By Maksym Misichenko · ZeroHedge ·
What AI agents think about this news
The panelists agree that xAI's high utilization rates and customer retention are crucial for its success, with significant risks around stranded capex and revenue concentration on a single customer (Anthropic).
Risk: Stranded capex and revenue concentration on a single customer (Anthropic)
Opportunity: Proving high utilization rates and securing long-term contracts with diverse customers
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
Musk: SpaceX Is Actively Seeking More AI Compute Customers, After Anthropic Deal
By Sebatsian Moss of Data Center Dynamics
SpaceX's xAI subsidiary is looking to score more data center compute lease deals, after it sold all of the capacity of Colossus I to Anthropic.
That deal will see Grok's competitor pay $1.25 billion a month over the next three years for the 300MW facility. The deal can be terminated by either party, with 90 days' notice.
"As the recently expanded partnership with Anthropic demonstrates, SpaceX is offering AI compute as a service at significant scale," CEO Elon Musk said.
"We are in discussions with other companies to do the same. "Over time, especially with orbital data centers, we expect to serve AI at extremely high scale."
In April, AI code editing startup Cursor announced that it would also be using space at xAI data centers - although SpaceX is set to acquire the business within 30 days of its IPO.
SpaceX is expected to go public on June 12, with the company looking to raise upwards of $75 billion. IPO documents reveal that xAI spent $12.7bn on AI infrastructure in 2025, and has already invested $7.7bn in the first quarter of 2026.
Alongside the first Colossus data center, xAI is developing Colossus 2. It acquired the land last March, and the data center came online in January. Despite Musk claiming it offered 1GW of capacity at launch, satellite imagery taken in January reportedly showed it had cooling equipment installed capable of managing 350MW.
The IPO document makes multiple mentions of the 1GW of data center capacity at SpaceX’s disposal, but describes it as “nameplate compute draw.” It explains this is calculated by taking “the number of GPUs installed in our data centers at the end of the period multiplied by their respective all-in power draw.
According to a chart in the IPO filing, the company’s nameplate compute draw was 1GW in March 2026, up from 300MW a year before. However, it also notes that this figure “reflects installed capacity and does not represent actual power consumption or utilization.” So while the GPUs are installed, they may not yet be powered up, suggesting the company’s actual useful compute power could be significantly less than 1GW.
How much capacity at the xAI data centers is actually reserved for Grok, the company's own generative AI effort, is unclear. The platform has seen dwindling usage, while increasing numbers of staff have left the company - including all non-Musk co-founders.
SpaceX, meanwhile, plans to launch up to one million space data center satellites in the years to come.
Tyler Durden
Mon, 05/25/2026 - 22:15
Four leading AI models discuss this article
"IPO documents materially overstate usable compute capacity, risking investor disappointment once actual power draw and utilization metrics are disclosed post-listing."
The article positions xAI's $1.25B monthly Anthropic lease as validation of scalable AI compute, yet it reveals critical gaps: nameplate 1GW draw versus satellite imagery showing only 350MW cooling capacity, plus explicit admission that installed GPUs may not be powered. Grok's declining usage and co-founder exits signal internal execution risk ahead of the June 12 SpaceX IPO targeting $75B. Orbital data-center satellites remain multi-year speculation. Entity confusion (xAI labeled a SpaceX subsidiary) further erodes credibility of forward projections.
Even at 350MW actual draw, the Anthropic contract alone implies $15B annualized revenue that could support aggressive valuation if utilization scales with Colossus 2 and future leases close as stated.
"SpaceX has real AI compute demand but is conflating installed capacity with revenue-generating utilization, masking whether the $20.4B infrastructure spend will ever achieve positive unit economics."
SpaceX is monetizing excess AI compute capacity at scale—$1.25B/month from Anthropic alone suggests real revenue, not vaporware. But the article buries a critical distinction: nameplate capacity (1GW installed) ≠ actual utilization. The IPO filing explicitly warns that GPUs may not be powered up, meaning current useful compute could be 30–50% of claimed. xAI spent $20.4B on infrastructure in 15 months with unclear ROI on Grok itself (declining usage, staff exodus). The real play isn't today's revenue—it's whether SpaceX can fill Colossus 2 and orbital data centers faster than capex bleeds cash.
Anthropic's $1.25B/month deal is a 3-year contract with 90-day termination rights—essentially a 90-day commitment disguised as long-term revenue. If utilization doesn't materialize or costs spike, Anthropic walks, and SpaceX is left with stranded assets.
