AI earnings: Big growth, bigger spending, tougher questions

The AI investment cycle is entering a more demanding phase in which strong demand must translate into profitable cash flows and measurable productivity gains.

• Cloud demand remains robust: Azure revenue grew 48%, Google Cloud expanded 103% and AWS grew 39%, but Amazon expects approximately $220 billion of capital expenditure in 2026, raising questions about returns and financing costs.
• AI monetisation is uneven: Meta's advertising impressions increased 10% and average ad prices 14%, while ServiceNow raised its 2026 AI annual contract value target to $1.7 billion; elsewhere, infrastructure and development costs are rising faster than revenues.
• Infrastructure suppliers are reporting exceptional profitability, including a 66.4% gross margin at TSMC, an 80% operating margin at SK hynix and a 57.5% gross margin at ASML, but capacity expansion could eventually increase competition and pressure prices.
• The earnings cycle is broadening beyond chipmakers: Seagate generated $4.2 billion of revenue and $1.5 billion of quarterly free cash flow, while Intel's data-centre and AI revenue grew 78% despite weaker guidance and a soft PC market.
• Broad economic productivity gains remain unproven because adoption requires implementation, training, workflow changes and human oversight, and some benefits may accrue to customers or competition rather than the companies buying AI tools.

Takeaway: Investors should distinguish companies that can finance AI expansion through recurring cash flow, apply valuation discipline and diversify beyond infrastructure, while waiting for evidence that adoption produces durable improvements in margins and returns on capital.