Elon Musk’s artificial intelligence startup xAI is accelerating its infrastructure push, acquiring a third building to expand its data center footprint as it seeks to lift training capacity to nearly 2 gigawatts of compute power.

The move underpins how the race to build ever more powerful AI models is increasingly being decided not just by algorithms, but by access to electricity, land, and specialized chips.

Musk disclosed the purchase on Tuesday in a post on X, saying xAI had bought a third facility called “MACROHARDRR,” without revealing its precise location. The name appears to be a deliberate play on Microsoft, a key backer of OpenAI. Earlier, The Information reported, citing property records and a person familiar with the project, that the building for the third supersized data center is planned outside Memphis, Tennessee, where xAI is already operating its flagship supercomputer cluster, Colossus.

Colossus, based in Memphis, has been billed by xAI as the largest AI supercomputer in the world. The system is central to Musk’s ambition of turning xAI into a credible challenger to OpenAI’s ChatGPT and Anthropic’s Claude. According to people familiar with the plans, xAI intends to expand Colossus to house at least one million graphics processing units, a scale that would put it among the most compute-dense AI installations globally.

The newly acquired warehouse is expected to begin conversion into a data center in 2026, The Information reported. It would complement both the existing Colossus cluster and a planned Colossus 2 facility. Crucially, both sites are located near a natural gas power plant that xAI is building in the area, alongside other power sources, highlighting how energy access has become one of the most significant bottlenecks in advanced AI development.

The near-2GW compute target is striking. For context, data center campuses operating at that level rival the power consumption of small cities. As AI models grow larger and more complex, training runs can take weeks and require an enormous, continuous energy supply. This has pushed leading AI companies to secure long-term power arrangements, invest directly in generation assets, and, in some cases, rethink where data centers are located.

Vertical integration appears to be a strategic choice for xAI. Unlike OpenAI and Anthropic, which rely heavily on cloud partners such as Microsoft and Amazon, Musk is pursuing a more self-contained model, combining proprietary data centers, power generation, and in-house model development. Supporters say this could give xAI greater control over costs and scaling, while reducing reliance on third-party infrastructure that may be constrained by competing demands.

The expansion also reflects the broader surge in capital spending across the AI sector. Tech companies have been pouring hundreds of billions of dollars into data centers, chips, and networking equipment to support a global frenzy for AI solutions. Nvidia’s dominance in AI chips has made access to GPUs a strategic priority, while power availability has emerged as a limiting factor even for the largest cloud providers.

However, the rapid buildout has not gone unchallenged. Environmental groups and local activists have raised concerns about the impact of large data centers, pointing to their heavy electricity consumption, water use for cooling, and increased strain on local grids. xAI’s proximity to a natural gas plant has intensified scrutiny, as critics question the climate implications of fossil-fuel-powered AI infrastructure at a time when governments and companies are pledging to cut emissions.

Musk has previously argued that AI progress requires massive compute and that reliable baseload power is essential to support it. Yet the tension between AI’s growth and sustainability is becoming harder to ignore, particularly as projects scale into the gigawatt range.

Strategically, the expansion signals Musk’s determination to keep xAI in the top tier of AI developers as competition intensifies. With OpenAI reportedly working on increasingly powerful models and Anthropic attracting significant enterprise adoption, the ability to train faster, larger, and more capable systems is becoming a decisive advantage.

xAI’s latest move reinforces a central reality of the sector as the AI arms race deepens: leadership is no longer just about who has the smartest models, but who can marshal the infrastructure, energy, and capital required to run them at unprecedented scale.

China has started distributing the second batch of 2026 crude oil import quotas to independent refiners, with at least one major facility in eastern China receiving its allocation, sources familiar with the matter revealed to Reuters on Tuesday.

More refiners, particularly in the Shandong province hub, expect notifications in the coming days, signaling Beijing’s calibrated approach to managing non-state imports amid stable overall volumes and a focus on demand flexibility.

The unnamed eastern refiner secured a combined 11 million metric tons (approximately 220,000 barrels per day) across the first two batches—equivalent to about 70% of its projected full-year quota. The initial batch, issued in late November, totaled around 8 million tons nationwide and was usable for cargoes arriving by year-end, providing a bridge amid quota exhaustion in late 2025.

This early release spurred a buying spree among teapots, boosting November crude imports to their highest daily level in 27 months at 12.2 million bpd, as refiners scrambled to utilize the fresh allowances before expiration. Independent refiners—commonly called “teapots,” clustered primarily in Shandong and processing ~20-25% of China’s crude (total imports ~11 million bpd)—anticipate their first two batches will similarly cover 70% of annual allowances, a shift from 2025 when early issuances comprised the full quota.

The reason for front-loading less volume remains unclear, though analysts speculate it allows greater flexibility in response to demand fluctuations, sanctions impacts, or inventory management. Beijing released an additional quota of about 10 million mt to qualified independent refineries for crude imports in 2025 in late November, and the final batch of 2025 quotas totaled more than 7.5 million tons—notably higher than the 6 million tons released in the same period the previous year.

China’s Ministry of Commerce has set the total 2026 quota for non-state firms at 257 million metric tons, unchanged from 2025 and reflecting cautious demand expectations in a slowing economy with 4.7% GDP growth in 2025. This cap covers independent refiners and some private giants, excluding state majors like Sinopec and PetroChina, which import freely.

For 2026, the first batch of fuel export quotas is stable year-on-year, with main recipients being state-owned oil companies Sinopec and CNPC, which received 13.76 million tons of allowances for refined fuels. Independent refiners often favor discounted sanctioned grades such as Russian, Iranian, and Venezuelan, due to price sensitivity, and the quota system, in place since the 1990s, regulates their imports to control refining capacity, fuel quality, and foreign exchange outflows.

Early 2026 allocations enable teapots to plan purchases amid narrowing Russian discounts of $2-4/bbl below Brent and heightened sanctions scrutiny. The November batch spurred a buying spree, boosting utilization rates to over 60% in December and drawing down bonded storage of Iranian crude. Second-batch issuances could sustain this momentum into Q1 2026, supporting sour barrel demand while Beijing manages overcapacity.

Independent refiners exhausted their previous quotas as early as October and were waiting for a new issuance at the end of the year. Shandong teapots, representing 80% of independent capacity (4 million bpd total), have faced headwinds: quota delays, fuel export curbs, and consolidation pressures. Recent revivals—three bankrupt plants resurrected under new owners—highlight resilience, but overall numbers have declined from peaks above 50 refiners.

Stable quotas signal muted growth expectations for 2026 refining throughput, prioritizing efficiency over expansion amid EV adoption and peak oil demand forecasts. Yet, front-loading 70% early provides certainty for discounted sour crude sourcing, potentially stabilizing Asian physical markets. China’s independent refiners boost crude buying after new import quotas, with the total allocation for 2026 set at 257 million tons, the same as a year ago.