There’s nothing better than some hot and spicy takes to warm up the cold of a Nordic new year. What do we reckon is going to unfold in the cloud ecosystem in 2025? Come toast your hands by fire and share in our 2025 tech predictions. Whether they have you nodding in agreement, or shaking your head furiously – they’re sure to take the chill off!
In a nutshell, we predict:
- – More Rust in Kubernetes
- – Increased experimentation on Serverless
- – An uptick in WASM adoption
- – Stronger focus on security
- – Bridging the gap between IaC and SW developers
- – AI/LLMs maturing beyond GPUs
- – Kubernetes (hopefully) moving towards being more user-friendly
More Rust in Kubernetes
We’ve noticed an increasing trend of operators being written in Rust, and UpCloud Lead Developer Nate Ham is “here for it.”
“With topics like AI and LLMs being all the rage, there will hopefully be an increasing interest in topics like ‘formal verification’ and writing more rigorous tooling. With Rust’s push towards safety, I would really like to see what a bigger buy-in from the community would look like.”
On the environmental side, Rust has proved itself a highly efficient language. The brilliant research in Energy Efficiency across Programming Languages How – Do Energy, Time, and Memory Relate? demonstrated that wide adoption of Rust and C could have a profound impact – reducing compute energy consumption by 50%. But unlike C, Rust has memory safety, and, let’s face it, is a whole lot easier to work with.
A Great Experimentation on Serverless
Will it, won’t it? For a good two years the ecosystem has been wondering whether serverless – running event-orientated back-end programs and processes in the cloud – has proverbial legs. Knative was accepted into the CNCF landscape as an incubating project in 2022, and the Linux Foundation even released a dedicated serverless on Kubernetes certification in 2024.
But what really got us talking was a presentation at the recent KubeCon + CloudNativeCon North America: Why serverless is trending again, by Fermyon CEO Matt Butcher and Akamai CTO Jay Jenkins.
Serverless workloads are essentially cost-prohibitive right now – moving where applications are deployed is insanely expensive because of Egress fees. But if this financial constraint wasn’t a constraint anymore, you literally could spin up a deployment where and when it was needed, and then shut it down.
“Egress costs are way too high, everybody knows it, everybody complains about it, but there’s very little that’s been done about it so far so that’s something that absolutely has to be fixed.” – Jay Jenkins
We’ve been working on some PoCs and have found interesting opportunities to integrate new services to the UpCloud platform. We’re still in the very early stages, but stay tuned for more later down the line. In the meantime, since we’ve removed Egress fees, we’re interested to see how your Serverless workloads run on UpCloud. Let us know if you’re experimenting and let’s see how we can help!
WASM 2.0
Over the past decade, WASM has been quietly making an impressive impact in the cloud native ecosystem – running in everything from Google Earth to Amazon Prime Video. But the release of WASM components over the summer of 2024 looks set to move the dial towards much greater, and wider, adoption in 2025.
Will WASM replace containers wholesale? Probably not, but components offer an incredible application for running workloads in IoT devices or at the Edge – anywhere that containers don’t reach.
On a side note, the global WASM community is outstanding and there’s some great talks in the WebAssembly developer room at FOSDEM . If you’re also attending FOSDEM in February, come have a few beers with us at the UpCloud Happy Hours on Friday 31st! RSVP here.
Stronger Focus on Security
Security, security, security! According to the CNCF’s 2023 Annual Survey (its most recent), security is the top challenge for organisations looking to adopt containers, cited by 44% of respondents. This is in line with the 2022 survey results where security came in as the second most important concern, only after lack of training.
It’s interesting to also note that last year concerns appeared to shift towards security and away from lack of training as organisations’ level of maturity with containers grows. While lack of training is consistently a more important challenge than security for companies with low adoption of containers (46% cite lack of training vs 33% cite security) security is a consistently more important challenge than lack of training for companies with high adoption (43% cite security vs 22% cite lack of training).
