Currents Research - February 2025

Introduction

Currents is DigitalOcean’s report on trends impacting developers at growing tech businesses around the world. For our latest edition, we focused on how these growing businesses are adapting to the increased adoption of artificial intelligence. The rise of AI has created opportunities for individual developers and businesses, but integrating new tools and technologies is not without challenges, especially for those who are not large enterprises with teams dedicated to AI. This report, gathered responses from over 1,000 developers around the world, examines how widespread adoption of AI really is, what challenges developers face when it comes to AI adoption, how usage of AI tools differs based on company size and other factors, and what the biggest upcoming trends are.

The findings reveal a significant growth opportunity, as 21% of organizations have yet to implement AI formally in any form. Despite this, there is clear interest in AI innovation: 66% of respondents are engaged in passion projects exploring AI, and 36% are actively investigating capabilities without having launched formal initiatives. For small and medium enterprises (SMEs) in particular, AI adoption could be a game-changer by allowing developers to take on multiple roles more easily—but they must first overcome key barriers, such as high Graphical Processing Unit (GPU) costs and lack of internal knowledge on how to best leverage these tools.

Key findings include:

• AI adoption is increasing, but still has growth potential: AI adoption has increased significantly over the past year—in November 2023, we found that 49% of respondents had used AI for business use, and a year later, 79% say their organization is integrating AI in some form. However, there is still clear room for more growth and maturity in organizations’ AI usage, as 32% say they are just starting to explore AI, and just 11% consider themselves an AI-driven business.

• Cost and knowledge are barriers for entry for SMEs: The high upfront costs of GPUs are a challenge for 34% of businesses, along with the lack of knowledge around how to optimize GPUs. Smaller companies feel this more acutely, and sometimes struggle to justify investments in AI as they find it challenging to demonstrate a clear return on investment from these activities, and have a lack of dedicated in-house AI resources.

• Trust, safety, and integration concerns: Other challenges with adoption for AI include integrating AI into existing workflows and choosing the right AI model. Looking at the most urgent issues in the AI space today, 47% say reliance on unverified LLM data is a top concern, along with ensuring AI data is unbiased.

• Upcoming trends in AI: While areas such as agentic AI are already growing quickly, we looked at what respondents feel is the next frontier of AI, and found that Advanced Multi Modal Systems, Real Time Audio Translation, and Automated Machine Learning are the next most compelling. Organizations currently are primarily using AI for improving internal processes and operations and enhancing their existing services with AI/ML.

These findings make it clear—accessible AI solutions are key to unlocking AI for businesses of all sizes, especially those without large teams dedicated to AI transformation. Affordable GPU compute and approachable AI tools will help alleviate the challenges many developers are currently facing when it comes to AI, and enable them to integrate AI solutions more efficiently.

Read on for the full report and results.

AI usage

79% of respondents are already using AI in some way, but many are in early stages of exploration—indicating that there’s huge potential for growth as tools become more accessible. Smaller businesses (those with under 100 employees) are using AI to augment their skills more often, while businesses with 500 or more employees are more likely to be just starting to explore AI, which could be due to the complexity of integrating AI at large organizations.

Of those already using AI tools, the most common application is to improve internal processes and operational efficiency (26%), followed by enhancing existing products with AI (25%) and developing new products with AI (22%).

Barriers to entry

GPUs, which are required for many compute-heavy AI tasks, are expensive to procure and run. While 29% of respondents stated that accessing compute wasn’t a challenge for them, 34% mentioned the high upfront cost of GPUs and 24% cited lack of internal knowledge around how to effectively utilize GPUs.

Smaller businesses with 1-10 employees are more likely to cite cost as a challenge when it comes to GPUs, with 38% of businesses that size reporting this as a challenge compared to 29% of those with 11-99 employees and 22% of those with 100-499 employees. Similarly, lack of knowledge around utilizing GPUs is more often a challenge amongst businesses with fewer employees, as they are less likely to have employees dedicated to infrastructure management and optimization.

Currents also looked at whether respondents started experimenting with AI using cloud services or their own computers, and examined the factors that go into why individuals or organizations eventually move to AI cloud services. While 63% of respondents said they started experimenting with AI for personal projects on their own computer, organizations are more likely to start on cloud platforms, with 44% stating that organizations started on cloud services. Organizations are also more likely to now use cloud services for AI, at 65%, compared to 39% for those with personal AI projects. The top reasons for moving to the cloud were access to advanced features (42%), cost optimization (38%) and scaling needs (37%).

Despite the rise in funding for AI-centric businesses, some still struggle to justify the spending that comes along with integrating AI into their organization—23% say they struggle significantly to justify AI spending, while 27% say they struggle somewhat, and 36% don’t struggle to justify this spending. When asked why they struggle to justify spending, the most popular reason was the challenge of predicting and demonstrating a return on investment (48%), followed by lack of in-house expertise needed to leverage AI technologies fully (38%) and the complexity of integrating AI with existing systems (32%).

