If users can run powerful models on their own hardware, why pay a monthly subscription? Is local AI going to disrupt the current business models we see today? I feel like the barrier to entry is dropping fast as open-source models improve.
3 answers
It won't kill SaaS, but it will force it to evolve. We will likely see a hybrid approach where the "heavy lifting" or training happens in the cloud, but the daily inference is done locally. Startups that only provide a wrapper for an API will struggle, but those offering sophisticated fine-tuning or unique data integration will still thrive. The value prop is shifting from "access to the model" to "how the model uses your specific data." Open-source projects like Llama and Mistral have proven that the community can produce results that rival closed-source systems, which is the real catalyst for this change.
Does the complexity of setting up local environments still act as a major deterrent for non-technical users compared to simple web apps?
Open source is the real hero here. The speed at which the community optimizes these models for local hardware is absolutely mind-blowing.
I agree with Laura. The community-driven optimization for local hardware is what is actually making this entire shift technologically feasible for most of us.
Joshua, that is the biggest hurdle right now. While tools like Ollama and LM Studio are making it easier, it still isn't "one-click" simple for a marketing manager. Until we have seamless integration into common productivity software, the web-based SaaS model remains the king of convenience for the general public and non-dev teams.