This suggests that as the industry’s overall adoption grows, the challenge of security will grow in step. This can also be observed by the rise in popularity of projects – like Kubescape and Kube-bench – which attempt to address this challenge, and by how popular the “kubernetes security” search term has become over time. Here’s a comparison against other popular kubernetes search terms.
Bridging the Gap Between IaC and Software Engineers
As cloud adoption continues to steadily grow, and more organisations experiment with hybrid and multi-cloud approaches, the demand for skills that allow companies to lift and shift their infrastructure from one place to another is on the rise. It’s not surprising that “multi-cloud” and Terraform are considered by some to be amongst the most sought-after tech skills in 2024. The web is filled with literature on how increasingly hard it is for organisations to hire talent with Cloud Infrastructure, DevOps and IaC skills.
One potential answer to this problem is whether organisations can upskill and/or reskill software engineers to fill the void. This is, of course, no easy task. The tools traditionally used for IaC are very different from those used by software engineers. However, some projects propose an alternative approach – allowing software engineers to define infrastructure in their programming language of choice.
Will this approach eventually become mainstream? The StackOverflow Annual Surveys from the past five years indicate that it might. Amongst IaC tools, Terraform is king, followed closely by Ansible. And while Chef and Puppet adoption have declined rapidly, Pulumi is on the up – overtaking Chef in 2024, which is a notable feat considering that Pulumi didn’t even rank back in 2020!
HashiCorp seems well aware of Pulumi’s allure. Just two years after Pulumi launched, it released CDKTF, which aims to address some of the pain points around Terraform, like limited reusability and the learning curve for HCL.
As for us, we’re fans of both Terraform and Pulumi. We updated our Terraform provider recently, and we’re actively working on our UpCloud Pulumi Provider. We’d love your feedback if you’re using either of these open source tools!
AI/LLMs maturing beyond GPUs
AI was the undisputed buzzword of 2024, with more than 100 models released throughout the year. (Check out this amazing overview from Hugging Face!) But now AI/ML workloads are moving into production, and LLMs continue to become more sophisticated, we predict a shift in focus on the engineering side from: “how big do our GPUs need to be?” to “how do we run AI workloads effectively?”
OpenAI, the force behind ChatGPT, leveraged cloud computing early on – scaling Kubernetes to 7,500 nodes to create a scalable infrastructure for large models like GPT-3, and for rapid small-scale iterative research. In 2025, we reckon cloud will be the de facto choice for optimising AI/ML performance across diverse infrastructure needs, rather than relying solely on raw computational power. After all, general purpose cloud computing is fantastic for running tasks like data preprocessing, lightweight inference, and distributed analytics – where CPUs can handle workloads efficiently. And as AI/ML workloads become more cost-effective, we may start to see a growth in CPU-based models (using vector databases) instead of all requests running against an inference.
We’d love to know if you’re running AI/ML workloads on UpCloud, and whether you have feedback, so we can keep improving our offerings!
The next 10 years of K8s
Kubernetes turned 10 last summer! Globally, an estimated seven million developers work with Kubernetes, and there is no doubt that it has had a profound impact on the cloud native landscape.
But, let’s face it, Kubernetes is still complex. Even with so many tools that have sprung up to make our lives easier (looking at you, GKE Autopilot!), as engineers, it would be great if we could lift some of the operational burden from end users. We’d love more operators (in Rust, please) and abstractions. Kubernetes’ core maintainers have stated:
“The community is prioritizing changes that both improve the user experiences, and enhance the sustainability of the project. The world of application development continues to evolve, and Kubernetes is poised to change along with it.”
At UpCloud, we’re also planning to keep simplifying the ways users can work with Kubernetes, and we’ll share more of this in our roadmap later in the month. We really want to continue lowering the barrier for adoption for SMEs, solopreneurs, and small dev teams, to unlock the benefits of containerization.
What do you think?
These are our hot takes for 2025, but we’d love to hear yours! What do you think? Connect with us on any of our social channels (links below) and join the debate.