Very small businesses with under 10 employees are more likely to struggle significantly with justifying spend on AI (25%), compared to 21% with 10-99 employees and 18% of larger businesses.

In addition to challenges around access and cost justification, we asked respondents about overall challenges in adopting and implementing AI. The most common challenges were around integrating AI into existing workflows (41%), choosing the right AI model for their needs (35%), and managing data privacy and security concerns (30%). For this question we found that integrating AI into existing workflows was a larger challenge for larger organizations, likely due to the number and complexity of their existing processes, while choosing the right AI model was cited by a higher percent of smaller businesses as a top challenge.

AI Models

Al models are the backbone for much of the Al work happening today-many of the innovative Al projects being developed use common Al models, rather than creating their own models. We looked into how respondents are thinking about Al models today, and especially how they are using open source models. We found that 83% use open source Al models as a starting point for their projects at least some of the time, with 39% using these models frequently or always. When asked about benefits of using open source, in addition to using open source code, respondents find knowledge sharing (54%), resource/tool sharing (37%), and skill development (40%) are valuable benefits of open source when it comes to Al.

The complexity of Al models has grown over time, and with this several challenges arise. We found that difficulties acquiring high-quality datasets were a top challenge for respondents (45%) followed by effectively scaling models to handle greater workloads (44% and ensuring data diversity (40%).

Future of AI

Finally, we looked into the future of Al in the minds of developers, including the most urgent issues in the Al space and what they see as the next big revolution in Al. Reliance on unverified LLM outputs was a top concern for respondents, with 47% citing this as an urgent issue in Al, while 39% mentioned ensuring Al is free of bias and 38% cited energy efficiency and the environmental impacts of Al.

Al is constantly evolving, and looking at the next big revolution in Al, we found several areas that developers feel will emerge next. Advanced multi-modal systems were the top choice (34%) | followed by automated machine learning (31%), Al-driven biotechnology (23%), and real-time audio translation (22%).

We also looked at how developers are engaging with Al outside of their everyday work, finding that 65% engage in Al-related passion projects or side jobs, demonstrating the strong interest in the area from developers. The top motivations for taking part in these side projects were exploring new technologies (69%) and enhancing professional skills (65%), which both surpassed other motivations like increasing income, cited by only 36%.

This skill expansion is also backed by the skills developers are expected to have today while full stack development is the most in-demand skill for developers, with 57% saying their organization looks for this, 28 say model training and deployment is a skill they look for, 25% say developers are expected to have expertise in managing ML lifecycles, and 25% cite Al deployment as a key skill developers they hire should have.

Conclusion

It’s clear that when it comes to Al, developers and SMEs face a sometimes conflicting reality-while there is immense potential for growing businesses to incorporate Al to streamline processes and enhance product offerings, there are significant challenges when it comes to Al integration. Cost of GPU compute and the lack of internal knowledge on how to best leverage Al and optimize GPUs are acute challenges for non-enterprise businesses, which is why they need access to cost-effective compute and simpler solutions to take advantage of all that Al has to offer.

At DigitalOcean, we understand these challenges and opportunities deeply. Our accessible, cost-effective, and developer-friendly solutions are designed to empower businesses of all sizes especially those without the resources of large enterprises. GPU Droplets allow anyone to access on-demand GPU compute at affordable prices, while our new GenAl product makes agentic Al accessible to even the smallest of teams by removing the complexity often associated with Al tools. By addressing key barriers and delivering intuitive tools, we help businesses leverage Al to streamline operations, enhance services, and innovate confidently in a competitive market.

As Al continues to reshape industries, DigitalOcean is ready to support developers and organizations in realizing their Al ambitions-join us by signing up for an account, reading a community tutorial, or speaking to our experts about your needs today.

Methodology

This survey was conducted through an online survey link from October 25, 2024 to December 31st, 2024 which garnered over 1,500 responses. The link was distributed to DigitalOcean email lists.

Approximately 24% of respondents were full-stack developers, 14% were CEOs, founders, or owners, and 11% were back-end developers. CTOs made up 6%, while systems architects and systems administrators each accounted for 3%. All other roles, spanning from DevOps specialists to marketing professionals, summed up to 34%. The remaining 5% fell into the “Other” category.

Respondents represent 96 countries, with 24% being in the United States, 9% in India, 3% in the UK, 4% in Germany, and 4% in Canada. The questionnaire was developed by DigitalOcean and was distributed via link through email to both DigitalOcean customers and non-customers.

Try DigitalOcean AI/ML solutions today

DigitalOcean has the AI/ML solutions you need, from powerful GPUs to a GenAI platform that makes AI accessible to anyone. Sign up or speak to our sales team about your AI/ML needs.