AI Soft Landing Scenario

AI infrastructure spending is shifting from an explosive expansion to a slower and more predictable cycle rather than bringing the AI trade to an end.

• After three years of spending growth that repeatedly exceeded forecasts, Wall Street expects hyperscalers' AI infrastructure capex to reach about $800bn in 2026, rise 54% to $1.2tn in 2027, and grow only 12% to $1.4tn in 2028.
• The slowdown is most dangerous for neoclouds, which rely on high-yield debt; bonds from one prominent issuer yield close to 15%, while its credit default swaps trade at distressed levels.
• Slower neocloud growth could ease the memory shortage: DRAM contract prices rose roughly 90–95% quarter on quarter in early 2026, while spot memory chips climbed from about $12 to a record $42.50.
• Markets price at least 50 basis points of further Federal Reserve hikes and rates near 4.5% through 2027, but cooler component demand could reduce goods inflation and allow rate cuts from the second half of 2027 into 2028.
• Stronger borrowers may acquire cash-strained competitors, creating consolidation among weaker high-yield AI issuers as the financing gap between investment-grade and high-yield companies widens.

Takeaway: Investors should position for a decelerating but continuing AI cycle, emphasizing borrower quality and the potential credit-selection and consolidation opportunities created by weaker financing conditions.