Wells Fargo: The wave of open-source models sweeps through the AI circle, Microsoft and ServiceNow will become the big winners

According to the Zhitong Finance APP, Wells Fargo recently released a research report indicating that the continuous advancement of open-source models is beneficial for large existing platforms such as Microsoft (MSFT.US) and ServiceNow (NOW.US).

The analyst team led by Michael Turrin stated that the development of open-source models is accelerating, with representative projects including Moonshot AI's Kimi K3, Z.ai's GLM 5.2, and DeepSeek's DeepSeek V4. It is expected that the U.S. market will also heat up, with Meta (META.US)'s "Thinking Machines" project and companies like NVIDIA (NVDA.US) also worthy of attention.

The Turrin team pointed out: "An increasing number of market participants advocate embracing open-source models, as evidenced by the open letter issued by NVIDIA on July 24, co-signed by Microsoft, Meta, Palantir, ServiceNow, and IBM. Although the attitude of leading U.S. laboratories is more cautious, overall acceptance of open-source models is positive for the entire ecosystem, effectively reducing intelligence costs and expanding the accessibility and use of AI."

Analysts further explained that the industry is shifting from "competing on the number of tokens" to "competing on value output." Their research results show that open-source models are increasingly favored by users due to lower costs, weaker lock-in effects, and stronger security and controllability.

Customers are increasingly willing to turn to "good enough" open-source models to supplement the expensive cutting-edge models, seeking a more balanced trade-off between cost and performance. This trend indicates that market preferences are gradually shifting towards output quality and actual value, driving cost curve optimization, ultimately benefiting the AI consumer end.

The Turrin team believes that the evolution of open-source models is a positive signal for large existing software vendors, as AI is inherently embedded in workflows. "The decrease in inference costs means that the threshold for obtaining intelligence in familiar mature applications is lower, while the market risk of a 'one model fits all' approach is also reduced. A typical example is Microsoft Copilot, where enhanced intelligent capabilities provide additional value to its already solid security and enterprise positioning."

In addition, the reduction in intelligence costs will also stimulate software development activities, boosting companies such as Datadog (DDOG.US), MongoDB (MDB.US), Atlassian (TEAM.US), and GitLab (GTLB.US), but will accelerate the squeeze on the survival space of small and medium-sized players and niche market participants