Invest all of your money in Serverless Startups. Okay, not really. [Disclaimer: Do not take financial advice from me as I am in no way qualified to give it] But Forbes recently listed their top 5 startup opportunities and wouldn’t you know it, Serverless is number 3.
Forbes says “By focusing on improving the developer experience and providing robust serverless solutions, startups can capture a growing market interested in flexible and cost-effective computing options.” I definitely agree on this one.
One thing I have seen happening in the cloud as of late (aside from the proliferation of Generative AI) is cost optimization. Organizations are trying to get their cloud bill under control. Many started their cloud journey just turning things on and throwing things onto the cloud but they are now looking at that bill and trying to find ways to be smart. Serverless can absolutely fulfill this goal.
Boston Consulting Group released a report recently in which they highlight how serverless platforms are growing in popularity among the three hyperscalers (AWS, Azure, and GCP). It is going to be the next generation of IT architecture in the same way that cloud is the current generation and on-prem was the previous.
In this article, I am going to talk about a few things that we should look for between now and January 2026. Some may seem repetitive to previous posts but everything will feel new!
WebAssembly Pods, Coming to a K8s Cluster Near You
Back in May I wrote about the Rise of WASM. To recap, WebAssembly (WASM) is a way to turn code into a binary that can be run natively in a browser. Imagine turning a Rust application into a web application.
Today, WASM is growing in popularity but still has a ways to go to catch up to containers. One of the limitations is that there is no way good way (yet) to orchestrate WASM applications in the same way we see containers being orchestrated. Um actually, you can run WASM on Kubernetes.
All the way back in February 2024, SpinKube was introduced as a Sandbox project in the CNCF. In short extends containerd to allow it to support WASM. With this, you would be able to run WASM apps in Kubernetes along with your container-based apps. We are probably a year or so away from seeing people run WASM applications as Kubernetes Pods.
I would be curious to see how this works in comparison to containers. The argument for WASM is that it has a quicker start time than containers but part of that can be the fact that it runs in browser. By running it in Kubernetes, it’s possible that the start time difference is negligible. It is still an early project in the Sandbox phase but I would recommend keeping an eye on it.
Alternatively, there are also startups and projects looking to design ways to run WebAssembly outside of the browser without Kubernetes. Orca is one that promises to do just that. I think it is important to watch this as WASM will be a major component in Serverless compute.
GenAI meets Serverless Fine-Tuning
In March I talked about the role of Serverless in AI. We are seeing the rise of “Inferencing-as-a-Service” where LLM providers offer serverless workers to interface with the LLM. Microsoft seems to have taken this a step further.
Recently, Microsoft announced a serverless method that allows developers to fine-tune their Phi-3 model. According to TechTarget, fine-tuning is “the process of taking a pretrained machine learning model and further training it on a smaller, targeted data set. The aim of fine-tuning is to maintain the original capabilities of a pretrained model while adapting it to suit more specialized use cases.” This is often an alternative to RAG.
Now to me personally, RAG is a superior method for “customizing” foundational LLMs. It is more secure and can be more cost-efficient. Many people consider fine-tuning to be so 2023. That being said, there are still valid use-cases. Fine-tuning a smaller model, for example, can be more efficient.
Part of the problem is that fine-tuning can be cumbersome. With this new offering from Microsoft, fine-tuning might be easier. With the serverless model, it may also become more cost efficient. I expect to see many LLM providers to offer something similar to this in the future. This will give users a choice on how they want to extend the foundational LLMs to their use-case.
Serverless as a Specialization
In 2022 if you said the word “prompt engineer” no one would know what you were talking about. All of the sudden, GenAI created this new role and specialization. Change in tech is so fast that at times, it creates entirely new job titles.
Admittedly, some of those titles are a bit “fluffy” and it’s debatable as to whether or not they will be relevant in the next 5 years. That being said, these roles can be an indicator on the direction of the industry.
I was reading Analytics Insight and came across an article talking about what to expect in 2025 with regards to tech. Similar to the aforementioned Forbes article, they predict the continued growth of the serverless business. They also predict the rise of the “Serverless Developer”.
This will be a developer who specializing in utilizing serverless platforms and developing applications to run on them. Serverless does remove the infrastructure management from users but there will always be a bit of a learning curve.
This isn’t exactly new. When git became popular, developers had to learn how to best use that. What about when CI/CD and DevOps became a practice? Or when your team decided to adopt Agile Development? Any kind of paradigm shift in how you do your job will require some kind of education to best do this.
As serverless becomes more popular, I expect to see consulting and education firms investing creating programs to help educate developers on serverless platforming best practices. Existing firms will expand their offerings and we may even see startups arise to either help train or further simplify the platform to make the specialization easier.
Closing Thoughts
The Cloud will be Serverless. That’s a given. Many reports show significant growth and it will be a multi-billion dollar industry by 2030. I wouldn’t be shocked if we saw half of the internet running serverless applications in the same way that we see half of the internet running on Kubernetes today.
We will continue to cover these trends here so stay tuned and learn more.
—Photo courtesy Bello Co on Pexels—