State of AI 2026: ROI rises as agentic tools and open source spread

AI is moving from experimentation to everyday operations, and companies say the payoff is increasingly measurable. NVIDIA’s 2026 “State of AI” industry surveys, fielded from August through December 2025 and based on more than 3,200 responses worldwide, suggest that enterprise adoption is rising alongside reported gains in productivity, revenue, and cost reduction.

## NVIDIA’s results span financial services, retail and consumer packaged goods, healthcare and life sciences, telecommunications, and manufacturing, offering a snapshot of how organizations say they are deploying AI and what they expect next. Across the combined surveys, 64% of respondents said their organizations are actively using AI in operations, while 28% reported they are still assessing projects and 8% said they are not using AI and have no plans to start.

The survey also points to regional differences. North America led with 70% reporting active AI use, compared with 65% in EMEA and 63% in APAC, where a larger share of respondents, 15%, said they were not using AI. Company size mattered as well: respondents from organizations with more than 1,000 employees reported higher adoption and stronger returns, with 76% saying they are actively using AI and only 2% saying they do not use it at all.

NVIDIA’s data suggests the center of gravity is shifting from pilots to scaled deployments, as executives push projects into production and focus on specific, high-impact use cases. Nasdaq was highlighted as one example of a large organization building an AI platform to improve internal operations and external products. Michael O’Rourke, senior vice president and head of AI and emerging technology at Nasdaq, described AI as a way to connect data across the company’s businesses and technologies to improve products and services.

When respondents were asked about goals, the top themes were operational efficiencies at 34%, improving employee productivity at 33%, and opening new business opportunities and revenue streams at 23%. More than half of respondents, 53%, said improved employee productivity was among AI’s biggest impacts on business operations. The telecommunications survey was especially emphatic, with 99% of respondents saying AI improved employee productivity, and about a quarter saying the improvement was major or significant.

NVIDIA’s report also emphasizes that productivity gains can cascade into broader operational outcomes. In the overall results, 42% said AI created operational efficiencies and 34% said it helped open new business and revenue opportunities. Manufacturing examples included Siemens integrating AI into tools used by manufacturers, and a PepsiCo deployment with Siemens and NVIDIA to build high-fidelity 3D digital twins for selected U.S. facilities. NVIDIA said the approach helped identify up to 90% of potential issues before physical changes, delivering a 20% throughput increase in initial deployments, nearly 100% design validation, and 10–15% reductions in capital expenditure.

On ROI more directly, respondents reported broad-based financial impact. In the combined surveys, 88% said AI increased annual revenue in some or all parts of the business, with 30% reporting gains greater than 10%, 33% reporting 5–10% growth, and 25% reporting less than 5%. Executives were more likely to report larger gains, with a little over 40% of C-suite and vice president respondents saying they saw annual revenue increases above 10%.

Cost reduction showed a similar pattern. Overall, 87% said AI helped reduce annual costs, and 25% said the decrease exceeded 10%. Retail and CPG stood out within the verticals, with 37% saying costs fell by more than 10%. NVIDIA cited Lowe’s as an example, describing how the retailer built AI-powered digital twins of more than 1,750 stores and used AI to speed asset discovery and generate 3D models from 2D product images in minutes at a cost of less than $1 per model.

Another major thread in the surveys is the rise of “agentic AI,” systems designed to reason, plan, and execute tasks based on high-level goals. NVIDIA said 44% of companies were deploying or assessing agents during the survey period, capturing what it characterized as an experimentation phase that has since moved toward broader deployments in early 2026. Telecommunications reported the highest adoption rate at 48%, followed by retail and CPG at 47%. In healthcare, NVIDIA pointed to Mona by Clinomic, an onsite assistant for intensive-care units, which it said reduced documentation errors by 68% and cut perceived workload by 33%.

Model choice and software strategy also emerged as a defining factor. Across respondents, 85% said open source is moderately to extremely important to their AI strategy, including 48% who said it is very to extremely important. Smaller companies were even more likely to emphasize open source, with 58% rating it very to extremely important, reflecting a preference for building solutions and tailoring models with their own data.

The survey suggests investment is following perceived success. Overall, 86% of respondents said their AI budget will increase in 2026, while 12% said it will stay the same, and nearly 40% predicted increases of 10% or more. Respondents said spending priorities include optimizing AI workflows and production cycles at 42%, finding additional use cases at 31%, and building or expanding access to AI infrastructure, whether on-premises or in the cloud, also at 31%.

Despite the upbeat picture, the same data highlights persistent barriers. Data issues were the most frequently cited challenge, with 48% pointing to having sufficient data or related problems. A lack of AI experts and data scientists followed at 38%, underscoring that scaling from pilot to production still depends on specialized skills. Another 30% cited lack of clarity on AI’s ROI, a reminder that even when benefits like productivity are real, they can be difficult to quantify consistently across roles and workflows.

NVIDIA said respondents came from a mix of roles, including C-suite and vice presidents at 27%, directors and managers at 33%, and AI practitioners at 40%. The geographic distribution spanned APAC at 32%, North America at 26%, EMEA at 21%, and the rest of the world at 20%, with responses sourced from NVIDIA distribution lists, social media, and third-party agencies in China and Japan.