A Bloomberg Opinion columnist argues that the tech industry's massive data center buildout could prove a poor investment unless companies make AI agents more useful and attract wider adoption.
Opinion · AI infrastructure
The data center buildout may hinge on whether AI agents become useful
Huge spending on computing capacity assumes future revenue. If AI agents, systems that carry out multistep tasks for users, fail to become both more capable and widely adopted, the return on that spending could fall short.
The dependency chain
Each step must hold for the infrastructure bet to pay off
1
Capital spending
Capacity to train and run models
→
2
Agents get more capable
Multistep tasks done reliably
→
3
Wide adoption
Users rely on agents at scale
→
4
End-user value
Revenue that justifies the buildout
Why the bet could still work
Agents are widely seen as a path to more autonomous and economically valuable AI.
Demand for infrastructure could still rise even without a breakout in agent use.
Why it could fall short
Agents fail to become useful enough to attract wide use.
Returns fall below what investors and operators are counting on.
Core claim
The infrastructure race is a bet on one application of AI reaching scale, not on AI demand in general.
Condition A
Agents become more capable
Condition B
Agents become more popular
Condition C
Both happen at scale, not just in pilots
What would clarify the outcome
An open question, likely resolved over time
Agent products mature and show real use
Companies disclose how their AI capacity is actually used
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article...