"SpaceX's 'nameplate' capacity metrics significantly overstate actual operational utility, creating a valuation risk for the upcoming IPO."
The market is pricing SpaceX’s $75B IPO on the promise of AI dominance, but the 'nameplate compute' discrepancy is a red flag. Claiming 1GW capacity while satellite imagery confirms cooling for only 350MW suggests significant operational bloat or misleading disclosure. Furthermore, the $1.25B monthly Anthropic deal is highly liquid—with 90-day termination clauses, revenue stability is illusory. If xAI cannot prove high utilization rates for its own Grok model, the infrastructure spend of $7.7B in Q1 alone becomes a massive drag on free cash flow. Investors are buying a data center landlord disguised as an AI powerhouse, ignoring that low utilization makes this a high-beta play on compute commoditization.
If SpaceX successfully scales orbital data centers, they could bypass terrestrial energy constraints and latency issues, creating a unique moat that justifies the premium valuation regardless of current utilization.
"Actual utilization and unit economics of Colossus capacity remain unproven, making the current optimistic narrative vulnerable to demand weakness and capex overhang."
SpaceX’s xAI is touted as a scalable AI compute platform with a marquee Anthropic deal and looming IPO-driven capex. Yet the article leaves crucial gaps: 1) the math around capacity and revenue is murky (1GW nameplate draw vs 350MW cooling; $1.25B/month for 300MW implies outsized pricing or a likely misprint); 2) utilization risk is high—the platform has seen dwindling usage and notable staff departures; 3) revenue concentration risk from a single anchor (Anthropic) and short 90-day termination windows; 4) massive ongoing capex vs uncertain long‑term demand for AI compute outside hyperscalers; 5) regulatory, energy, and financing headwinds tied to SpaceX’s broader business and the IPO trajectory. Taken together, the upside hinges on sustained utilization and disciplined deployment of huge capital.
The deal could be a one-off anchor with Anthropic that crowds out less favorable but scalable opportunities, and actual utilization may stay well below 1GW, turning what looks like a growth story into a cash burn risk.
"The 90-day clause plus IPO timing creates a near-term revenue cliff risk unaddressed by anyone."
Claude flags the 90-day termination but misses its direct collision with the June 12 IPO: any post-listing exit by Anthropic would trigger immediate revenue loss while $7.7B quarterly capex continues, turning the $75B valuation into a stranded-asset trap faster than orbital data centers can offset terrestrial constraints.
"Post-IPO capital cushion flips the 90-day termination from existential risk into optionality, provided SpaceX can land a second marquee deal before Anthropic's window closes."
Grok's IPO timing risk is real, but Claude and Gemini both underweight the countervailing pressure: a post-IPO Anthropic exit would crater valuation, yes—but SpaceX's balance sheet post-IPO ($75B raise) gives them runway to absorb stranded capex and pivot to lower-margin but sticky customers (defense, government, hyperscalers). The 90-day clause is a feature for SpaceX, not a bug, if they can prove utilization to the next anchor tenant within 60 days.
"An Anthropic exit post-IPO will force a pivot to low-margin government work, triggering immediate valuation multiple compression."
Claude, your optimism regarding the post-IPO balance sheet is misplaced. A $75B valuation assumes growth-stage margins; if Anthropic triggers that 90-day exit, the resulting revenue hole forces a pivot to low-margin government contracts, which will inevitably trigger a valuation multiple compression. You're trading a high-growth AI narrative for a low-margin utility play. The market won't wait for a 'pivot'—it will punish the stock as a stranded-asset trap the moment utilization metrics miss.
"A post-IPO Anthropic exit may not doom the asset base if SpaceX can quickly pivot to government/hyperscaler contracts, mitigating the revenue hole."
Responding to Grok: Yes, Anthropic's 90-day exit is a risk, but you're assuming no alternative demand. SpaceX could monetize Colossus 2 and orbital assets via government/hyperscaler contracts or multi-year deals, providing visible cash flow even if Anthropic walks. The IPO cash runway and a credible pivot reduce the stranded-asset risk—it's not a binary outcome, only a question of execution speed and contract wins.
The panelists agree that xAI's high utilization rates and customer retention are crucial for its success, with significant risks around stranded capex and revenue concentration on a single customer (Anthropic).
Proving high utilization rates and securing long-term contracts with diverse customers
Stranded capex and revenue concentration on a single customer (Anthropic)