tech
Goldman Sachs says the AI boom is bigger than investors think
The bank expects token consumption to increase 24 times by 2030, largely driven by the rise of enterprise agents.
TL;DR
- Goldman Sachs analysts suggest current hyperscaler capital expenditure estimates for 2027 are too conservative.
- They project capital spending could reach $1.1 trillion by 2027, with a bullish scenario reaching $1.4 trillion.
- The forecast is based on the expectation that AI computing power demand is in its early stages, with token consumption predicted to grow 24 times by 2030.
- Despite significant spending, companies are struggling to prove AI's return on investment, and productivity gains may not yet exceed model running costs.
- Cloud providers like Google Cloud and AWS have reported substantial backlogs, signaling strong demand.
- AI-related investment as a percentage of GDP is expected to be comparable to historical investment booms in areas like railroads and electrification.
- Physical constraints such as data center availability, memory, power, and labor are identified as potential bottlenecks for the AI buildout.
- While earnings growth is expected for AI infrastructure suppliers, crowded trades and high valuations increase the risk of volatility.
- Evidence of widespread AI-driven productivity gains remains limited, with few companies quantifying specific benefits or earnings impacts.