The Real Reason Google Pays $920M a Month to Rent SpaceX's AI Computers

Illustration of a massive AI brain glowing through server racks in a large data center, with a space rocket flying above.
AI Summary

Google has entered into a $920 million monthly lease agreement with SpaceX for 110,000 GPU infrastructures, demonstrating that computing capacity has emerged as a top strategic resource in the tech industry.

The Hidden ‘Receipt’ Behind Your Smartphone Assistant Getting Smarter

Imagine you are the owner of the biggest and most successful bakery in town (Google). Every day, countless customers come to buy bread, but suddenly, they start ordering tens of thousands of elaborate, 10-tier wedding cakes all at once. You already have many great ovens in your shop, but you simply cannot handle the waterfall of orders. Eventually, you are forced to rent the massive ovens of another large bakery across the street (SpaceX), even though you often compete with them. You sign a contract to pay a huge monthly rent. It might hurt your pride a little, but it’s a much better choice than missing out on orders and closing your shop.

Simply put, Google, currently the world’s leading technology company and the absolute ruler of the internet, has made the same decision as this bakery owner. According to regulatory filings released on June 5, Google has agreed to rent AI computing infrastructure from Elon Musk’s SpaceX for a whopping $920 million (approximately 1.2 trillion won) per month through a multi-year infrastructure deal [[Google to pay SpaceX $920 mln monthly under new… The Daily Star](https://www.thedailystar.net/news/tech-startup/news/google-pay-spacex-920-mln-monthly-under-new-deal-4191751)]. Why is Google renting computers from a competitor?

Why It Matters

We usually think of global tech giants like Google, Apple, and Microsoft as having infinite resources and technology. However, with the explosive growth of AI technology, even the world’s largest internet companies have hit a massive physical limit. That limit is ‘Compute Capacity’—the power needed to train AI and make it work intelligently.

Google’s massive deal, which amounts to $11 billion a year, clearly shows what is becoming the most precious resource in the AI era. This amount is equivalent to the annual budget of many small nations. While nations and corporations competed fiercely for oil or territory during the industrial revolution, today’s tech companies are paying astronomical costs to secure a new virtual resource called ‘computing power.’ This contract symbolizes that AI computing capacity has become a top-tier core strategic resource, much like rocket launch capabilities or rare mineral resources, determining the very survival of a company [Google will pay SpaceX $920 million monthly to lease 110,000 Nvidia…].

This ‘war of money’ between corporations also has a major impact on the daily lives of ordinary people. Chatbots that answer your questions naturally like a human, medical AI that diagnoses plant diseases just by taking a photo with a smartphone, and amazing features that translate calls in real-time during overseas travel are not magic that fell from the sky. They are the physical fruits of massive data center computers crunching billions of numbers behind the scenes. The fact that companies like Google are pouring astronomical sums into infrastructure means that the quality of the AI services we use, our monthly subscription fees, and the overall pace of daily technological advancement are directly tied to these massive hardware rental agreements.

The Explainer

What exactly is Google renting for $920 million a month? The core of this deal is approximately 110,000 NVIDIA GPUs (Graphics Processing Units) and the supporting CPUs (Central Processing Units), massive amounts of memory, and other key computer equipment needed to run them constantly [[Google pays SpaceX $920M monthly for xAI compute Let’s Data Science](https://letsdatascience.com/news/google-pays-spacex-920m-monthly-for-xai-compute-e9933fc6)]. GPUs were originally created to render 3D graphics for video games, but their status has skyrocketed to become the world’s most precious and expensive semiconductors, acting as the brain of AI.

To use a simple analogy: if a CPU, the smart brain of a general computer, is a single ‘senior mathematics professor’ who solves complex calculus problems step-by-step over several days, a GPU is like a crowd of 10,000 ‘quick and smart middle schoolers’ who can solve thousands of simple, repetitive problems like addition and subtraction simultaneously.

Modern AI models must learn from hundreds of billions of sentences and millions of images simultaneously to understand human language and the world. This massive task requires the ability to process countless simple calculations at once rather than one highly complex logical step. This is why a crowd of trained middle schoolers (GPUs) is overwhelmingly more advantageous than a single math professor (CPU) for AI training. Through this contract, Google has essentially outsourced and leased 110,000 smart brains at once [Google will pay SpaceX $920 million monthly to lease 110,000 Nvidia…].

