Good morning.

It’s no secret that AI is power-hungry, and Big Tech is moving quickly to lock in the electricity needed to keep it running. Case in point: Alphabet’s $4.75 billion acquisition of Intersect Power—one of the tech giant’s largest infrastructure deals to date.

Alphabet is buying Intersect Power in an all-cash transaction, bringing a major clean-energy developer directly into Google’s orbit. The deal gives Google tighter control over electricity supply as it scales AI-driven data centers amid mounting grid constraints.

Let’s dive into today’s edition.

SoftBank Group said it will acquire digital infrastructure investor DigitalBridge in a deal valued at $4 billion, as the Japanese group accelerates its push into artificial intelligence–linked infrastructure. The transaction values DigitalBridge at $16 per share, a 15% premium to its prior close, and is expected to close in the second half of 2026.

Infra Push: The deal will expand SoftBank’s exposure to the physical backbone of AI, including data centers, cell towers, fiber networks, small cells, and edge infrastructure. DigitalBridge manages approximately $108 billion in assets and holds stakes in companies including Vantage Data Centers, Zayo, Switch, and AtlasEdge.

Big Picture: The acquisition aligns with SoftBank founder Masayoshi Son’s strategy to capitalize on surging demand for AI computing capacity, following the group’s broader investments alongside OpenAI and Oracle in large-scale U.S. data center and power infrastructure projects.

TSMC said it has begun mass production of its 2-nanometre (2nm) chips, marking the rollout of what it calls the industry's most advanced semiconductor technology. The company said volume production of its N2 process began in the fourth quarter of 2025, as planned. Production is taking place at Fab 20 in Hsinchu and Fab 22 in Kaohsiung, both in Taiwan.

Pocket-Sized Power: The 2nm chips use a nanosheet transistor architecture, allowing significantly higher transistor density and improved energy efficiency compared with previous generations.

Why does this matter for AI? These gains are especially critical for AI workloads, which are increasingly constrained not by software innovation but by power availability, heat dissipation, and infrastructure costs. Smaller nodes enable more compute per chip while reducing electricity consumption, easing one of the biggest bottlenecks facing data centers running large AI models.

NVIDIA has agreed to pay about $20 billion for the assets and technology of AI chip startup Groq, marking the largest deal in NVIDIA’s history and underscoring its bid to deepen leadership in AI hardware.

Why it matters: AI inference — running models in production — is becoming the dominant cost and power bottleneck for data centers. Whoever controls low-latency, energy-efficient inference chips controls the economics of AI at scale.

What’s happening:

Caterpillar has emerged as an unlikely beneficiary of the artificial intelligence boom, with its stock rising 62 percent in 2025 as demand for power equipment surged. Data center developers, facing long waits for grid connections, are increasingly turning to on-site natural gas generators to meet the enormous electricity needs of AI systems.

That shift has driven rapid growth in Caterpillar’s power and energy business, now the company’s fastest-growing segment. Power-generation sales reached $7.8 billion in 2024, up 22 percent from the prior year, and continued to climb in 2025.

In response, Caterpillar is investing $725 million to expand engine production in Indiana and plans to more than double turbine output by 2030.

Data center developers are increasingly relying on aircraft-derived turbines and fossil-fuel generators to meet AI power demand, as supply chain constraints and multi-year grid connection delays stall access to cheaper, cleaner electricity.

According to the Financial Times, manufacturers of aeroderivative turbines—derived from jet engines—and diesel generators are seeing a surge in demand as data center operators bypass grid connections while waiting for larger gas turbines to come online.

By the numbers:

Copper prices are on track for their strongest year since 2009, rising more than 40% in 2025, as artificial intelligence infrastructure emerges as a major new source of demand. Beyond EVs and renewables, AI data centers are now driving incremental copper consumption, requiring large volumes for power transmission, internal wiring, transformers, and cooling systems.

Domino Effect: Analysts say the rapid buildout of AI compute is colliding with tight global supply, slow mine development, and limited inventories, pushing prices to record highs. JPMorgan expects copper to average above $12,000 per metric ton in 2026, citing data center expansion as a key upside risk, while even cautious forecasts see AI spending setting a durable floor under prices.

Big Picture: The rally shows a growing constraint in the AI boom where digital intelligence increasingly depends on physical infrastructure, and copper is emerging alongside power, land, and chips as a key bottleneck to scaling AI globally

Goldman Sachs Group is co-leading financing for a large-scale Texas project to build private power campuses totaling 5 gigawatts to support energy-intensive artificial intelligence data centers. The bank is working alongside Newmark Group to raise both equity and debt for the development, according to GridFree AI, the project’s sponsor.

Key Details: The project, known as South Dallas One, will supply power to a cluster of data centers using modular, natural-gas–fired generation, thereby avoiding immediate reliance on the Texas grid. Each campus will deliver about 1.5 gigawatts of capacity, built in 100-megawatt modules, allowing faster deployment than traditional grid-connected infrastructure.

Fasttracked Power: GridFree AI said initial power could be online in less than 24 months from lease signing. Fuel will be delivered via multiple pipelines, including one owned by Energy Transfer, and the company plans to expand similar AI-ready power campuses globally, including future nuclear-powered sites.