Lately I have been thinking
What is the Business model of AI ?
Where does the money come from and where does it outflow ?
Then I saw the following headline from Gartner.
According to Gartner’s analysis, hardware will claim a staggering 80% of all gen AI spending in 2025. The forecast shows:
* Devices will account for $398.3 billion (99.5% growth)
* Servers will reach $180.6 billion (33.1% growth)
* Software spending follows at just $37.2 billion (93.9% growth)
* Services will total $27.8 billion (162.6% growth)
The first point about devices makes sense. since the last 1-2 years, there is a huge push to add NPUs onto every new device being sold on the market. Currently there is very little mainstream consumer hardware that does not classify itself as AI device. AI seems to be a marketing trick to make us buy brand new expensive computers. With windows 10 reaching end of life at the end of 2025, a lot of AI devices might end up coming into the hands of consumers. The software support for the AI stuff is not yet in place or most of these devices are far too under powered to run AI tasks.
Next I wanted to see where the money for the servers and overall CAPEX for AI is coming from. Data for the next piece of analysis comes from: AI capex 2025
Capital Expenditures for AI (CAPEX)#
| Company | 2024 CAPEX | 2025 CAPEX | Total 2024-2025 |
|---|---|---|---|
| Microsoft | $55.7B | $80B | $135.7B |
| Amazon | $83B | $105B | $188B |
| $52.5B | $85B | $127.5B | |
| Meta | $37.3B | $60-65B | $100-102B |
| Tesla | $11.3B | $11B | $22.3B |
| Apple | Not specified | $11B | Not specified |
AI Revenue#
| Company | 2024 AI Revenue | 2025 AI Revenue | Notes |
|---|---|---|---|
| Microsoft | $4.7B | $13B | $10B from OpenAI (at-cost pricing) |
| Amazon | $2.77B | $5B | - |
| 3B | $7.7B | Includes Google One AI and Workspace | |
| Meta | Not specified | $2B-$3B | GenAI-driven revenue estimate |
| Tesla | $0 | $0.1B | No apparent GenAI revenue |
| Apple | $0 | $0 | No direct AI revenue reported |
Big Tech AI Performance#
| Company | 2025 AI Revenue | 2025 CAPEX | Revenue to CAPEX Ratio |
|---|---|---|---|
| Microsoft | $13B | $80B | 16% |
| Amazon | $5B | $105B | 5% |
| $7.7B | $85B | 9% | |
| Meta | $2-3B | $60-65B | 3-5% |
Other Tech AI Performance#
| Company | AI Revenue | Total Company Revenue | AI as % of Total |
|---|---|---|---|
| ServiceNow | $250M ACV | ~$10B+ | ~2.5% |
| Adobe | $100M ARR | ~$20B+ | ~0.5% |
| Salesforce | $100M ARR | ~$35B+ | ~0.3% |
AI only company Financial Performance#
| Company | Revenue | Annual Costs/Losses |
|---|---|---|
| OpenAI | $10B ARR | Unknown |
| Anthropic | $4B ARR | $3B annual loss |
| Anysphere (Cursor) | $500M ARR | |
| Perplexity | $150M ARR | $68M loss |
| Midjourney | $200M ARR | Unknown |
| Ironclad | $150M ARR | Unknown |
| Replit | $100M ARR | Unknown |
| Glean | $100M ARR | Unknown |
Where is the CAPEX outflowing into?#
When it comes to where this CAPEX is flowing, the vast majority of the pile goes to Nvidia as you can see from the following table. Most of the CAPEX goes into buying GPUs. I also added other companies for comparsion.
AI Capex between 2022-2025
Lets explore the data of the AI infrastructure investments from NVIDIA, AMD, and hyperscalers like GCP, AWS, and Azure.
Loading AI timeline data...