Goldman Sachs' most recent survey of chief information officers indicates that corporate commitment to artificial intelligence is deepening even as broader IT spending momentum cools.
Analyst Gabriela Borges reported that Goldman's Overall IT Spending Index declined to 65.0 from 68.0, and the IT Capital Spending Index dropped to 59.0 from 65.5. Both measures have now fallen in each survey since November 2024.
Borges noted that part of the softness in aggregated IT budgets appears linked to uncertainty around AI, saying there "may be some crowding out of category spend not tied to AI" as companies defer larger budget commitments until they have a clearer view of how AI will affect core budget categories.
Despite the weaker readings for total IT spend, AI-specific intentions are increasing. Over a three-year horizon, 42% of CIOs now expect generative AI to represent more than 10% of their IT budgets, up from 35% in the prior survey. The weighted average expectation for generative AI as a share of IT budgets stands at 10%.
Goldman also found that two-thirds of AI inference costs are expected to be covered by reallocating existing budgets. That reallocation includes an 18% shift away from spending on application software, a pattern that Goldman interprets as evidence AI is already compressing near-term software and labor budgets.
Hardware demand, by contrast, is strengthening sharply in the survey responses. Sixty percent of respondents said they now plan to increase server spending, a rise from 37% previously. Expectations for higher storage spending climbed to 75% from 47%.
Borges connected these shifts to supply conditions in memory markets, noting DRAM and NAND tightness as a factor driving higher average selling prices for hardware components.
On the software front, CIOs still lean toward procurement rather than in-house development. Fifty-six percent of survey participants expect to "build less and buy more," a stance Goldman suggested could temper investor worries about widespread enterprise investment in home-grown AI stacks.
What this means in practice
- Enterprises are prioritizing AI allocations even as overall IT and capital spending indices drift lower.
- Budget reallocation is the primary mechanism funding AI inference costs, with notable diversion from application software.
- Demand for servers and storage has surged in expectations, influenced by tighter DRAM and NAND supply and higher component prices.
These survey responses suggest a phase in which companies are reshaping near-term procurement and project priorities around AI capabilities, while remaining cautious about large, non-AI commitments until outcomes and budget impacts are clearer.