Open-source AI models gain ground as companies cut costs

Cost concerns drive AI model selection

Enterprises are increasingly choosing open-source AI models over expensive proprietary alternatives as mounting bills force a more pragmatic approach to artificial intelligence deployment, according to Werner Vogels, Amazon's chief technology officer.

Speaking at the UN's AI for Good summit, Vogels described a clear trend: "We see a shift happening between the cheaper open source models and the bigger expensive models." The migration reflects growing anxiety about AI costs after several high-profile budget overruns made headlines. Uber reportedly exhausted its entire 2026 AI budget in just four months, while another company burned through half a billion dollars in a single month after failing to cap employee AI usage.

The economics are straightforward. Frontier models from OpenAI, Anthropic, and Google DeepMind bill by the token—roughly a word and a half of text—and those charges accumulate quickly at scale. Open-source models, sometimes called "open weight" models, can be downloaded for free, though users must pay for their own cloud computing infrastructure. Even with infrastructure costs factored in, this approach often proves cheaper than running the most advanced proprietary models.

Why it matters

This shift signals a maturation in enterprise AI adoption. After an initial wave of experimentation driven by hype around large language models, organizations are entering a more disciplined phase focused on return on investment. The question is no longer just what AI can do, but what it costs to deploy and maintain over time—a calculation that's reshaping architecture decisions across industries.

Architecture decisions prioritize efficiency

Vogels emphasized that cost considerations should drive technical choices from the start. "Cost is a very important part of your architecture, you need to take that into account," he said. "Do you really need to have the biggest, highest-end model to solve this? The answer is no, you don't."

The shift also reflects demand for transparency in AI systems, particularly in sensitive sectors like healthcare, government, and humanitarian work. "Transparency becomes extremely important," Vogels noted. "People want to know what is the data that goes into it." In fields serving vulnerable communities, understanding how an AI system was trained and how it makes decisions can be as critical as its performance. "If these people serve vulnerable communities. If they don't trust the system, they won't use it," he said.

Open-source models offer advantages here because developers can inspect and modify code and fine-tune models on their own data. However, even most open-weight model providers don't fully disclose the data used for initial training.

New tools for research access

At the summit, Vogels also introduced a new Amazon open-source AI tool designed to help researchers quickly locate relevant scientific datasets. The system connects the AWS Registry of Open Data—which houses more than 1,100 datasets from organizations including NASA, NOAA, and the NIH—to AI assistants, enabling natural language searches instead of complex catalog navigation. The tool aims to replace processes that previously took hours, lowering barriers for scientists at under-resourced institutions and accelerating research in fields like climate science and public health.

These details were first reported by Fortune.