As the artificial intelligence race shifts from a battle of pure algorithms to a contest for physical infrastructure, tech giants are sparing no expense to lock in energy supplies through mergers, acquisitions, and vertical integration. Against a backdrop of increasingly tight power supply, "speed" and "grid access rights" have become critical assets determining victory.
Google recently spent $4.75 billion to acquire renewable energy developer Intersect Power, a move whose core logic is to bypass the lengthy grid queue. Through this deal, Google directly acquired Intersect's secured grid connection permits, critical equipment orders, and a development pipeline of 8 to 10 gigawatts (GW), enabling it to secure large-scale power supply faster than its competitors. In contrast, while rivals like Amazon are also making moves, Google has clearly traded a premium price for a time advantage in terms of cost per kilowatt and construction speed.
Simultaneously, Elon Musk's xAI is demonstrating extreme engineering speed. On January 17, 2026, EST, Musk announced the official launch of Colossus 2, the world's first single-site AI training cluster with gigawatt-scale computing power. The project went from groundbreaking to operation in under a year and is planned for an upgrade to 1.5 GW by April this year. This pace far exceeds the roadmap plans of competitors like Anthropic and OpenAI, signaling the official entry of AI training into the industrial-scale computing era.
These developments indicate that the scarcity of power infrastructure is reshaping the valuation system and competitive landscape for data centers. Whether it's Google's high-premium acquisition or xAI's "blitzkrieg" construction, both highlight that in the current reality of grid congestion and equipment shortages, possessing self-controlled power and land resources has become a core moat for tech giants.
Paying a "Fast Lane" Premium: Google's Strategic Logic Google's acquisition of Intersect is not just an energy investment but a buyout of supply chain priority. Reports indicate that Amazon recently acquired a bankrupt 1.2 GW solar project for $83 million, whereas Google's cost per gigawatt in the Intersect deal is approximately six times that of Amazon's. This significant premium reflects Google's thirst for speed: Intersect's top-tier grid connection permits and regulatory approvals allow its projects to be ready by around 2028, whereas ordinary projects typically face waits of several years.
More crucially, Intersect holds a favorable position in the supply chain. In Memphis, Tennessee, giant transformers reserved by Intersect are being produced at a factory operated by Hyosung HICO, a subsidiary of South Korea's Hyosung Heavy Industries. Such critical equipment currently faces delivery lead times of up to four years, but Intersect has locked in capacity through long-term partnerships. Intersect CEO Sheldon Kimber likened this capability to "selling not just coffee beans, but high-priced lattes," implying that its mature project assets are more valuable than simple power development.
Google's Data Center Energy Lead stated that this transaction will bring new generation capacity to the grid, rather than merely consuming existing resources. This "bring your own power" model—where data centers solve their own power production and feed it back to the grid—is seen by Google as the "entry ticket" to play.
Texas: The "Disneyland" of Energy Geographically, Google and Intersect have set their sights on the Texas Panhandle. This region not only boasts abundant wind and solar resources but, more importantly, Texas has a unique, relatively independent free-market grid (ERCOT). Intersect CEO Kimber calls Texas the "Disneyland of energy," considering it an ideal place for rapid, cost-effective deployment.
Intersect has secured substantial land reserves in places like Hereford, Texas, and designed a hybrid energy model: using wind, solar, and battery backup to provide most of the power, while using on-site small gas turbines as "balancing power." This model allows renewable energy to account for up to 70% of the mix, while bypassing the 5 to 7-year delivery wait for large gas turbines by connecting to underground natural gas pipelines.
This strategy enables Google to avoid the difficulties of building power plants in the heavily regulated US Northeast. As one observer noted, while efforts were made to restart plants elsewhere, Texas's free-market mechanism already allows data centers to grow rapidly in a "petri dish" like environment.
xAI's "Blitzkrieg": Vertical Integration of Industrial-Scale Computing While Google accelerates through acquisition, Musk is demonstrating astonishing execution through vertical integration. The launch of Colossus 2 means xAI transformed a plot of land into a supercomputing cluster with an energy consumption exceeding the peak electricity usage of San Francisco in under a year.
Analysis suggests xAI's computing power scale curve is rising steeply, while competitors' plans for similar scale remain slated for 2027. Musk revealed the cluster will expand to 1.5 GW in April and boldly claimed that within five years, xAI's computing power will exceed the sum of all other companies. xAI has adopted a strategy of fully self-building infrastructure, designing facilities from the ground up to suit computing loads, rather than relying on cloud providers like Microsoft Azure or Amazon AWS.
This aggressive "build first, legitimize later" expansion also carries regulatory risks. The US Environmental Protection Agency (EPA) pointed out that xAI uses natural gas turbines at its Memphis site to fill power gaps, with some equipment allegedly operating without proper permits. This highlights the tension between pursuing ultimate speed and meeting compliance requirements.
Market Dynamics: Controversies from Water to Nuclear Power As the competition for resources by data centers intensifies, their impact on the environment and public resources is drawing widespread attention. Data shows that while Colossus 2's water usage is significant, it is equivalent to only 2.5 In-N-Out Burger restaurants, lower than public expectations.
Other giants are also seeking solutions. Microsoft is collaborating with the Midwestern grid operator MISO to use AI for optimizing grid management. Meanwhile, Meta Platforms' deal to purchase capacity from an existing nuclear power plant in Ohio has drawn criticism from Princeton University energy expert Jesse Jenkins, who argues it "eats into" public resources and drives up electricity prices rather than adding new clean energy supply. Notably, Google is a funder of Jenkins' research.
Overall, whether through Google's capital maneuvers or Musk's engineering marvels, the facts point to the same conclusion: in the AI era, whoever solves the physical bottleneck of power the fastest will hold the dominant position in computing power.
Comments