Industry executives and experts share their predictions for 2025. Read them in this 17th annual VMblog.com series exclusive.
By Laurent Doguin, director of developer relations and strategy, Couchbase
As we enter 2025, businesses find themselves at a technological crossroads where AI regulation, data privacy and infrastructure decisions demand immediate attention. While companies race to harness AI’s potential, many are discovering that its high operational costs and growing regulatory scrutiny require strategic reassessment. We’re witnessing a significant shift toward hybrid infrastructure models, combining cloud, edge and on-premises solutions as leaders try to balance innovation with compliance. The era of AI experimentation is maturing into more real-world implementation, with businesses focusing on specialized AI tools and edge computing solutions to drive efficiency while managing costs. Additionally, the industry faces a challenge in developing the next generation of developers who must learn to incorporate AI tools without becoming overly dependent on them.
If businesses don’t act on AI, regulation will act for them
AI regulation has been on everyone’s mind in 2024, with governments around the world debating legislation. It’s been two years since ChatGPT came out, and there are still many questions around how exactly AI should be regulated. We have seen much back and forth in regulations between The Cloud Act, Privacy Shield program and Schrems 1 and 2. These were attempts at regulating user data transfer and ownership. The future of AI is based on user data and as such, independently of any AI regulation, there will be a need to sort out data regulations too.
How AI is used in regulated industries such as healthcare will define how strict or loose regulations are in other sectors. For example, if some industries or countries completely block AI from certain spaces, then companies will be forced to create ecosystems that ensure AI never touches those areas.
Transparency will also be key. It will be essential in spotting and addressing AI that can be too realistic, such as deepfakes, and how that affects society, politics, privacy and creativity. We need to understand how to identify deepfakes and prevent their creation, and distinguish between AI-generated art and art created by real people who take pride in their work. That transparency around AI will be particularly important on social media and for content targeted at younger audiences. More than anything, we will need to ensure AI-generated content is not used for cyberbullying and harassment, fake news or anything that would damage the fabric of society. As companies mature their use of AI, and more use cases come into production, revenue attribution models will be discussed in various industries.
New industries will master the basics of AI, unlocking unexpected innovation
Everyone is waiting for the big thing in AI – something that breaks the mold of chatbots, semantic search and information retrieval. Since many people have been amazed by the degree of AI anthropomorphism that ChatGPT gave us, we are awaiting a revolution that is similarly amazing. But what if this is a silent revolution already happening? It’s hard to quantify the effect of tools and improved productivity that basic AI tooling already provides.
For this innovation to happen, we need to get the basics right – security, privacy and ensuring AI fits into existing systems without causing chaos. Many organizations are still figuring out the operational details and testing the waters with chatbots and semantic search. When businesses get past the mundane elements and figure out how everything works together, we’ll start seeing the real breakthroughs. Most developers are being cautious and waiting to see how new AI innovations are harnessed in other industries before fully committing to it themselves. They know that an AI tool that doesn’t work simply won’t be worth the cost or effort. Developers will need opinionated AI stacks or AI platforms to simplify choices and alleviate all the integration code needed to make sure these new technologies can work together.
Additionally, one burgeoning area to watch is AI-powered smart glasses. Google Glass may have been ahead of its time, but new collaborations, such as between Meta and Ray-Ban, are bringing the technology back into the field. Adding AI personal assistants into those devices could unlock a lot of interesting functionality, and perhaps even help people with disabilities or learning difficulties. It wouldn’t be surprising if we see rapid evolution of this in the next year. The future of AI is more than ever at the edge as it offers contextualized, improved user experiences.
Developers will find a happy medium for AI-reliance
We’re already seeing consultancies having trouble selling development days. Some are pivoting and selling AI code assistant training instead of traditional development practices. Over the next couple of years, developers will need to make sure that more junior workers entering their teams are not stuck in the trap of being “AI-native.” Those that have an overreliance on AI, such as heavily leaning on copilot AI tools to double check their work or even do parts of it for them, could see a huge gap between junior and senior developers who have worked in the industry since before the AI boom. Senior developers will know if the AI hallucinated, how accurate the answer is and possibly re-prompt the assistant to get to a working solution, while junior developers may not fully understand the reasoning behind that solution if an AI tool suggested it.
Companies should invest in helping junior developers grow without being overreliant on AI tools if they want to become senior developers in the future. They will need to understand, for example, why a chatbot suggested a particular solution so that they know how to solve the problem themselves. Companies will need to invest more in education going forward, especially through hands-on experience. But the beauty of AI is that it can also provide training. Developers might soon start to use AI to offer hands-on, immersive and real-time learning experiences – much like learning a language by immersing yourself in the culture. It can guide developers through coding processes and help them learn on the job. The value of AI might be in how it allows us to work better with what we already have rather than to generate something completely new.
AI costs and rug pulling
While industries are still in the early AI adoption stage, it is proving to be one of the most expensive technologies to run. Between the price and rarity of the hardware and the associated electricity consumption, we are currently in a time where only a select few can do exactly what they want with LLMs. Many companies are currently countering this by offering generous free offers, but that can only last so long. Companies will need to think about how they mitigate impacts when the era of AI freebies comes to an end. As the industry matures and developer knowledge of AI stacks spreads, the composite nature of architecture will drive the adoption of small language models for specific, dedicated tasks and possibly diminish the pressure on LLMs availability.
AI driven cloud exodus
Companies will increasingly adopt hybrid models that combine on-premises, edge and cloud resources to meet specific latency, security and data sovereignty requirements. Edge providers, especially those with global reach, will play a large role in enabling compute power closer to users. Demand for AI and machine learning will drive growth in specialized hardware like GPUs, TPUs and quantum processors. This trend may lead to new partnerships or even competition between hardware startups and traditional CSPs, especially as hardware costs rise and AI workloads require unique infrastructure.
Mature open-source, cloud-native software will increasingly enable companies to deploy cloud-like environments on their own infrastructure. This trend empowers businesses to move away from traditional CSPs, cutting costs and enhancing control. The Cloud Native Computing Foundation will see its importance grow along its user base.
There will also be an increased focus on data privacy and governance. Data regulations and privacy concerns will prompt more companies to explore self-hosting or to rely on local, region-specific providers. This shift, driven by political and regulatory pressures, may lead to more fragmented cloud services that align with local compliance needs.
The road ahead
Moving forward, AI project success needs a more nuanced approach. The push toward AI regulation will require heightened transparency, particularly in addressing deepfakes and AI-generated content. Companies will need to rationalize their cloud spending, potentially leading to an exodus from traditional cloud service providers in favor of hybrid solutions. The rise of specialized hardware, edge computing and small language models will offer new pathways for cost-effective AI deployment. By thoughtfully integrating these emerging technologies while maintaining robust security and privacy standards, organizations can transform these challenges into opportunities. Those who successfully balance innovation with governance, empower their developers with proper AI training and remain agile in their approach will be best positioned to thrive in 2025 and beyond.
##
ABOUT THE AUTHOR
Laurent Doguin is the Director of Developer Relations and Strategy at Couchbase, provider of the developer data platform architected for critical applications in our AI world. Previously, he was a Developer Advocate at Couchbase where he focused on helping Java developers. Prior to Couchbase, Laurent held developer roles at Clever Cloud and Nuxeo.
