$100 Billion Spent on Data Centres in Q2 2025
AI Media Is the Next Layer
The smart ones are preparing to license. In the second quarter of 2025, the world’s leading tech firms — Amazon, Microsoft, Google, Meta, Apple, Nvidia, and Tesla — spent over $100 billion USD on infrastructure.
The vast majority of this investment went into AI data centres. Not general-purpose cloud. Not storage. These are custom-built for inference, rendering, simulation, and generative workflows.
They’re not building for us. They’re building for the machines that will serve us.
https://youtu.be/XITavRKbXF4?si=qSfo1vPUgQbBc5CF
BLOOMBERG - The Mysterious Rise of China’s Desert AI Hubs - 1 Aug 2025 #China #AI #Technology
A Bloomberg investigation found China is building a small city of AI data centers in a remote desert and looking to buy 115,000 of Nvidia’s best chips to power them despite a US export ban.
Q2 2025: AI Data Centre Investment by the Big 7
Amazon is preparing AWS for real-time AI video and automated retail systems.
Microsoft is expanding Azure to host OpenAI workloads for every enterprise vertical.
Google is scaling Gemini across YouTube, Search, and Docs. Expect synthetic video soon.
Meta is investing in a foundation for agent-driven storytelling and immersive generative media.
Apple is enabling offloaded model runs through Private Compute — tied into personal hardware.
Nvidia is deploying end-to-end AI factories that run their own outputs in closed loops.
Tesla is training robotics and autonomous agents, fed by simulation-scale infrastructure.
Together, they spent more than any other single quarter in history. And this isn’t peak spend.
Will next quarter be bigger?
Yes. Everything points towards Q3 will top this. Nvidia’s forward guidance includes additional spend in China and the EU. Microsoft’s new Azure regions are coming online. Google is prepping multi-modal integrations across services. Tesla is building out Dojo 2. The datacentre arms race is accelerating.
Amazon is preparing AWS for real-time AI video and automated retail systems. Microsoft is expanding Azure to host OpenAI workloads for every enterprise vertical. Google is scaling Gemini across YouTube, Search, and Docs. Expect synthetic video soon. Meta is investing in a foundation for agent-driven storytelling and immersive generative media. Apple is enabling offloaded model runs through Private Compute — tied into personal hardware. Nvidia is deploying end-to-end AI factories that run their own outputs in closed loops. Tesla is training robotics and autonomous agents, fed by simulation-scale infrastructure. Together, they spent more than any other single quarter in history. And this isn’t peak spend.
Here’s a table summarizing key estimated Q2 2025 AI data centre and capital expenditures by the "Big 7" tech companies, reflecting their historic AI investments and strategic focus:
| Company | Q2 2025 AI/CapEx Spend | Key AI Investment Focus |
|---|---|---|
| Amazon | $31.4B^1 | AWS: Real-time AI video, AgentCore, massive infrastructure; $100B+ full-year capex guidance^1^3 |
| Microsoft | $22.6B^4 | Azure: OpenAI workloads, verticalized AI, deep enterprise integration; $80B FY 2025 target^4^5 |
| $22.4B^5, $85B FY25 guidance^2 | Gemini: AI across YouTube, Search, Docs; synthetic video; $85B FY 2025 capex^2 | |
| Meta | $15–16B (Q2 est.)^2 | Generative media, agent-driven storytelling, foundation models; $60–65B FY 2025 forecast^2 |
| Apple | ~$7–8B (Q2 est., slower ramp)^2 | Private Compute, local/offload hybrid AI, personal hardware integration |
| Nvidia | $30B revenue (CapEx not split)^6 | End-to-end AI factories, sovereign clouds, GPU supercenters in EU/US; heavy manufacturing buildout^7 |
| Tesla | $2–3B (Q2 est.)^2 | Robotics, autonomous agents, simulation infra; major new Gigafactories, battery plants^7 |
- Group spend: Combined Q2 2025 CapEx for these seven exceeded $110B, making it the highest single-quarter AI data centre investment in history. Yearly group capex guidance is north of $320B, and Q2’s pace is not yet the peak^2^5.
- Trends: Investments are concentrated in cloud AI workloads (AWS, Azure, Google Cloud), GPU supercenters (Nvidia), simulation for robotics (Tesla), and foundational consumer/enterprise AI infrastructure (Meta, Apple).
This unprecedented scale anchors the ongoing "AI arms race"—a level of investment never seen in tech before, as each company races to dominate the next phase of generative and agentic AI^2^1.
So who pays for this?
You do. We all do. In cash, in data, in content.
Increased cloud bills — from individual storage to enterprise hosting.
New subscriptions — AI assistants, creative tools, productivity upgrades, premium features.
Higher device costs — phones, laptops, and cameras will quietly increase in price.
Behavioural capture — more of your actions will be interpreted as training fuel.
And eventually, through media. Because the content these systems deliver will come from somewhere. If you're not licensing in, you're being used to train it.
What creators and rights holders need to understand
This isn’t about uploading to the cloud or even being streamed. This is about structuring your work to fit the pipelines these systems require.
That means:
Titles that can be parsed
Characters and scenes that can be remixed
Legal rights that can be traced
Files that carry provenance and usage data
What we do at The AI Film Company
We create structured, machine-readable, assignable IP for AI-native systems. Our Titles are authored into Sandboxes. Each Sandbox contains Components performance, voice, environment, behavior, design. Every asset is wrapped in a .dave file for validation and rights control.
If you wish to know more about creating structed licensed entertainment IP contact us.
