Early this morning, OpenAI CEO Sam Altman took to Twitter to showcase the company's earnings, stating, "For our API business alone, we added over $1 billion in ARR (Annual Recurring Revenue) last month." He continued, noting that while most people only see the achievements of ChatGPT, the performance of the API team has been equally astounding.

Altman's move may be aimed at boosting investor confidence. Recently, OpenAI was reported to be planning a fundraising round of $50 billion, with a new valuation estimated to be between $750 billion and $830 billion. Simultaneously, at a Davos forum hosted by The Information's CEO Jessica Lessin, OpenAI CFO Sarah Friar discussed another business opportunity—"value sharing." She proposed that in the field of drug discovery, other companies could use OpenAI's technology to develop new drugs. However, once a new drug is successfully developed, OpenAI would take a share of the profits generated by its AI technology for the client. This suggests that OpenAI might be considering a shift from a "tool-selling" model to a "profit-sharing" business model. OpenAI appears unsatisfied with merely collecting "software usage fees" and instead wants a "cut" when its clients strike it rich.

OpenAI CFO Sarah Friar at the Davos Forum Recently, there has been widespread discussion about OpenAI's revenue pressure, with predictions that they would change their strategy to find ways to profit this year. Who would have thought that changing the business model would mean this? Once this report surfaced, it truly caused a huge stir, once again putting OpenAI in the spotlight. Some believe this could represent a massive颠覆 to the perception of AI as a mere tool. Imagine if you were using Photoshop, and Adobe demanded a royalty on every design piece you created. If this practice were to become an industry standard, the entire cost calculation logic of the business model would be彻底改变 for startups building their businesses on AI APIs.

Others have analyzed, "This might sound far-fetched, but scientists are indeed fascinated by the potential of large language models as 'idea synthesizers' and 'research assistants,' and OpenAI is actively seeking to license private data in fields like biology and pharmaceuticals for model training."

However, some industry insiders have lamented, "It's embarrassing, even a disgrace, that a company which started as a non-profit has come to this point."

"So, OpenAI plans to take a cut from the intellectual property (IP) that users develop using its software, while their own software was built by infringing on the IP of others." OpenAI has long faced controversy over the sources of its AI training data, which includes copyrighted articles, books, code, and artworks. Previously, The New York Times and individuals including several authors have sued OpenAI for using their data to train models without authorization.

Is this model acceptable for enterprises? The answer might be yes. However, it could also lead OpenAI to lose more commercial clients.

Behind OpenAI's Shift in Strategy: AI is Accelerating Drug Discovery Part of the reason OpenAI is floating this revenue-sharing idea is the observable effectiveness of AI as a scientific research tool. Recently, pharmaceutical and biotechnology companies have begun using various forms of AI for drug discovery. Several major pharmaceutical companies have announced deep collaborations with OpenAI, experimenting with using OpenAI's models to analyze vast amounts of data and generate hypotheses or test methods. Last October, Thermo Fisher Scientific, a company providing services and equipment to pharmaceutical firms, stated it would use OpenAI models to accelerate drug development and identify which therapies are unlikely to succeed. OpenAI also appears to be developing increasingly sophisticated AI models specifically tailored for biology and drug discovery, aiming to push AI towards more directly assisting pharmaceutical companies in the discovery process. For example, OpenAI has recently held talks with life sciences diagnostics provider Revvity, Xero, and other biotech companies, seeking authorization to use their more specialized data to train its AI models. In the AI + drug discovery space, OpenAI is not alone; Anthropic and Google's DeepMind have also engaged in discussions with early-stage biotech startups regarding data licensing or partnerships. Sarah Friar is undoubtedly familiar with earlier AI drug discovery companies like Recursion, which have struck deals with pharmaceutical companies entailing large payouts if their technology successfully identifies a drug. However, such success stories, if they exist at all, remain few and far between. Although the competition is just beginning, it is already intense: OpenAI's rivals Anthropic, Google DeepMind, and Alphabet's subsidiary Isomorphic Labs, which focuses on using AI for drug discovery, have all discussed data licensing or partnerships with early-stage biotech startups. Last weekend, Sarah Friar hinted in a blog post that OpenAI could also arrange this type of value-sharing in fields like energy and finance. "IP-based licensing agreements and outcome-based pricing will share the value created," Friar wrote.

Large language models are already adept at discovering structures and patterns that humans might miss. OpenAI's models can sometimes connect concepts from different fields, suggesting novel experiments related to everything from nuclear fusion to pathogen detection. Despite their many limitations and errors, scientists still seem enthusiastic about them. While it might sound fantastical to imagine OpenAI earning more revenue from IP licensing or royalties than from advertising, Sarah Friar's remarks send an extremely clear signal about their intentions. The question is whether she will still be talking about this after OpenAI completes the tens of billions of dollars in funding it is currently seeking from investors.