A recent survey of over 1,100 developers, CTOs, and founders by DigitalOcean reveals that AI agents are delivering real return on investment (ROI), with 67% of organizations using them reporting productivity gains. However, scaling these agents in production remains a challenge, with only 10% of respondents currently doing so. The main obstacle to scaling is the high cost of inference, with 49% of respondents citing it as the top barrier to growth.
The survey found that 52% of companies are actively implementing AI solutions, including agents, which is a significant increase from the 35% reported in the previous year. Of those respondents, 46% are specifically deploying AI agents, which are being used for a range of tasks, including code generation and refactoring (54%), automating internal operations (49%), and building customer support and chatbots (45%). OpenAI and other companies are also working on AI agents, with Google releasing the Agent Development Kit as an open-source framework, allowing for the creation of coordinated multi-agent systems.
The survey also found that 60% of respondents believe that applications and agents represent the greatest long-term value in the AI stack, with 37% expecting budget growth in this area over the next 12 months. In contrast, only 14% of respondents expect budget growth in infrastructure, and 17% in platforms. Nvidia and AMD are among the companies that are working on improving the infrastructure for AI, with AMD working with DigitalOcean to optimize inference costs for companies like Character.ai.
As AI agents move from pilots to production, the companies that scale successfully will be the ones that are able to absorb the complexity of optimizing GPU configurations, managing parallelization strategies, and fine-tuning model serving infrastructure. DigitalOcean is investing in inference optimization with its Gradient AI Inference Cloud, which aims to reduce the cost of running AI at scale. With the right infrastructure in place, companies like Character.ai are able to double their production inference throughput and reduce their cost per token by 50%. As Wade Wegner, Chief Ecosystem and Growth Officer at DigitalOcean, notes, solving for inference shouldn’t fall on developers, and cloud providers need to absorb this complexity to enable the widespread adoption of AI agents.
The shift from experimentation to production is well underway, with 2026 shaping up to be the year when more teams move agents into production. As Andrej Karpathy notes, “vibe coding” is becoming increasingly popular, where developers describe what they want in plain language and let the AI write the code. With the right tools and infrastructure in place, the potential for AI agents to deliver significant productivity gains and cost savings is substantial, and companies like Ring and Y Combinator are already seeing the benefits of using AI agents in their operations.
In conclusion, the survey findings suggest that AI agents are delivering real ROI, but scaling them in production remains a challenge due to the high cost of inference. As companies like DigitalOcean invest in inference optimization and cloud providers absorb the complexity of optimizing GPU configurations, the potential for AI agents to transform industries and deliver significant productivity gains and cost savings is substantial. With the right infrastructure and tools in place, 2026 is likely to be the year when AI agents graduate from pilot to product, and companies that are able to scale them successfully will be the ones that reap the benefits.
Leave a Reply