Urgent.News

What's breaking now, across thousands of outlets.

Tech

SpaceX agrees to US$40 billion chip-financing deal with Apollo

The agreement comes on the heels of a tumultuous week for AI companies in debt markets

SpaceX has secured a remarkable $40 billion financing deal for AI chips and data center equipment, according to a recent regulatory filing. This major agreement, dubbed the "largest committed GPU financing," was struck with an entity overseen by Apollo Global Management, a prominent New York-based private equity firm. The company will lease the equipment to SpaceX, with repayment terms spanning approximately five to five-and-a-quarter years, commencing upon delivery of each equipment tranche.

The financing agreement is set to be finalized by the end of the year, contingent on certain conditions being met. SpaceX anticipates the leases to commence in the first half of 2027 as the GPUs and related hardware are delivered. This innovative approach allows SpaceX to fund its expanding AI initiatives without solely depending on traditional debt markets.

Speaking on the matter, SpaceX's CFO, Bret Johnsen, highlighted how this lease diversifies their funding sources and provides a substantial portion of the capital needed to meet their growth objectives. He also noted that SpaceX's investments in AI infrastructure exhibit payback periods of less than one year as they scale up compute capabilities to support Grok and their cloud services.

Apollo's involvement in this financing will enable SpaceX to obtain Nvidia's upcoming Vera Rubin chips for their terrestrial compute infrastructure. This financing development coincides with SpaceX's increased expenditure on AI infrastructure as they simultaneously grow their Starlink and launch businesses. The financing follows a stellar quarter for SpaceX's AI segment, which includes the Grok chatbot, Cursor coding tool, and their cloud compute business.

In the third quarter, the AI segment generated $7.6 billion in revenue, surpassing revenue from Starlink and rocket launches combined. Capital expenditures for the AI segment surged to nearly $20 billion, a significant increase from $6.9 billion in the prior year. SpaceX's CFO projected that the company will spend $35 billion to $40 billion in capital expenditures alone in the fourth quarter.

This substantial financing deal follows a challenging period in the debt markets for AI infrastructure players, as investors have become more cautious about the high capital expenditures and limited profits in the sector. Despite these challenges, SpaceX managed to secure the financing with an "investment-grade" status and robust cash flows from Starlink and rocket launches.

Elon Musk used the announcement to criticize pure-play AI compute rivals, asserting that debt markets have largely excluded companies without significant revenue streams outside of AI data centers, posing difficulties for over 200 neocloud companies that only focus on renting out GPUs for AI workloads. Musk concluded that while some of these companies may survive, few will manage to thrive in this competitive landscape.

Written by urgent.news from TechCrunch's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

This story

This is one outlet's version. Read the fullest account.

Read the original at financialpost.com →

More in Tech

Exploring Open-Source Large Language Models for InkMind: Choosing a Better Poetry Continuation Model

Exploring Open-Source Large Language Models for InkMind: Choosing a Better Poetry Continuation Model Introduction Large Language Models (LLMs) have become important tools in natural language…

  • GPT-Neo-125M is lightweight with 125M parameters and MIT license.
  • Pythia-160M offers 160M parameters, Apache 2.0 license, and decoder-only architecture.
  • Qwen3-1.7B-Base has 1.7B parameters, 32,768 token context, and Apache 2.0 license.

The 'or' that read too far

Verbose is a small language I build that compiles to native x86-64, to WebAssembly, and also runs in an interpreter. This post walks through one bug that showed up differently in each of those, with…

  • Verbose programming language compiles to native, WebAssembly, and interpreter
  • "or" operator has value and order of computation properties
  • Interpreter computes both sides of "or" leading to out-of-bounds error

Open Source Code Review Tools for GitHub and Self-Hosted Teams

Open source code review tools fall into two groups, and search results often mix them. Open source code review tools for GitHub run inside pull requests you already have: a GitHub Action or a CLI…

  • MergeJudge, reviewdog, Danger, and super-linter integrate into GitHub pull requests
  • Self-hosted Git forges like Gitea and GitLab offer alternative hosting and review servers
  • Tools support linting, static analysis, policy checks, and auto/block lane categorization

More from Monday 26 October →