Fermyon Serverless: Solving Key Market Problems According to Leading Analyst Firm
Since Fermyon’s founding we have been focused on enabling software developers to do what they do best: create amazing application software while minimizing all of the other painful duties that often come with the territory (setting things up for Operations, or creating local copies of resources, or administrative tasks required before writing a line of application code).
We care so much about this core philosophy that our foundational user story was (and still is) “from blinking cursor to deployed application in 66 seconds.” We didn’t want to stop at a basic microservice or webhook-style application that could be deployed that quickly, we wanted (and have succeeded) at enabling full-stack applications to benefit from that same guiding user story by releasing “no-ops” key value storage and “no-ops” SQL databases. Even more recently, we released “no-ops” AI inferencing in private beta.
And in so doing, realized that what we’ve basically created is “serverless the way it ought to be” rather than serverless of today which seems to impose more burden on developers than relieving burden — and even bleeding over to burden on operations teams who were supposed to be “out of the picture” when serverless was first gaining steam.
We were recently pleased to have some of these key features evaluated in a recent IDC Link titled “Fermyon’s WebAssembly PaaS Targets an AI-First World” in which IDC provides their viewpoint and guidance.
According to IDC, a leading provider of global IT research and advice, the serverless market is expressing the following needs:
- Unify the developer experience and reduce context switching;
- Lower the costs of executing cloud-based computing; and
- Reduce operational complexities and operational costs created by developers’ tool choices.
Fermyon Spin and Fermyon Cloud unify the developer’s local development experience while also unifying the developers’ cloud-based operational experience by, essentially, eliminating all of the setup requirements (save for the definition of an SQL schema for the SQL service) associated with moving a developers’ application to the cloud for high-performance execution. With AI inferencing Fermyon has gone one step further by actually making the cloud compute resource (the GPU) available to the local application. This is exactly the kind of unification of the developer’s experience IDC is referring to.
Because Fermyon utilizes WebAssembly as its under-the-hood secret sauce, the underlying portability (works on all GPUs) and performance (AI inferencing cold-start times of <51 milliseconds) was an ideal match for AI inferencing as it now enables developers to get the same pay-per-use pricing characteristics while also having industry leading cold-start performance.
We also believe, as do investors and market observers with whom we’ve spent time, that AI inferencing is poised to become “just another resource” in the software developers’ toolkit: in the same way that SQL paved the way for developers to easily access data repositories known as databases, AI inferencing is paving the way for developers to access data repositories known as AI Models. In other words, that we aren’t headed for a world of “bespoke AI developers” who impart their wizardry on the rest of us but, instead, to that world where all developers simply have access to the proper resources to bring generative AI to their full stack applications. Full stack applications, in turn, require developers to have access to the SQL and Key Value data sources that can, for example, store and quickly retrieve prompts, or data embellishments that “fine tune” AI models, or vectors that contain stored information about more complex bits of information that are fed into AI models via inferencing.
IDC concurs that software developers far and wide will need to add these basic AI building blocks to their repertoires.
We were honored to read what IDC wrote in its IDC Link about the characteristics of the market and about Fermyon’s fit within those characteristics. We hope you’ll take the 5 minutes it takes to read their research opinion in full.