Goldman Sachs recently released a blockbuster research report, "Tracking Trillions: The Assumptions Shaping the Scale of the AI Build-Out," predicting that from 2026 to 2031, cumulative capital expenditures on global AI infrastructure will reach$7.6 trillion.
How amazing are the numbers?
Goldman Sachs predicts that AI capital expenditures in 2026 will be approximately$765 billion, and climbs year by year to 2031$1.6 trillion/year.
Four key variables determine whether RMB 7.6 trillion is a bubble or inevitable
The Goldman Sachs report specifically pointed out that the market should not only focus on the revenue side (demand) of AI applications.Four variables on the supply sideThis is the key to determining whether this huge investment can be established:
- Chip economic life: How long is the depreciation cycle of AI chips?
- Data center construction costs: The cost of building large data centers has soared from hundreds of millions of dollars to billions of dollars, and the construction cycle has been lengthened to 2-3 years.
- Power bottleneck：一个超大规模AI数据中心可能需要1GW以上的电力供应，相当于一座小型城市的用电量。
- supply chain constraints: From GPU to high-end network equipment, can it continue to meet the explosive growth demand?
What does it mean for the average user?
This US$7.6 trillion infrastructure investment will eventually be passed on to every user:
- AI services may continue to rise in price: Infrastructure costs are so high that the era of free or low-cost AI services may be drawing to a close.GitHub CopilotPay-as-you-go,ChatGPTSubscription price increases are precursors.
- Local AI devices may have a renaissance: When the cost of cloud inference remains high, local AI inference (on-device AI) promoted by Apple, NVIDIA, etc. will be more attractive.
- The value of open source models is highlighted: Enterprises may prefer to use an open source model for on-premises deployments to avoid the ongoing costs of cloud APIs.
What should investors watch?
Goldman Sachs recommends investors closely monitor three indicators:Chip depreciation cycle(Whether there are signs of accelerated replacement),Power expansion progress(Can the power grid transformation keep up?), andReal revenue growth from AI applications(You can’t just look at the number of users, but paid conversions).
Summarize
AI infrastructure is becoming the largest single industry investment direction in human history.Without electricity, chips, and data centers, no matter how powerful the AI is, it can only remain in papers..
focus on AI Dash, discover the most noteworthy AI tools and industry trends.