Our Take
xAI is losing money faster than it can earn it, and the filing shows no path to profitability — only plans to spend more.
Why it matters
This is the first public audit of Musk's AI financials. For investors betting on SpaceX's $1.75 trillion IPO valuation, the xAI unit's burn rate and unproven monetization are material risks that SEC disclosure now locks into the record.
Do this week
Infrastructure leaders: audit your own AI capex-to-revenue ratio against xAI's 10:1 loss rate to validate whether your spending targets are realistic.
SpaceX's IPO filing reveals xAI's scale and spending
xAI lost $6.4 billion on $3.2 billion in revenue in 2025, more than four times the prior year's $1.56 billion loss (per the SpaceX SEC filing). Capital expenditures for AI infrastructure climbed to $7.7 billion in Q1 2026 alone, annualizing to roughly $30.8 billion and more than doubling year-over-year spending.
The revenue increase came primarily from "AI solutions and infrastructure revenue" totaling $465 million, which includes $365 million from X and Grok subscriptions and $88 million from data licensing. Advertising contributed an additional $116 million. Those figures matter because they show where money actually comes from: subscriptions and licensing, not enterprise inference or API services.
Grok itself reaches 117 million monthly active users for AI features out of 550 million total across X and Grok combined. That is one-fifth of the ecosystem actively engaging with Grok AI. The filing commits xAI to scaling Grok to "multiple trillions of parameters," described as a "step change in reasoning in depth and overall intelligence." Two data centers, Colossus and Colossus II, came online in 122 and 91 days respectively and collectively provide about 1 gigawatt of compute power for training and inference.
The SpaceX filing also introduces a longer-term vision: orbital AI compute satellites starting as early as 2028. The filing states "the future of AI will be determined by control of the physical stack," positioning vertical integration from satellites to data centers to software as a competitive moat.
The math doesn't work yet, and capex is accelerating
xAI is spending $10 to earn $1. Competitor Anthropic, by contrast, is on track for 130% revenue growth to $10.9 billion in Q2 and is approaching operating profit (per company reporting). OpenAI and Anthropic are both eyeing public debuts in 2026, but neither has disclosed burn rates of this magnitude tied to a single model.
The filing shows capex acceleration, not moderation. Q1 2026 capex alone exceeded the entire 2025 capex run rate, suggesting Musk intends to build more compute capacity before demand fully materializes. This is a bet that subscription revenue, licensing, and advertising will eventually grow into the spending, or that orbital compute will substantially reduce the cost structure by 2028 or later.
For SpaceX shareholders, xAI's losses now live inside the audited financials of a company valued at $1.75 trillion. That means scrutiny of AI capex will carry the weight of SEC disclosure going forward, not internal estimates.
Plan for sustained compute intensity
If xAI's spending trajectory becomes standard for frontier model training, the cost floor for competitive AI development is rising sharply. Teams evaluating in-house model training versus API consumption should model capex at $30B+ annualized for frontier-grade systems. That calculus shifts the ROI bar for build-versus-buy decisions.
Watch the Q3 and Q4 2026 filings for capex trends and Grok user growth rates. If capex keeps accelerating while user adoption flatlines, the orbital compute timeline becomes the primary investor thesis, not current revenue.