What unprecedented AI capex means for markets and investors

AI capital expenditure is becoming a global, increasingly leveraged growth cycle whose continuation depends on monetization, infrastructure availability and investor appetite.

• J.P. Morgan projects global AI capex of $900 billion in 2026 and approximately $5.5 trillion by 2030, with Amazon, Alphabet, Meta, Oracle and Microsoft together set to spend about as much in 2026 as since ChatGPT launched in November 2022.
• The buildout is spreading beyond U.S. hyperscalers: the EU aims to triple data-center capacity within 5–7 years and has mobilized more than €10 billion for up to seven AI gigafactories, while ByteDance is reportedly investing about $39 billion in a Brazilian data center.
• Early monetization evidence comes from cloud businesses: AWS revenue rose 37% year over year to a $169 billion annualized run rate with $496 billion of backlog, Google Cloud revenue rose 82% with $514 billion of backlog, and Oracle Cloud Infrastructure revenue increased from $2.1 billion in Q1 2025 to $5.8 billion in Q4 2026.
• Financing is shifting from hyperscalers’ free cash flow toward debt and equity: Alphabet raised $85 billion in equity in June 2026, while operating cash flow across hyperscalers is still expected to exceed $900 billion by 2027.
• Physical constraints could slow or reduce returns: data centers may represent 9%–17% of U.S. electricity demand by 2030 versus 3%–4% today, construction takes 18–30 months, and shortages of memory, GPUs, power, land and skilled labor are raising costs and delaying projects.
• Investors should monitor whether spending reaches projected levels, whether cloud monetization supports unit economics, and whether debt, equity issuance and future AI IPOs can absorb the financing needs without pressuring valuation multiples.

Takeaway: The investment case remains supported by accelerating cloud revenue and backlog, but the scale-up is becoming more dependent on external financing and vulnerable to power, equipment, permitting and political constraints, raising the hurdle for earnings breadth and cash conversion.