Important points
Analysts say this year will be important for private AI companies, especially OpenAI, as investors focus on returns. For companies whose sole business is selling AI models, this is “make or break,” Deutsche Bank wrote in a Jan. 20 note. “OpenAI may be the most at risk, as it is particularly scaled and appears to have yet to find a viable business model to cover its reported $9 billion cash burn last year and likely $17 billion this year,” investment bank analysts Adrian Cox and Stéphane Avredan said in a statement. Of its estimated 800 million weekly users, only a “small fraction” pay. At the same time, AI leaders are working on $1.4 trillion worth of data center projects. OpenAI had more than $20 billion in revenue last year, up from $6 billion in 2024, according to a blog post by finance director Sarah Friar. The company is widely expected to go public later this year or in early 2027. The company has struck deals with Nvidia and Microsoft, among others, and raised billions of dollars in the process, with a potential valuation of $500 billion. The company secured $22.5 billion from SoftBank late last year, on top of $40 billion already committed by investment firms. OpenAI has partnered with many hyperscalers, but compared to other large competitors whose AI handbooks are backed by sound business fundamentals, OpenAI’s moat is “relatively shallow,” Cox and Abourdan wrote, adding that “the path to success appears to be getting narrower and narrower.” “Pressure will only increase as we move closer to an IPO, which is expected to be discussed in early 2027 and could exceed $1 trillion,” they said. In a blow to OpenAI, Apple on January 12th chose to leverage Google’s technology in its AI products. On January 16th, OpenAI announced that it would soon be testing advertising on ChatGPT. Founder Sam Altman said the 2024 transition is a “last resort” business model. Dimitri Zabelin, senior investment research analyst covering AI and cybersecurity at PitchBook, says this represents a new phase for underlying model developers, as “investor scrutiny has shifted from scale to reliable improvements in returns, or at least unit economics.” “The key question is whether enterprise monetization, pricing power, and lower inference costs can outweigh increases in compute intensity,” he said, but added that “OpenAI’s access to strategic compute and capital partners remains unusually deep” thanks to multi-year capacity agreements that demonstrate support for its scaling roadmap. Competitor Anthropic, founded by a group of former OpenAI employees, is also rumored to be aiming to go public as early as this year. Zavelin said the companies are benefiting from regulatory tailwinds “especially as they continue to become more integrated into government operations at home and abroad through sovereign AI initiatives.” Market participants expect the U.S. Federal Reserve to take a more dovish stance on interest rates, but concerns about intervention are roiling markets, which could further accelerate generative AI financing despite bubble concerns, according to S&P Global. But Deutsche Bank analysts are not convinced. “It will prove nearly impossible for small independent companies to withstand the computational costs of the accelerated models,” they said. “We can’t rule out the possibility that something like Perplexity will end up in the hands of a hyperscaler by the end of the year. Anthropic may be an exception. It has a slower cash burn than OpenAI, is particularly popular with programmers and (paying) enterprises, and has a more dynamic pricing model.”
