xAI in advanced talks to lease up to gigawatts of data‑center capacity in Saudi Arabia, tapping Humain and a 200 MW backup deal to boost Grok and Colossus scale.
Since its founding in March 2023, Elon Musk’s AI company, xAI, has pursued an aggressive strategy to build compute capacity fast — launching the Colossus supercomputer in Memphis in December 2024 and unveiling the Grok chatbot to scale against incumbents like OpenAI’s ChatGPT and Anthropic’s Claude 1.
Colossus, initially powered by 150 MW through over 100,000 Nvidia GPUs—and later expanded—was built in under 130 days. But domestic infrastructure alone could not keep pace with xAI’s growth trajectory, prompting exploration of international partnerships.
In mid‑July 2025, Bloomberg reported that xAI entered advanced negotiations to lease data‑center capacity in Saudi Arabia. Two options are on the table: a long‑term gigawatt-scale deal through Humain, and a near-term 200 MW facility under construction by an unnamed local entity 2.
Humain, officially launched on May 13, 2025 under the Public Investment Fund (PIF) and chaired by Crown Prince Mohammed bin Salman, aims to position Saudi Arabia as a global AI node 3.
Its plan targets 1.9 GW of data‑center-grade infrastructure by 2030, expanding to 6.6 GW by 2034, with total investment estimated at $77 billion 4.
Humain has already secured partnerships valued at tens of billions:
These partnerships signal a coordinated surge of AI infrastructure capital backed by sovereign resources.
This unnamed partner is building a 200 MW facility slated for near‑term rollout—likely within 12 to 18 months. It offers immediate compute availability for training model cycles and Grok-related workloads, serving as a bridge while Humain’s infrastructure develops 6.
If Humain delivers its gigawatt-scale buildouts, xAI could lease several gigawatts of compute, aligning precisely with the projected needs of next-gen Grok models. But most of the capacity is still in planning or early construction, likely delivering beyond 2026 7.
Leasing scale gives xAI flexibility without requiring high CapEx and allows modular deployment across sites.
In July 2025, xAI secured $5 billion in debt (through bonds and term loans) and a further $5 billion in strategic equity, as arranged by Morgan Stanley, branded as one of the most oversubscribed debt offerings in recent memory. This injection was earmarked for global data-center infrastructure and Grok model development 8.
From the equity side, SpaceX invested approximately $2 billion, nearly half of the equity tranche. Meanwhile, Kingdom Holding Company (linked to PIF) previously funneled about $800 million into xAI across earlier rounds 9.
Total capital xAI has raised to date exceeds $12 billion, including prior rounds in 2024 from investors like Andreessen Horowitz, Sequoia, BlackRock, and Fidelity. Company funding now approaches $17–18 billion, and it’s exploring another $10–20 billion round targeting a valuation between $170–200 billion, though Elon Musk has stated that xAI isn’t actively seeking capital now 10.
The raised funds support both Colossus expansion in Memphis and global leasing plans—including the Saudi discussions.
Electricity rates in Saudi Arabia fluctuate between $0.02 and $0.05 per kWh, representing a 20–40% cost advantage versus typical U.S./European grid rates. Over a 200 MW site, that could save $25–40 million per year in operational expense, especially significant over multiple years of training costs 11.
By leasing, xAI avoids major upfront CapEx. It gains compute capacity without long-term asset ownership, yet retains access to growing global infrastructure as Humain scales out.
Though not involving political commentary, it is neutral to note xAI’s alignment with PIF’s sovereign-backed infrastructure development strategy, offering stable compute availability and integration into broader cloud and AI service networks.
Even with its appealing upside, the arrangement faces several neutral concerns that are typical in large infrastructure agreements:
Construction and delivery risk Humain’s multi-gigawatt buildout is planned over years—with build schedule risk, permitting hurdles, and logistics uncertainty.
Data handling and governance Hosting compute overseas raises questions about data residency, access, and compliance regimes. While no criticism is intended, these are standard issues for international compute deployment.
Export-control and hardware supply While chip export policies on advanced GPUs may ease, supply chains remain subject to licensing and geopolitical frameworks that could slow delivery.
Financial exposure Although leasing mitigates direct CapEx, rental commitments at scale could reach hundreds of millions per year—requiring durable revenue models.
ESG and environmental footprint Diesel or gas-based power generation and carbon intensity of remote compute centers are noted concerns; neutral awareness is prudent, even if no regulatory action is identified publicly.
These are operationally typical points of attention for multi‑year infrastructure deals, not political critique.
The global AI infrastructure market is rapidly scaling: major tech firms (Meta, Google, CoreWeave, Anthropic) are each committing billions to data centers. Estimates place industry-wide AI infrastructure spending at over $320 billion in 2025 alone 14.
Saudi Arabia’s strategy, through Humain and sovereign partnerships, is intended to attract global AI players. For xAI, tapping in creates a diversification route beyond U.S.–based compute, without signs of any conflict or political motive.
Explore the essential 2025 toolkit for freelancers—tailored by role and income level. Boost productivity, automate workflows, and grow smarter.
Read more →
