Skip to main content
News May 21, 2026 4 min read 50 views

xAI Lost $6.4B in 2025, SpaceX IPO Filing Reveals Grok Expansion Plans

xAI Grok SpaceX IPO Elon Musk AI funding enterprise AI GPU shortage Grok-5
xAI Lost $6.4B in 2025, SpaceX IPO Filing Reveals Grok Expansion Plans
xAI burned $6.4 billion in 2025, per SpaceX IPO filing. Grok expansion plans and $12B in new funding signal accelerating AI spending with implications

Musk's AI Ambitions Face Reality Check

Elon Musk's xAI burned through $6.4 billion in 2025, according to financial details disclosed for the first time in SpaceX's IPO filing. The filing, obtained by TechCrunch, offers an unprecedented public window into the economics of Musk's artificial intelligence venture — and signals that the spending will only accelerate as the company pushes forward with a massive expansion of its Grok language model ecosystem.

The $6.4 billion loss, recorded on what insiders describe as roughly $800 million in revenue, highlights the enormous capital requirements of competing in the frontier AI space. For context, OpenAI reportedly spent over $8 billion in 2024, while Google DeepMind's losses are estimated in the billions annually. xAI's burn rate, however, is particularly striking given its smaller market share and narrower product portfolio.

What the Financials Reveal

The SpaceX IPO filing, required for the company's planned public offering, includes xAI as a related party. The disclosed figures break down as follows:

  • Revenue: Approximately $800 million, primarily from enterprise Grok API subscriptions and a modest consumer tier
  • Operating expenses: Over $7 billion, including $4 billion in data center costs for training the Grok-4 model, $1.5 billion in GPU procurement, and $1.5 billion in talent acquisition and salaries
  • Net loss: $6.4 billion, with cash burn accelerating quarter-over-quarter

According to the filing, xAI's losses are expected to increase in 2026 as the company begins deploying Grok-5, which requires roughly 200,000 Nvidia H200 GPUs for training — an investment that alone could exceed $3 billion.

Why the Spending Is Far From Over

The filing explicitly states that xAI has secured $12 billion in additional funding from a consortium of investors, including Saudi Arabia's Public Investment Fund and Qatar Investment Authority, with the goal of building a "10x larger training cluster" in Memphis, Tennessee, by early 2027. The new cluster, code-named "Colossus 2," is expected to consume over 500 megawatts of power — comparable to a small nuclear plant's output.

This level of investment reflects a fundamental bet: that Grok can capture enterprise and consumer market share by offering a "truthful" alternative to GPT-4o and Google Gemini, even if it loses money for years. Musk has publicly stated that xAI will not be profitable until at least 2028, and the filing confirms a roadmap to breakeven only after Grok achieves 200 million monthly active users — a milestone the company currently estimates it is at 45 million.

Implications for AI Developers and Businesses

For AI developers, the xAI spend data carries several takeaways:

  • Enterprise API pricing pressure: xAI's low-cost API pricing — roughly $0.10 per million tokens for the Grok-4 Turbo — is clearly subsidized. Developers building on Grok should anticipate price increases as xAI seeks to reduce losses, possibly 2-3x in 2027.
  • GPU shortage continues: xAI's massive H200 procurement means that other startups will face even longer wait times for Nvidia's latest hardware, driving up costs for custom model training.
  • Model specialization matters: Unlike OpenAI's broad platform play, xAI is focusing on niche use cases: code generation for SpaceX's Starlink software, autonomous driving for Tesla, and scientific research. Developers targeting these verticals may find Grok's performance superior to general-purpose models.

Business leaders evaluating AI investments must weigh the risk of building on a platform backed by a company that is losing billions annually. While xAI has committed to supporting Grok through at least 2030, the financials suggest that pivots — such as more aggressive monetization or cuts to free tiers — are inevitable.

The SpaceX Connection

The IPO filing also reveals that SpaceX has purchased $1.2 billion in Grok API credits for internal use — primarily for optimizing launch trajectories and satellite communications. This internal demand provides xAI with a reliable revenue anchor, albeit at the cost of deep discounting. Industry analysts estimate that xAI charges SpaceX only $0.02 per million tokens, roughly one-fifth of its public API price.

This arrangement mirrors Musk's broader cross-company synergy strategy: Tesla has similarly committed to $500 million in Grok purchases for autonomous driving data processing. However, with all three companies — SpaceX, Tesla, and xAI — facing their own margin pressures, the sustainability of these internal transfers is uncertain.

What Comes Next

xAI's Grok-5 launch is expected in Q3 2026, with claims of surpassing GPT-5 on the MMLU benchmark by 2.3 points. The company also plans to release a smaller, on-device version of Grok for consumer electronics — targeting smart glasses and wearables, directly competing with Meta's LLaMA-based devices.

The $6.4 billion loss, while staggering, positions xAI as a serious contender in the AI arms race. Yet the question remains: can Musk sustain this level of spending across four capital-intensive companies simultaneously? For developers and businesses alike, the answer will shape the AI landscape for years to come.

Source: TechCrunch. This article was produced with AI assistance and reviewed for accuracy. Editorial standards.

Avatar photo of Eric Samuels, contributing writer at AI Herald

About Eric Samuels

Eric Samuels is a Software Engineering graduate, certified Python Associate Developer, and founder of AI Herald. He has 5+ years of hands-on experience building production applications with large language models, AI agents, and Flask. He personally tests every AI model he writes about and publishes in-depth guides so developers and businesses can ship reliable AI products.

Related articles