Corporate America is getting hooked on open-source AI

Corporate America is increasingly adopting open-source artificial intelligence (AI) models over closed alternatives, driven by rising costs and the desire for more flexibility. AT&T, for instance, has shifted from using closed models from Anthropic and OpenAI to open models, which can be downloaded and customized without fees or approvals.

This move has already slashed the company's AI expenses by up to 80% compared to earlier this year, with usage of open models now accounting for 40% of its total AI consumption. Such a trend is evident across many corporations, including Airbnb and Deloitte, with open models now representing 58% of AI usage in the U.S., up from just 10% a year ago.

These models come in two forms: open-source, with publicly available code, and open-weights, with publicly available numerical calculations determining the model's reasoning. Many popular open models originate from Chinese companies like Moonshot AI, DeepSeek, and Alibaba, offering powerful alternatives to closed models from OpenAI and Anthropic at a fraction of the cost.

Experts argue that open models provide a more cost-effective solution, likening them to affordable, reliable cars, as opposed to expensive sports cars. The ascent of open-source AI models also raises questions about AI regulation and control, as companies like Anthropic and OpenAI prepare for potential IPOs and grapple with the escalating costs of research and development.

However, some tech leaders like Nvidia CEO Jensen Huang and Meta CEO Mark Zuckerberg advocate for open-source technology to foster a more democratic and balanced AI ecosystem.