Anthropic overtakes OpenAI in the enterprise market: LLM spending reaches $8.4 billion

Claude leads among companies, while OpenAI slips to second place in a multi-billion dollar race.

The new mid-year report from Menlo Ventures, published on July 31, 2025, marks a turning point in enterprise adoption of large language models (LLMs). Enterprise spending on LLM technologies has exceeded $8 billion for the first time, with accelerated growth compared to the previous half-year. But the most significant figure is Anthropic’s overtaking of OpenAI: Claude has become the most used model in enterprise environments, with a 32% market share, while OpenAI falls to 25%. A clear signal that leadership in foundation models is far from settled.

Anthropic captures the enterprise LLM market with a vertical rise in usage share

While OpenAI held over 50% of the enterprise LLM market in 2023, Anthropic now leads the sector with 32% of production workloads. This rapid rise is largely tied to the performance of Claude 3 models, particularly Claude Sonnet 3.5, perceived as more robust, less hallucinatory, and more predictable in mission-critical tasks. Companies seem to favor it not just for its reasoning capabilities, but also for its architectural transparency, data governance, and safety-first design.

OpenAI, with 25% of the market, remains strongly present but has lost dominance compared to 2023. Google, thanks to the efficiency of Gemini 2.5, stands at 20% of workloads, followed by Meta (with Llama) at 9%, and DeepSeek at 1%.

Inference overtakes training and dominates enterprise workloads

Another significant trend is the nature of the workloads. According to Menlo, 74% of AI startups and 49% of large enterprises now use models almost exclusively for inference, a sharp increase from 2024. Custom training and fine-tuning, which were key elements in internal AI strategies until recently, are now marginal. Only 9% of teams report investing in internally trained models.

Driving this shift is the evolution of APIs and general-purpose models, which are now more powerful and frequently updated by providers: 66% of teams have adopted the latest version from their vendor, only 11% have switched providers, and 23% remain unchanged.

Closed-source models dominate: open-source loses ground in the enterprise

The initial promise of open-source models, which in 2023 accounted for up to 20% of enterprise workloads, appears to be fading. The report indicates that today only 13% of enterprise use cases rely on open models. This is mainly due to a growing performance gap, as well as challenges in scalability, privacy management, and integration complexity.

Meanwhile, closed-source models have solidified their leadership, powering 87% of enterprise applications. Companies increasingly choose turnkey solutions that offer technical support, regulatory compliance, and long-term technology roadmaps.

Long-horizon agents and advanced automation: the next $10 billion frontier

Menlo Ventures highlights that the current phase is only the beginning. Companies are now aiming for the next wave of AI innovation: long-horizon agents. These are systems capable of executing multi-step autonomous tasks, such as end-to-end process management, complex code generation, technical document synthesis, or multi-source strategic analysis.

These agents require advanced orchestration, persistent interaction with external tools (databases, browsers, IDEs), and models with extended contextual memory. According to Menlo’s estimates, this category could represent the next $10 billion enterprise AI platform, with increasingly specialized models and modular architectures.

Market implications: from model leadership to ecosystem warfare

Anthropic’s rapid rise demonstrates that success in the LLM market depends not only on model quality but on the ability to build an ecosystem of tools, API compatibility, compliance, and competitive pricing. Claude has integrated well into existing toolchains and earned trust among developers, legal teams, and decision-makers.

Meanwhile, Google has positioned itself as a third force, thanks to lower token costs and tighter integration with the Workspace ecosystem. OpenAI, while still strong in consumer markets and research, will need to reassess its enterprise positioning to regain momentum.

A market flooded with investment but still seeking profitability

Even though enterprise LLM spending has reached $8.4 billion, the market remains marked by a wide gap between investment levels and real revenue. Major players have raised over $90 billion in equity and debt, but total revenues remain relatively modest.

This imbalance makes business model efficiency a critical issue. Companies that offer scalable, sustainable, and compliant AI will gain a strategic edge. Those relying only on hype risk being trapped in a spiral of high burn rates and growing financial pressure.

____________

FOLLOW US FOR MORE.

DATA STUDIOS

datastudios.org