Menlo: enterprise generative-AI budgets rise
- Menlo Ventures said enterprise generative-AI spending hit $37 billion in 2025, up from $11.5 billion in 2024, even as ROI remains uneven. (menlovc.com) - The tension is stark: Menlo sees broad deployment and $19 billion flowing to applications, while an NBER-linked survey found 89% saw no productivity change. (menlovc.com) - That matters because capital is still pouring in — including Encord’s $60 million round for physical-AI data infrastructure — despite shaky proof of gains. (crowdfundinsider.com)
Enterprise AI budgets are going up fast. That is the simple headline. Menlo Ventures’ latest enterprise AI report says companies spent $37 billion on(menlovc.com)m is landing at the same time many companies still cannot point to clean, measurable productivity gains. (menlovc.com)t more experimentation. Menlo’s view is that enterprise generative AI has moved from pilot mode toward real budget lines, with more than half of 2025 spending (crowdfundinsider.com)odels, which tells you companies are buying tools meant to change work now, not someday. (menlovc.com) ### Why does that matter? Because application spend is the closest thing to a real conviction signal. A company can dabble with a model API or run a lab project from an innov(menlovc.com)inside the business thinks the tool is worth operational risk, procurement hassle, and change management. Menlo also says enterprise AI now equals more than 6% of the software market, which is a huge number for something that barely existed three years ago. (menlovc.com) ### So why are people still doubtful? Because the payoff is still murky. A large (menlovc.com)AI on employment or productivity, and 89% reported no change in productivity measured as sales per employee. In other words, companies are spending like this is a platform shift, but many still cannot prove the gains in the language finance teams care about. (theregister.com) ### Is that a contradiction? Not really — but it is an uncomfortable phase. New technology often works in narrow pockets before it shows up in company-wide metrics. Coding, support, and con(menlovc.com) is messy, data is bad, managers do not redesign jobs, and employees use the tools unevenly. Basically, local wins do not automatically become enterprise ROI. (theregister.com) ### What does Menlo think is happening? Menlo is arguing that demand is real even if the market mood has turned more skeptical. Its report explicitly pushes back on the “bubble” story and says adop(theregister.com)re now at least 10 AI products above $1 billion in ARR and 50 above $100 million, led by model providers like Anthropic, OpenAI, and Google but spreading into departmental software. (menlovc.com) ### Where is venture money going now? Not just into frontier models. Encord raised $60 million in a Series C round in March 2026 to build what it calls an AI-native d(theregister.com)s that run on video, sensor, and multimodal data. That is a useful tell. Investors are still funding picks-and-shovels companies that help AI systems work in production, especially where reliability and data quality matter more than demo magic. (crowdfundinsider.com) ### Why is physical AI part of this story? Because it shows where the market is (menlovc.com)hifts toward data pipelines, evaluation, monitoring, and governance. Encord said data on its platform grew from about 1 petabyte to more than 5 petabytes in a year, and revenue from physical-AI customers grew tenfold. That is not consumer hype — that is infrastructure strain. (crowdfundinsider.com) ### What is the bottom line? Enterprise AI is in the expensive middle stage. Bud(crowdfundinsider.com)rket is no longer asking whether companies will buy AI. It is asking which purchases will survive the next budgeting cycle.
