Goldman Says AI Spending Shifts Beyond Chips as Intangible Investment Nears $1 Trillion

Goldman Sachs said investors are underestimating the scale of non-hardware spending required for artificial intelligence adoption, including software, data infrastructure, workforce retraining and corporate reorganization. The bank estimates U.S. AI-related hardware capital expenditure has reached $360 billion, or 1.1% of GDP. But it said related intangible investment could total about $700 billion in the U.S. and approach $1 trillion globally, based on historical links between information-technology hardware and software, data and organizational spending. Goldman estimated non-tech companies are spending about $153 billion annually on internal AI tool development through IT departments. AI-related organizational capital spending exceeds $40 billion a year, while workforce restructuring costs are currently about $10 billion annually and could reach $800 billion to $900 billion over the full adoption cycle if AI ultimately displaces 6% to 7% of workers. Snowflake, Databricks and Palantir have more than tripled revenue since ChatGPT’s 2022 launch, while their combined enterprise value has risen from below $100 billion to more than $650 billion by late 2025, Goldman said.