Jul 6, 2026, 3:16 PM ETFireside Chat, Interview

AI: What Investors Should Know

Total capital expenditures (CapEx) related to AI infrastructure are projected to exceed $3 trillion by the end of 2026.
Goldman Sachs analysts characterize the current environment as containing two distinct market conditions: an "earnings bubble" in public markets driven by unsustainable downstream demand, and a "valuation bubble" in private markets where unproven business models hold robust valuations.
The sustainability of current capital spending hinges on enterprise AI adoption; without wholesale enterprise integration, the infrastructure investment cannot justify returns on invested capital, as consumer usage generally does not generate direct revenue.

Consumer AI adoption is robust with over one billion daily active users, whereas enterprise adoption remains disappointing due to significant hurdles in integrating unstructured, disparate data sets.
A documented disconnect exists between executive optimism and frontline reality, driven by AI's inability to currently consolidate diverse organizational data without hallucinations or errors.
Despite CEO optimism, mass labor replacement is not expected; the technology is positioned to increase existing workforce efficiency rather than eliminate entry-level jobs required to develop the human judgment necessary to validate AI outputs.
The sector is shifting toward task-specific Small Language Models (SLMs) over monolithic Large Language Models (LLMs), which offer reduced hallucination rates, lower power consumption, and higher efficiency for specific business purposes.

Circular financing practices are identified as a risk factor, particularly when suppliers fund customers to maintain purchasing volume before the end users generate sufficient independent cash flow to sustain the expenditure.
Public discourse regarding potential market bubbles is viewed as a positive, "healthy" development that differentiates the current cycle from the 1999–2000 internet boom, potentially insulating the broader market from severe fallout if high-profile IPOs fail.
Geopolitical winners in the AI space are contingent on which value layer dominates:
- Model Layer: China may emerge as the winner due to its potential to act as a low-cost producer.
- Chip Layer: The United States is favored to maintain dominance by controlling major AI chip applications.
- Application Layer: The value distribution remains undetermined, dependent on whether cost savings can be realized by enterprises.
The industry anticipates that large-cap AI companies may not achieve profitability for another four to five years, complicating current IPO trajectories.