Originally, this massive computing facility was built by xAI, the AI company founded by Elon Musk, with the intent of creating their own powerful AI. Currently, xAI is part of SpaceX and jointly operates this vast infrastructure [[Google will pay SpaceX $920M per month for compute TechCrunch](https://techcrunch.com/2026/06/05/google-will-pay-spacex-920m-per-month-for-compute/)]. So why doesn’t Google build its own data centers? From Google’s perspective, they could not afford to wait the long time it takes to buy 110,000 chips, find land the size of several football fields, build massive structures, and bring in enough electricity to power a whole city. Because the pace of AI development is much faster than the speed of building physical structures, they chose the realistic path of paying expensive rent to become a tenant in a competitor’s massive ‘empty room’ (data center) that is already completed and ready to be plugged in.
This massive lease agreement is scheduled to last for 32 months, from October 2026 to June 2029 [[Google pays SpaceX $920M monthly for xAI compute Let’s Data Science](https://letsdatascience.com/news/google-pays-spacex-920m-monthly-for-xai-compute-e9933fc6)]. During this period, Google will save physical construction time while easily securing a powerful rocket engine to significantly expand its own cloud capabilities [Google to pay SpaceX nearly $1 billion monthly for cloud capacity].

Where We Stand

In fact, Google is not the first big-name tenant of SpaceX’s massive data center. Anthropic, a tech company often cited as one of the ‘big three’ alongside Google and OpenAI in the AI industry, had already signed a larger deal with SpaceX. Anthropic promised to pay $1.25 billion per month (approximately $15 billion annually) through 2029 to lease the entire available computing resources of the ‘Colossus 1’ data center located in Memphis, Tennessee [Google will pay SpaceX $920M per month for compute - MSN].

Google’s new contract scale is roughly half of the total computing power Anthropic uses exclusively at Colossus 1, in terms of monthly rent [Google will pay SpaceX $920M per month for compute]. So, which SpaceX data center will Google be using? SpaceX has not officially specified which facility’s computers Google will be using. However, CEO Elon Musk has previously hinted that ‘Colossus 2,’ a larger and newer data center he is ambitiously building, would not be rented out and would remain an exclusive research facility for xAI, further piking the industry’s curiosity [Google will pay SpaceX $920M per month for compute].

An interesting fact is that running these giant digital machines day and night requires an enormous amount of analog energy. Cooling the furnace-like heat emitted by 110,000 semiconductor chips running at full capacity 24/7 and providing a stable power supply is a massive civil engineering project and an extreme engineering challenge in itself. Surprisingly, some of the vast power currently operating this massive high-tech data center is coming from 27 giant methane gas turbine generators mounted on trailers [[Google to pay SpaceX $920M a month for compute capacity at xAI data centers Hacker News](https://news.ycombinator.com/item?id=48417490)]. This is a fascinating highlight of the paradoxical physical reality: AI, the world’s smartest and most sophisticated future-oriented technology, is dependent on the crude reality of burning fossil fuel methane gas on a construction site for its survival.

What’s Next

This astonishing and massive computer rental deal was revealed to the world just ahead of SpaceX’s planned large-scale IPO (Initial Public Offering). For SpaceX, this announcement is huge positive news. SpaceX has now perfectly demonstrated to global investors that it is far more than just an adventurous company launching rockets—it is a powerful platform company that has built world-class AI infrastructure and is generating massive fixed rental income from it. The trillions of won in astronomical monthly rent collected from top-tier clients like Anthropic and Google will serve as the most reliable and certain cash flow to maximize SpaceX’s enterprise value ahead of its listing [Google to pay SpaceX $920 million a month for compute capacity].

In the future, the size of AI models will grow even larger than we can imagine. For AI to speak more intelligently, draw more sophisticated pictures, and solve more complex math problems, it needs exponentially more text data and much stronger computing power to swallow and digest it all. The recent case of Anthropic, which moved into SpaceX first, clearly shows how powerful physical computing capacity can be. Before the infrastructure deal with SpaceX, Anthropic had to impose significant restrictions, such as usage limits, on its services because its own computing power was insufficient. However, on the very day this massive rental deal was officially announced, it was able to immediately and substantially raise the usage limits for its AI models, like opening wide the gates of a congested highway [Google will pay SpaceX $920M per month for compute]. In this way, a company securing computing power directly translates into the total amount of intelligence and convenience provided beyond the smartphone screens of ordinary consumers like us.

