49% of US firms report that on-premises software delayed AI projects

Google Cloud and the research firm Hypothesis Group surveyed 1,250 business and technology decision-makers in the US. Google Cloud reports that 49% of them had AI or other technology projects delayed, cut, paused or canceled. On-premises software, which runs on local computers and servers, caused this, they say. Jeanette Manfra, the Google Cloud head of risk and compliance, wrote about the results.

What the research says

Manfra says that patches and upkeep use the time of engineers. AI work could use that time. AI tools find more flaws than the teams can repair.

The numbers

• 49% say that on-premises software delayed, cut, paused or canceled AI or other technology projects.
• 80% say that the software a provider sells must not set their AI.
• 78% say that the leaders of the last era are not always right for the AI era.

What is not known

The research includes only the US. Google Cloud made it and gives no method. It does not test other causes. Manfra also names tools of Google Cloud as part of the answer.

Insight: Two reports name two different causes of weak AI results

Gartner said on 19 October 2026 that 1 in 5 AI projects succeed. It told leaders to name an owner for each tool. Google Cloud blames legacy software and wants buyers to have more than one provider. This reads as a plan to make legacy software the cause of weak AI results. Find which cause your projects show before you accept one.