AI Janus face
Substack Post Week July 20-24
The global economy is entering a period where the distinction between headline strength and reality is becoming important (and more uncertain than ever). Across financial markets, trade data, corporate investment, and credit markets, the same pattern is emerging: large nominal numbers coexist with weaker real transmission mechanisms. Capital is moving aggressively into selected sectors, particularly AI infrastructure, while credit conditions are becoming more cautious and private demand remains fragile (at best).
The central question is not whether the AI investment cycle, private credit expansion, or infrastructure buildout are real. They are. The more important question is whether the scale of financial resources being committed today is producing a proportional increase in employment and sustainable economic growth. Recent developments suggest that the answer is becoming more complicated (closer to “no”).
Alphabet’s recent financial results provide one of the clearest examples. Google’s parent company reported the first negative quarterly free cash flow in its history as a public company, with free cash flow falling to approximately negative $5.9 billion. This does not represent a solvency problem. Alphabet remains one of the most profitable companies in the world. However, the scale of AI-related capital expenditure demonstrates how the nature of the investment cycle has changed.
Alphabet spent roughly $44.9 billion on capital expenditures during the quarter, almost double the amount from a year earlier, and management expects investment to continue increasing. Other hyperscalers are following the same path. The AI race has become an industrial mobilization requiring enormous amounts of semiconductors, electricity, data center capacity, specialized construction, but also financing. What was once perceived primarily as a software revolution resembles today a capital-intensive infrastructure cycle.
The challenge is that higher investment does not automatically translate into equivalent increases in real capacity. Data centers require advanced processors, memory, networking equipment, cooling systems, transformers, land, and energy connections. Many of these inputs are becoming scarce and expensive. As a result, companies may be spending significantly more while obtaining less additional capacity than headline capital expenditure figures imply.