Google, the eternal king of internet search, is also expected to use the 110,000 computing power units secured through this lease agreement as a foothold to accelerate the strengthening of its own AI cloud (virtual server rental) capabilities, which had momentarily stalled [Google to pay SpaceX $920M a month in compute deal - AOL]. Therefore, the cutthroat competition among global IT giants that we will witness daily in the news has moved beyond a polite battle of wits over who can write more sophisticated and clever math code. It will now solidify into a rough physical battle over who can monopolize power plants producing vast amounts of electricity and secure one more box of NVIDIA semiconductor chips worth tens of millions of dollars each to build the massive physical skeletons covered in concrete first [SpaceX to rent AI capacity to Google for $920 million per month]. The scale of the invisible bill the world must pay to put a kinder and smarter AI friend in your smartphone is already easily exceeding trillions of won per month.

AI’s Take

MindTickleBytes AI Reporter’s View: We often imagine AI as a formless ‘cloud’ floating in the sky or the elegant magic of software composed of invisible 0s and 1s. However, this astronomical deal between Google and SpaceX has vividly proven the raw reality of the ‘physical infrastructure’ industry: behind that sophisticated curtain of magic lie 110,000 red-hot metal chips and 27 massive methane gas generators burning fossil fuels with a roar.

The fact that even Google, the absolute power that has reigned at the top of the IT ecosystem for the past 20 years by monopolizing human knowledge and vast amounts of search data, is willing to bow its head to yesterday’s competitor and pay a massive cost of over 1 trillion won per month for its immediate survival is highly significant. This isn’t just about buying a few more semiconductor chips. It clearly states that the sheer ‘computing power’ of machines standing firmly on the ground and running with a roar is the heart of 21st-century technological hegemony and the ultimate, irreplaceable power that cannot be easily bought even with money. The era of software eating the world has passed, and a completely new era has opened where those who preoccupy the massive and heavy physical hardware to run that software will rule the world.

References

  1. [Google will pay SpaceX $920M per month for compute TechCrunch](https://techcrunch.com/2026/06/05/google-will-pay-spacex-920m-per-month-for-compute/)
  2. Google to pay SpaceX 920 million per month for AI compute…
  3. Google will pay SpaceX $920 million monthly to lease 110,000 Nvidia…
  4. [Google to pay SpaceX $920 mln monthly under new… The Daily Star](https://www.thedailystar.net/news/tech-startup/news/google-pay-spacex-920-mln-monthly-under-new-deal-4191751)
  5. SpaceX Locks Google Into A $920 Million-Per-Month Compute Deal…
  6. Google to pay SpaceX $920 million a month for compute capacity at xAI data centers
  7. [Google pays SpaceX $920M monthly for xAI compute Let’s Data Science](https://letsdatascience.com/news/google-pays-spacex-920m-monthly-for-xai-compute-e9933fc6)
  8. [Google to pay SpaceX $920M a month for compute capacity at xAI data centers Hacker News](https://news.ycombinator.com/item?id=48417490)
  9. Google will pay SpaceX $920M per month for compute
  10. Google to pay SpaceX nearly $1 billion monthly for cloud capacity
  11. Google to pay SpaceX $920 million a month for compute capacity
  12. Google to pay SpaceX $920M a month in compute deal - AOL
  13. Google will pay SpaceX $920M per month for compute - MSN
  14. SpaceX to rent AI capacity to Google for $920 million per month
Test Your Understanding
Q1. Approximately how much is the monthly amount Google agreed to pay SpaceX for computing power?
  • $100 million
  • $920 million
  • $1.5 billion
Google agreed to pay $920 million monthly in exchange for the computing capacity of xAI data centers owned by SpaceX.
Q2. Which key component, numbering approximately 110,000, is included in the data center infrastructure Google is renting?
  • CPU (Central Processing Unit)
  • GPU (Graphics Processing Unit)
  • SSD (Solid State Drive)
Through this deal, Google secured access to approximately 110,000 NVIDIA GPUs, which serve as the core 'brain' for AI computations.
Q3. Which AI company contracted to lease the entire 'Colossus 1' data center from SpaceX before Google?
  • OpenAI
  • Meta
  • Anthropic
Anthropic contracted with SpaceX before Google to lease the Colossus 1 data center located in Memphis, Tennessee, for $1.25 billion per month.
The Real Reason Google Pays...
0:00