Dragonfly Partner Haseeb Qureshi: Apollo AI Return Analysis Conclusion Reversed
Haseeb Qureshi criticized the Apollo report claiming that AI generated zero financial returns outside of the Mag 7, stating that this conclusion is untenable.
The report inferred that corporate AI spending did not yield returns based on profit margins and per capita income being equal outside of the Mag 7.
Qureshi believes that measuring variable contributions requires comparing companies that use more AI with those that use less; overall profit margins cannot answer this question.
He pointed out that the Mag 7 mostly consists of AI sellers, while the other companies are primarily buyers, making the split between Mag 7 and other companies an incorrect analytical unit.
With current rising interest rates, oil prices reaching $116, high inflation, and consumer confidence at historical lows, buyer companies maintaining profit margins in this context aligns with the reality that AI supports the economy.
Capital continues to flow towards models and infrastructure sellers, while buyer spending is viewed by some institutions as a cost center; however, under macro pressure, profit margins have not significantly declined, prompting a reassessment of AI's actual contribution to non-tech companies amidst methodological disputes.
Source: Public Information
ABAB AI Insight
Haseeb Qureshi previously worked as an engineer at Airbnb before shifting to investments in crypto and AI, and is now a partner at Dragonfly, continuously commenting on AI capital allocation and macro impacts, and has invested multiple times in related infrastructure and application layer companies.
Apollo's chief economist Torsten Slok released a report based on S&P 493 profit margins and per capita income being equal, categorizing AI spending as a cost center for non-tech companies. Qureshi, however, interprets the same data by comparing usage intensity differences and macro environments, suggesting that AI may be offsetting rising interest and energy costs.
This is similar to historical debates about the lag in productivity statistics during early technology cycles, such as the initial lack of immediate reflection of productivity data from IT investments in the 1990s, or the skepticism regarding application spending returns before the widespread adoption of mobile internet. Currently, AI is in a phase of concentrated selling and buyer penetration.
Essentially, this represents a shift in pricing power: as foundational model capabilities diffuse, return measurements shift from seller revenue growth to buyer profit maintenance under pressure, and the judgment of capital allocation transitions from merely observing tech giants' profits to an analytical framework that can distinguish usage intensity differences, with conclusions that do not employ comparative methods facing repricing.
ABAB News · Cognitive Laws
- Aggregated data being equal cannot prove variable contributions; comparisons are needed to isolate effects.
- When seller growth and buyer stability coexist, the pressure environment reveals actual support.
- When methodological errors occur, reversed conclusions can change capital flows more than data absence.