US Tech Leaders at AI Inflection Point, Compute Demand Outstrips Supply
Quick Look
- US tech leaders are at an "inflection point" with agentic AI, but compute demand far exceeds supply, a gap expected to persist until late 2027.
- This comes amid skepticism about AI infrastructure spending returns, amplified by cheaper Chinese AI models.
AI-generated summary
Why It Matters
US tech leaders are at a critical juncture with the advent of economically productive agentic AI tools. This follows significant capital expenditure on core infrastructure like data centers and semiconductors. However, questions persist about the return on these investments, especially with the emergence of cheaper AI models from China.
Instead of an AI bubble, US tech leaders are standing at a pivotal “inflection point” where the arrival of economically productive agentic AI tools has provided early vindication of the industry’s unprecedented capital expenditure on core infrastructure such as data centres and semiconductors, Sheridan said on the sidelines of Goldman Sachs’ Asia Communacopia + Technology Conference in Hong Kong on Monday.
“There’s a pretty big disconnect between the demand and the availability of compute,” Sheridan told the South China Morning Post. “We don’t think that imbalance closes until well into the second half of 2027.”
His remarks come amid recurring questions about whether the historic levels of AI infrastructure spending in the US – on track to top US$700 billion this year – will be able to deliver proportional investment returns.
Such scepticism has only been amplified as Chinese companies have open-sourced cheaper AI models, prompting some analysts to predict an erosion of margins for US model providers.
What to Watch
AI outlook — possibilities, not facts
The imbalance between compute demand and availability will not close until well into the second half of 2027.
Likely · Long term
Open Questions
- What specific AI tools are considered 'agentic' and 'economically productive'?
- What are the projected profit margins for US AI model providers?
- How will the availability of cheaper Chinese AI models specifically impact US companies?
- What are the specific technological advancements driving the demand for compute?
