The profit game of Anthropic's trillion-dollar IPO sprint: Aggressively advancing AI pharmaceuticals and demanding profits from Amazon

Anthropic is currently extending AI commercialization into the life sciences sector. On Tuesday, it launched its dedicated AI product for scientists and pharmaceutical companies, Claude Science, listing 3D protein structure rendering and drug discovery as core application scenarios, aiming to open up new revenue streams for its planned IPO. This is Anthropic’s first product tailored for scientists and will be available to global individual paid users and enterprise subscribers. Eric Kauderer-Abrams, head of Anthropic’s life sciences business, stated in an interview: "We believe that work in science, especially in the fields of life sciences and healthcare, can generate a large-scale positive impact for humanity." In terms of commercial layout, Anthropic has already signed cooperation agreements with several pharmaceutical companies and acquired biotech startup Coefficient Bio in April of this year. Novo Nordisk has applied Claude in drug discovery, clinical documentation, and regulatory filing, while AstraZeneca is using Claude to expand its R&D scale. Anthropic simultaneously released the latest version of its Sonnet model, Sonnet5, on Tuesday, claiming its agent performance is close to its flagship model Opus but at a lower cost. This positioning aims to fill the gap between high-end flagship models and lightweight models, giving enterprise customers options balancing performance and cost. While expanding into new customer groups and upgrading underlying technology, Anthropic is also re-examining the profit margins of existing core collaborations. According to The Information, Anthropic recently renegotiated parts of its cooperation agreement with Amazon, shifting Claude’s pricing model from computing hours to token-based billing. This switch in billing logic is expected to increase Amazon’s costs in using Claude for shopping, programming, and office AI products. Drug discovery becomes the core breakthrough In terms of specific applications, Eric Kauderer-Abrams stated that Claude Science can accelerate the early stages of drug discovery, including molecular design, and revealed that the company’s next focus will be on the clinical stage, exploring improvements in laboratory experiments and potential integration with robotics. Kauderer-Abrams described the launch of Claude Science as a move to narrow the gap between the forefront of technology and practical application. "There remains quite a gap between what most people can currently access and use and what’s already possible today. The main purpose of this product is to narrow that gap as much as possible, allowing all scientists in their fields to fully leverage the maximum potential of existing technology." AI’s importance in pharmaceuticals is continuously rising. Eli Lilly previously invested in Nvidia chips and earlier this year invested in Insilico Medicine, a company focused on AI-driven drug discovery. Sonnet new version released, targeting high performance-cost applications Anthropic simultaneously released the latest version of its Sonnet model, Sonnet5, on Tuesday, claiming its agent performance is close to its strongest model Opus but at a lower cost. This positioning aims to fill the gap between high-end flagship models and lightweight models, providing enterprise customers with a balance of performance and cost. Sonnet 5 is available for all packages starting today: Free and Pro editions use it by default, Max, Team, and Enterprise users can also access it. To test Sonnet 5’s true production capabilities, Anthropic compared two benchmarks: agent search performance with BrowseComp and computer operation ability with OSWorld-Verified. With different levels of compute input, Sonnet 5 shows stable improvements compared to Sonnet 4.6. Opus 4.8 is still the highest precision choice in both tests, but Sonnet 5 offers quite good performance at a much lower cost. Meanwhile, Anthropic's agent product Cowork, geared toward non-technical users, and its programming tool Claude Code have sparked concerns over job replacement risks in fields like software engineering, consulting, and law, putting its commercial expansion under dual scrutiny from the market and regulators. Adjusting billing model, "seeking profit" from core allies While expanding new customer groups and upgrading underlying technology, Anthropic is also re-examining existing core partnership profit margins. According to The Information, Anthropic recently renegotiated with Amazon, switching Claude's pricing from computing hours to token-based billing. This change in billing logic is expected to increase Amazon’s costs for using Claude in shopping, programming, and office AI products. Given Anthropic’s recent rise in inference infrastructure gross margins to over 70%, this pricing adjustment directly reflects its efforts to increase profitability and optimize its cost structure, and more importantly, signals its strong bargaining power to the market. Against the backdrop of sprinting for an IPO, seeking profit from core allies shows Anthropic has shifted from early-stage scale expansion to an intense focus on profit quality and commercial monetization. Biosecurity risks trigger access controls As AI’s boundaries continue to expand, Anthropic faces increasing safety scrutiny while accelerating commercialization. CEO Dario Amodei pointed out on Tuesday that, compared with cybersecurity, society still lacks awareness of AI-related risks in biology. He and other AI developers have warned that bad actors could use related models to create new pathogens or biological weapons. Amodei stated that only individuals or institutions whose credentials have been verified should be allowed to use powerful biological AI models with potentially dangerous application scenarios. "In pharma companies, people are always handling dangerous biological materials, and they have their own operating procedures. We can learn from those processes—perhaps we shouldn’t open access to anyone before verifying their identity." Notably, Claude Science currently runs on existing models and does not possess advanced biological functions. Anthropic plans to establish a trusted access program for biological applications, modeled after practices in cybersecurity. IPO prospects stir market expectations Anthropic is currently at a crucial pre-IPO window. The company may go public as soon as this year, with a potential valuation exceeding $1 trillion. It completed a round of financing last month, reaching $900 billion in pre-closing valuation. Competitively, OpenAI has also made scientific applications a strategic focus. In April, it launched GPT-Rosalind, an advanced inference model for biological research, drug discovery, and translational medicine, focusing on turning medical research results into clinical solutions. The launch of Claude Science marks an essential step for Anthropic in transforming its AI capabilities into sustainable enterprise revenue. However, how to balance accelerated commercialization and control of biological safety risks will be a core challenge it cannot avoid before entering the capital market. Risk Disclaimer The market involves risk, and investment requires caution. This article does not constitute personal investment advice, nor does it take into account any individual user's specific investment goals, financial situation, or needs. Users should consider whether any opinions, views, or conclusions in the article suit their particular circumstances. If you invest based on this, you are responsible for the outcome.