accounts for 24 percent of electricity sales. PJM wholesale power costs jumped 75.5 percent year-over-year to $136.53/MWh in Q1, while capacity prices hit the regulatory cap of $329.17/MW-day as 94 percent of forecast load growth through 2030 comes from data centers. NextEra's CEO John Ketchum has pivoted from pure renewables to an 'all forms of energy' model specifically to supply 24/7 baseload for AI workloads. The deal signals that AI's power hunger has reached mega-cap M&A scale, turning electricity access into the new strategic constraint for hyperscalers.

Google is negotiating omnibus licensing agreements with Blackstone, KKR, and EQT to give their portfolio companies access to Gemini AI models, treating private equity as a distribution channel for enterprise AI. The approach contrasts with OpenAI and Anthropic, which are building dedicated consultancy entities that embed engineers in portfolio companies. Google's strategy focuses on platform licensing rather than services, leveraging a $750 million partner fund and existing consulting relationships with Accenture, Deloitte, and others. Blackstone already has stakes in OpenAI, Anthropic, and CoreWeave, positioning itself as a distribution hub for multiple AI labs rather than committing to any single provider. The race to capture PE-controlled mid-market companies reflects AI's shift from one-off pilots to portfolio-wide deployments.

The private equity giant is capitalizing on AI infrastructure demand through public markets.

Network equipment demand surges as hyperscalers build out data center infrastructure.

Gautam Adani is building the world's largest renewable-powered AI infrastructure network. His $100 billion commitment targets 5 GW of data centers by 2035, powered entirely by the Khavda project's 30 GW capacity. Adani Energy Solutions shares trade at 211x earnings, reflecting the AI proxy mania that has gripped Indian markets. The integrated play spanning ports, power, and data centers creates a $250 billion ecosystem that could reshape India's position in the global AI race. The real test comes when hyperscale loads meet the reality of 100 percent renewable power promises.

CoreWeave's AI infrastructure boom hit reality with a $452 million Q4 loss, nearly double Wall Street's $0.49 per share estimate. The company faces securities class action lawsuits alleging it hid delays and overstated scaling capabilities. With $14 billion in debt, rising interest costs, and dependence on third-party data center developers, CoreWeave's $55.6 billion backlog looks increasingly precarious. Shares remain 189 percent above fair value estimates despite the legal overhang, suggesting the AI infrastructure bubble has further to deflate.

Meta stock fell 13% this week after the company raised 2026 capex guidance to $145 billion, driven by AI infrastructure and a Texas data center that ballooned from $1.5 billion to $10 billion. Legal defeats in New Mexico and Los Angeles compounded investor anxiety, while the company delayed rolling out a new AI model due to performance issues. Meta is simultaneously cutting hundreds of jobs to contain costs while pursuing what executives call "frontier ambition" in superintelligence. The 28% drop from recent highs reflects a broader reckoning: investors are no longer willing to fund massive AI bets without clear monetization timelines. Analysts still see 62% upside to $850, but that requires faith in returns years away.

Amazon raised quarterly capex to $31.4 billion, implying an annualized pace exceeding $118 billion as AWS chases AI demand. The company plans $200 billion for data centers and specialized chips, including $50 billion for US government AI infrastructure with 1.3 gigawatts of power capacity. CFO Brian Olsavsky confirmed AWS drives the largest share of spending, with elevated levels expected through 2026. Amazon issued $12 billion in bonds last year to fund the buildout, positioning against Alphabet's $85 billion capex target. The scale reflects hyperscalers' recognition that AI infrastructure is winner-take-all: fall behind now, lose the cloud wars permanently.

Nuclear startups are suddenly bankable as hyperscale data centers devour electricity faster than grids can supply it. X-Energy raised $1.02bn in its public debut with Amazon leading the charge, betting small modular reactors can solve AI's power bottleneck within the decade. The IPO values X-Energy at $3.2bn despite zero commercial reactors operating, a premium that reflects Silicon Valley's desperation for carbon-free baseload power. Amazon needs 40GW of new capacity by 2030 for its data center expansion, equivalent to powering 30 million homes. Traditional renewables cannot match nuclear's density and reliability, making these reactors the only viable path to AI scale without carbon guilt.

Anthropic committed $21 billion to Broadcom for nearly one million Google-designed TPU v7p AI chips, delivered as fully assembled rack systems for direct data center deployment. The deal bypasses Google as an intermediary and forms part of Anthropic's broader $10 billion commitment to Google's TPU infrastructure as the Claude maker scales from $1 billion to $7 billion in annualized revenue in nine months. Broadcom CEO Hock Tan is targeting $100 billion in AI chip revenue by 2027, positioning the company as a key alternative to Nvidia's GPU dominance through custom silicon partnerships.