Hope you all enjoyed the First day of Summer, the longest day of the year! Those of you who are in school, enjoy your break. Those who aren’t, I hope you still get a vacation.
This month, I read a lot of stuff similar to what you have read in the past. A lot of reports suggesting that serverless cloud is growing and expected to hit unprecedented numbers by 2030.
This month was a bit lighter on the news side but I did want to take the time to write you all.
A Decade of Serverless
AWS opened the floodgates for serverless with the creation of Lambda in November 2014 (GAing in April 2015). The creation of this Functions-as-a-Service (FaaS) framework resulted in the first “serverless” platform. Now technically, one could argue that PaaS was serverless too but most PaaS platforms had some form of infrastructure that you had to manage.
According to Technology Magazine, the FaaS industry is a $30bn global market now. This shouldn’t come as a surprise as many companies are trying to find ways to reduce their overhead. Of course, just like with anything, there are pros and cons to functions.
Many of the cons such as vendor lock-in can be attributed to the opinionated natures of FaaS runtimes. But with containers, you can potentially bypass this. This could be why Cloudflare launched containers in beta this month.
Serverless continues to grow and innovate such as the shift from functions to containers and WASM in serverless circles. Functions will continue to exist but having more options will reduce friction seen in serverless.
Even heavily regulated industries like healthcare are starting to see the truth of serverless.
As a side note, I found an interesteing article where someone built an app for under $10 a month with serverless.
A Serverless Data Lake?
Yes. Now that Databricks has acquired Neon they are looking to introduce a serverless data lake called Lakebase. Now admittedly, data storage and analytics aren’t my strong suit so I had to dig into this.
To level-set, a data lake is essentially a massive database for storing data in it’s raw format. Once the data is collected and stored, data scientists and analysts can then clean the data via an ETL pipeline or something similar so that they can try to gather valuable insights.
With Neon and their serverless Postgres offering, Databricks is looking to offer a data lake that is serverless in nature. That is, the storage of the data and the processing of the data are separate.
Per Databricks, this will allow businesses to keep their data with their analytics platform. This shows how serverless continues to innovate and how it is WAY more than just functions.
If you want to learn more, you can review this blog post!
Cloud Run GPUs @ Google Cloud get a few Upgrades!
Google Cloud Run GPUs are now GA for Services! This is true for the NVIDIA L4 model of GPU. But also…. GPUs for Cloud Run Jobs are in Public Preview (Google Cloud terms for “beta”). What does this mean for serverless an AI? Well with services (Cloud Run instances needing an HTTP/S endpoint) you can create endpoints using tools like Ollama or vLLM in a container to inference with LLMs such as Gemma or DeepSeek. These GPUs are on demand and will only run when a service is running (e.g. someone hits the endpoint).
Jobs are used for batch processes usually. So say you wanted to use an LLM to read documents in a storage bucket and store the results in a database. A Cloud Run Job would be perfect for this use case. Now it can use GPUs as well!
Closing Thoughts
Serverless is the best platform to start building applications in the cloud. There will always be a need for VMs, container orchestrators like Kubernetes and other application platforms but serverless should always be the starting point.
We are seeing major companies doing big things with serverless and helping us evolve past the notion of simple functions. We see serverless GPUs, Serverless datalakes and more.
I will leave you with this article from InfoWorld that talks about many misconceptions around serverless and how serverless is going to be an incredibly powerful tool in the age of AI.
Until next time!
Update
In order to provide more value to my readers, I am going to be making some format changes that will be visible in August 2025. You may see some samples in the July edition. I haven’t decided exactly what I want to do yet but I will make a formal announcement of the formal changes and release schedule in the July 2025 edition.
—Photo courtesy Nitin Dhumal on Pexels—