Nearly half of Gen AI adopters want it open source – here’s why

NAPA, Calif. – At the Linux Foundation Member Summit, Linux Foundation Research announced a new report that dives deep into the sometimes awkward relationship between open source and artificial intelligence (AI).

Called “Shaping the Future of Generative AI,” the report — produced by Linux Foundation AI & Data and the Cloud Native Computing Foundation (CNCF) — reveals that open-source software is crucial in shaping the future of generative AI. That much we already knew! Indeed, AI can’t exist without open-source programs such as PyTorch and TensorFlow. What this report, which surveyed 316 AI professionals, brings to the table is an analysis of open source and Gen AI’s significant new trends.

Also: The best open-source AI models: All your free-to-use options explained

Would you be surprised to know that 94% of organizations currently use Gen AI, with 42% reporting high or very high adoption rates? 

On average, 41% of those using Gen AI report that their organization’s code infrastructure supporting AI is open source. This percentage rises to 47% for high adopters of Gen AI. It’s far higher than that when you look behind the curtain and see how machine learning generates large language models (LLM) in the first place.  

Why? Nearly half (46%) of organizations cited cost efficiency for choosing open-source Gen AI solutions. Open-source tools reduce both upfront licensing costs and long-term dependence on proprietary vendors. For example, AI application frameworks like LangChain and LlamaIndex enable organizations to develop and deploy AI models at a fraction of the cost of proprietary solutions.

That can amount to some serious savings. BloombergGPT, Bloomberg’s 50-billion parameter finance LLM, cost around $3 million to build.  

Another cost savings can come from the rise of using cloud native technologies to run scalable GenAI programs. Kubernetes, for instance, has emerged as a key enabler for orchestrating scalable Gen AI workloads, with 50% of organizations using it to host some or all of their Gen AI inferencing workloads. 

In addition, just over half, 52% of respondents, said transparency and trust in the source code were primary reasons for choosing open-source Gen AI solutions. This allows organizations to verify model behavior, identify potential biases, and ensure regulatory compliance.

That last part is very important, as governments get more involved in regulating AI. It’s also far from easy. As Brian Warner, Fidelity Investments director and open source project office architect, said in his speech on licensing, AI licenses alone are a complicated mess. “There’s no shortage of LLMs which are claiming to be open source, but in reality, there are some wonky licenses that aren’t even close to the open-source definition.” This makes using AI programs legally a real headache.

Also: We have an official open-source AI definition now, but the fight is far from over

Licensing worries and all, open-source AI will only continue to grow. Seventy-one percent of respondents report that open source positively influences their decision-making, and 73% of organizations expect to increase their use of open-source Gen AI tools over the next two years.

Hilary Carter, senior vice president of research at the Linux Foundation, emphasized the importance of openness in AI’s future, stating, “82% of organizations believe open source AI is critical for ensuring a positive AI future, and 83% agree that AI needs to be increasingly open to foster trust, collaboration, and innovation.”

As the details of how open source is used in AI continue to evolve, the report concluded that open source will remain indispensable. Both the Linux Foundation and I recommend that organizations prioritize open source in their Gen AI strategies to remain competitive and aligned with industry trends. Like all software, open source is essential for further AI advances.