Claude’s Profit Surge: $70 Billion Revenue and $17 Billion Cash Flow by 2028

In the AI world of 2025, two distinct narratives are emerging. OpenAI is focused on launching voice, video, and plugin ecosystems to make ChatGPT accessible to everyone, while Anthropic is concentrating on profitability. Recent forecasts indicate that Anthropic could achieve $70 billion in revenue and $17 billion in free cash flow by 2028. Both are AI giants, but one is betting on scale while the other is betting on stability.

AI Giants Diverge: One Pursues Traffic, the Other Profit

In 2022, OpenAI and Anthropic were seen as part of the same camp, both pursuing artificial general intelligence (AGI) and even sharing some researchers. Anthropic’s founder, Dario Amodei, was previously the VP of research at OpenAI.

However, by 2025, the industry landscape has clearly differentiated. OpenAI continues to strengthen consumer products: ChatGPT subscriptions, voice assistants, multimodal inputs, and the GPT Store—all aimed at individual users. In contrast, Anthropic focuses on enterprise clients with offerings like Claude API, Claude for Work, and integrations with Slack and Notion, all centered around B2B.

In November, The Information revealed an astonishing forecast—Anthropic expects to reach $70 billion in revenue and $17 billion in free cash flow by 2028. This year, their goal is to generate $3.8 billion in revenue from enterprise APIs, with gross margins projected to soar to 50%.

During the same period, OpenAI’s annualized revenue is around $10 billion, but high computational costs persist. More subscriptions have not significantly improved their profitability structure. OpenAI opts for rapid user penetration, expanding through brand and product ecosystems, whereas Anthropic chooses stable enterprise contracts, trading cash flow for long-term survival.

Different Business Models: The Divide Between Traffic and Profit

OpenAI’s strategy is platformization. From ChatGPT Plus and Team plans to the GPT Store, it sells experiences, traffic, and ecosystems. The more users register, the more valuable it becomes. The subscription model is the most direct way to monetize traffic, but it is also one of the most costly.

Every interaction with ChatGPT incurs GPU computation costs. High inference costs trap OpenAI in a paradox: the more users, the greater the losses. According to The Information, OpenAI’s annualized revenue is about $10 billion in 2025, but its computational expenses remain substantial, and a true profitability milestone has yet to emerge.

If OpenAI is like a supermarket, then Anthropic is akin to a supplier. It has built Claude 3 as the backbone of enterprise AI, providing APIs and customized services. Platforms like Notion, Slack, and Zoom utilize the Claude model, generating stable revenue from enterprise contracts.

Internally, Anthropic predicts $3.8 billion in revenue for 2025, with a target of $70 billion by 2028 and free cash flow rising from -94% to 77% gross margin.

This marks the clearest divide in the current AI industry: OpenAI sells “entry rights” to bring as many people into the AI world as possible, while Anthropic sells “certainty,” ensuring enterprises are willing to pay for stability, safety, and control. One thrives on user growth, while the other focuses on cash flow.

Shared Origins, Divergent Paths: The Fork in Technology Ideals and Business Realities

From a “bloodline” perspective, OpenAI and Anthropic are like sibling rivals. They both believe that large language models are the pathway to general intelligence, utilizing transformer architectures and reinforcement learning with human feedback (RLHF) to train their models.

However, as AI enters the commercial phase, their beliefs begin to diverge. OpenAI’s focus is on multimodality, aiming to make AI capable of hearing, speaking, reading, and writing. This leads to the introduction of image generation, voice dialogue, video comprehension, and coding assistance. In this framework, the model’s goal is to maximize capabilities, prioritizing stunning experiences even at high costs.

Conversely, Anthropic emphasizes “safety and constraints.” Its proposed “Constitutional AI” stresses that models should adhere to a clear set of behavioral guidelines rather than relying on human oversight. This design is more suitable for enterprise use, reducing output risks and enhancing compliance.

From Claude 2 to Claude 3, each model generation optimizes for reliability, controllability, and explainability. This system has helped Anthropic build a reputation in the enterprise market, providing confidence in its $70 billion forecast.

Two technologies have ultimately shaped two brands: OpenAI embodies future imagination—creation, expression, generation; Anthropic represents the order of reality—compliance, stability, risk control. One asks what AI can do, while the other asks what AI should do. These two questions are defining the worldview of AI’s next phase.

The Future of AI: Is It About Scale or Cash Flow?

As AI companies begin to forecast cash flows, the industry’s character has shifted. In the past two years, capital chased breakthroughs in model capabilities; now, it is concerned with whether business models can sustain hardware bills.

The divergence between OpenAI and Anthropic reflects this turning point. OpenAI is pursuing the boundaries of “general intelligence,” while Anthropic is calculating the limits of profit margins. One is pulling back the curtain on imagination, while the other is laying the foundation for business.

In the coming years, the AI industry may witness two rhythms: one continues to expand, centered on products and ecosystems; the other converges, focusing on contracts and cash flow.

Both paths lead to the future, albeit via different routes: OpenAI represents speed and scale, while Anthropic signifies stability and patience. However, the deeper change is that the myth of AI may be rewritten by accounting.

People are no longer just asking how smart the models are; they are starting to ask, “How much can it earn?” When innovation returns to profitability and models enter financial statements, this AI revolution can be considered truly underway. The story of AI has finally shifted from a technological myth back to a business reality. When model parameters are inscribed in profit statements, the future of algorithms also comes with a cost price.

Perhaps this is another form of maturity for intelligence—not becoming more human-like but learning to make money.