Open-Source AI Models Gain Ground as Enterprises Seek Lower-Cost Alternatives, Citi Says

Open-source artificial intelligence (AI) models are rapidly gaining traction as businesses search for cost-effective alternatives to proprietary AI systems, according to a new research note from Citi. The shift comes as tighter regulations limit access to some advanced AI models while several leading developers slow the release of next-generation proprietary systems, encouraging enterprises to explore open-weight and open-source AI solutions.

Citi said enterprise demand for open-source AI models has accelerated as organizations balance rising development costs with growing restrictions surrounding access to premium AI platforms. Businesses are increasingly adopting open AI models for specialized tasks, particularly as the performance gap between proprietary and open-source systems continues to narrow.

A major catalyst behind this trend is the launch of Z.ai’s GLM-5.2 model, which Citi believes has significantly improved the competitiveness of open-source AI. The bank noted that several open models developed by companies such as Z.ai, DeepSeek, MiniMax, and Moonshot AI now perform much closer to proprietary AI systems across both intelligence and cost-efficiency benchmarks.

Industry adoption metrics further reinforce the momentum behind open-source AI. Citi highlighted that AI startup Cohere has tripled its projected annual recurring revenue target for 2027, reflecting growing enterprise demand. Meanwhile, OpenRouter reported that open-source models accounted for 65% of its processed AI tokens in June, up sharply from 34% in January. Fireworks also experienced substantial growth, with its open-source token volume doubling from April to June to approximately 30 trillion tokens.

Investor confidence in open AI infrastructure is also strengthening. Citi pointed to several high-profile funding rounds and acquisitions designed to reduce AI deployment costs and improve infrastructure flexibility. These include Baseten’s $1.5 billion funding round, Upscale AI’s $190 million capital raise, and Qualcomm’s planned acquisition of Modular, all of which underscore increasing support for open software and hardware ecosystems.

Despite this rapid expansion, Citi cautioned that significant challenges remain. The AI industry continues to face an intense shortage of skilled researchers and engineers, making talent acquisition a critical competitive advantage. Infrastructure limitations also present growing concerns, particularly as opposition to new data-center construction gains momentum across the United States.

According to Citi, more than 300 data-center bans or moratoriums have been enacted by local governments since 2023, with over 275 introduced since the beginning of this year alone. These restrictions could slow the expansion of computing capacity needed to support increasingly sophisticated AI models.

While open-source AI is expected to continue capturing market share, Citi believes established AI leaders will maintain a competitive edge thanks to their deep talent pools, extensive computing resources, and existing infrastructure. Even so, the bank expects open-source AI models to play an increasingly important role in enterprise AI adoption as organizations prioritize flexibility, affordability, and performance in their long-term AI strategies.