Elon Musk has quietly acquired APR Energy, a mobile power generation company, for more than $1 billion to secure electricity supplies for xAI's data centers that power the Grok artificial intelligence chatbot. The deal, which became public through a Federal Trade Commission filing, reveals how aggressively Musk is addressing one of the biggest bottlenecks in AI development: reliable, fast-deployable power infrastructure.

Why Does an AI Company Need a Power Generation Business?

Training and running massive AI models consumes staggering amounts of electricity. xAI's Colossus and Colossus 2 supercomputers in Memphis, Tennessee, which train the Grok AI model, have previously relied on mobile gas turbines to supplement grid power. By owning APR Energy outright, Musk gains direct control over a fleet of mobile generators that can be deployed near data centers within days, rather than waiting years for traditional power plants to be built and connected to the grid.

APR Energy operates gas turbines and diesel- and natural gas-fired generators mounted on trailers. These units can begin generating electricity in less than 10 minutes and be installed in just days, making them far faster than conventional infrastructure. For a company racing to scale AI computing power, this speed advantage is critical.

What's the Timeline for Grok 4.6, and How Does It Compare to Competitors?

On July 18, Musk announced that xAI's next-generation Grok 4.6 model, featuring 2 trillion parameters, is entering its final training phase and is expected to complete training this week. The model aims to match the inference efficiency of its predecessor while potentially surpassing Moonshot AI's recently released Kimi K3, which features 2.8 trillion parameters.

The competitive landscape is heating up. Kimi K3, released on July 17, has already demonstrated strong performance on third-party benchmarks. On the Arena AI front-end coding leaderboard, K3 topped global rankings with a score of 1,679, surpassing Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol. However, Moonshot AI acknowledged that K3's overall user experience still lags significantly behind these top-tier closed-source models.

When Musk commented "impressive" on a review of Kimi K3, the Chinese AI startup's official account responded playfully on Weibo, posting "@ElonMusk Welcome to the '2 Trillion+' club." The full model weights of Kimi K3 are set to be open to developers worldwide before July 27.

How to Understand the Computing Power Behind Grok 4.6

• GPU Infrastructure: The Grok 4 series models rely on the "Colossus" computing cluster for training. Colossus 2, which became operational in January 2026, launched with approximately 555,000 GB200 and GB300 series GPUs, equivalent to roughly 1.4 million H100 GPUs in raw computing power.
• Expansion Plans: xAI previously revealed that the ultimate goal for Colossus 2 is a capacity of 2 gigawatts. In June 2026, the company planned to add more than 400 megawatts of computing capacity by deploying over 220,000 GB300 chips, with a final target of 1 million GPUs.
• Model Scale Progression: xAI completed training of its 1.5-trillion-parameter base model V9-Medium in May 2026, which serves as the core foundation of Grok 4.5. The jump to 2 trillion parameters in Grok 4.6 represents a significant scaling effort.

What Are the Environmental and Regulatory Challenges?

The acquisition comes as xAI faces environmental scrutiny over its power generation methods. The NAACP, the Southern Environmental Law Center, and Earthjustice have filed a lawsuit against xAI, alleging violations of the Clean Air Act. The groups estimate that operating the turbines without emissions permits could produce more than 2,000 tonnes of nitrogen oxide (NOx) emissions annually, potentially worsening air quality in the Memphis area.

This environmental tension highlights a paradox in Musk's business empire. Tesla and SolarCity position him as a clean energy advocate, yet the surging electricity requirements of AI computing have made natural gas-fired generation the fastest available solution compared with building renewable energy facilities or expanding the national electricity grid. By acquiring APR Energy, Musk is prioritizing speed and scale over renewable alternatives.

The FTC cleared the acquisition to proceed without further antitrust review, concluding it was unlikely to create a monopoly. However, the deal underscores how critical power infrastructure has become in the AI arms race. As companies compete to train larger models faster, controlling electricity supply has become as strategically important as controlling computing hardware.

Industry analysts note that while parameter count remains a primary metric for measuring AI model scale, actual competitiveness depends on multiple factors: training data quality, algorithmic optimization, inference speed, and computing power supply. Whether xAI can maintain the operational efficiency of previous generations while expanding Grok 4.6 to 2 trillion parameters will be key to whether it can pose a substantial challenge to Kimi K3 and other top-tier models.