The numbers are staggering: Alphabet, Amazon, Meta, and Microsoft will collectively invest around $650 billion in AI-related infrastructure during 2026. That is a 58% jump from the estimated $410 billion spent last year. But there is a troubling data point: according to Goldman Sachs, all that investment contributed "basically zero" to United States economic growth.
Who is spending what
I have been tracking these companies' quarterly reports for a while, and every quarter the CapEx figures get more aggressive. Here is the estimated breakdown for 2026:
| Company | Estimated CapEx 2026 | CapEx 2025 | Growth |
|---|---|---|---|
| Microsoft | ~$80B | ~$55B | +45% |
| Alphabet (Google) | ~$75B | ~$52B | +44% |
| Amazon (AWS) | ~$100B+ | ~$75B | +33% |
| Meta | ~$60-65B | ~$38B | +63% |
| Total | ~$650B | ~$410B | +58% |
To put these numbers in perspective: $650 billion is more than the GDP of countries like Argentina, Colombia, or Sweden. It is a historic bet on a technology that has not yet demonstrated proportional returns.
Where the money goes
The bulk of the investment goes to three main areas:
- Data centers: building new AI-specialized facilities with advanced cooling and dedicated power
- GPUs and chips: massive purchases of NVIDIA GPUs (Blackwell and upcoming Rubin) plus development of custom chips (Google TPU, Amazon Trainium, Meta MTIA)
- Energy: nuclear and renewable energy contracts to power data centers that consume enormous amounts of electricity
Goldman Sachs: economic impact basically zero
The most unsettling finding comes from a Goldman Sachs analysis: the massive AI investment contributed "basically zero" to US GDP growth in 2025. That is, the hundreds of billions spent on AI infrastructure have not yet translated into measurable productivity at the macroeconomic level.
This does not mean AI does not work — it means the economic benefits have not materialized at the scale that would justify the investment. In my experience working with AI tools, individual productivity improves noticeably, but translating that to a country's GDP is a different story entirely.
Is this a bubble?
The question everyone is asking. Here are the arguments on both sides:
Arguments that it IS a bubble
- AI revenue for these companies is not growing at the same pace as investment
- Most companies paying for AI still do not see clear ROI
- The pattern resembles pre-2000 telecom: overinvestment in infrastructure
- Goldman Sachs says the economic impact is zero
Arguments that it is NOT a bubble
- NVIDIA reported $68.1B in quarterly revenue — real demand is enormous
- Unlike the dot-com era, these companies are profitable with strong cash flow
- AI is already generating real revenue (Azure AI, Google Cloud AI, AWS Bedrock)
- Technological advances keep accelerating with each model generation
How this affects you as a user or developer
The massive investment has direct effects on your tech life:
- Cloud prices are rising: AWS, Azure, and GCP have already adjusted prices upward. If you use cloud services, expect 10-20% increases in 2026
- More free or cheap AI tools: companies need users to justify the investment, so they will keep launching accessible AI products
- Tech employment: demand for AI engineers, MLOps, and data center specialists is at historic highs, but other tech roles are under pressure
Common questions for tech investors
Should I sell my Big Tech stocks?
I am not a financial advisor, but historically companies that invest aggressively in infrastructure during technology shifts (think Amazon in cloud during 2010-2015) ended up dominating the market. The risk is real, but so is the opportunity.
What happens if the bubble pops?
Unlike the dot-com era, these companies have profitable core businesses (advertising, cloud, e-commerce). A correction would be painful but not existential for them. The most affected would be AI startups with no revenue of their own.
Which AI companies will survive?
Those with distribution (users) and technical differentiation. Anthropic, OpenAI, and the cloud providers are well positioned. Startups that only build API wrappers face more risk.
Global context
The AI race is not just American. China is investing comparable amounts through companies like Alibaba, Tencent, and Baidu, plus direct state support. The European Union, meanwhile, debates regulations while trying not to fall behind in the technology race.
