Ultraman internal letter exposed: OpenAI's leading advantage narrows, warning of "difficult times" ahead

OpenAI CEO Sam Altman warned employees in an internal memo that the company's technological lead is narrowing and that the external environment is expected to be "quite challenging" for some time. This rare warning signal indicates that the AI giant, valued at $500 billion and having received over $60 billion in investments, is facing unprecedented competitive pressure.

On November 21, according to tech media The Information, Altman acknowledged in last month's internal memo that Google's recent advancements in artificial intelligence "may pose some temporary economic headwinds for our company" and warned employees that "I expect the external atmosphere to be quite difficult for a while." He admitted that Google's newly created AI seems to have surpassed OpenAI in terms of development methods, stating, "We know there is still work to be done, but we are catching up quickly."

The report suggests that this internal statement foreshadows the significance of Google's launch of Gemini 3 this week. Software developers have noted that this AI model excels in automating website and product design-related tasks as well as programming, which is one of the most important revenue drivers for AI companies like OpenAI.

More critically, OpenAI's competitors are making breakthroughs on multiple fronts. According to previous reports from The Information, Anthropic, which was founded only four years ago, may surpass OpenAI this year in revenue from selling AI to developers and enterprises. Meanwhile, although ChatGPT still significantly leads Google's Gemini chatbot in usage and revenue, the gap is narrowing.

Challenges to Technological Leadership

The report states that Google's breakthroughs in AI pre-training have surprised many AI researchers. Pre-training is the first stage in developing large language models, during which researchers expose the model to data from the web and other sources to help it learn the relationships between the data.

Altman acknowledged in the memo that "Google has been doing excellent work recently in every aspect," particularly in pre-training. This area has been one where OpenAI has struggled to make progress, and Google has also faced challenges.

These challenges have previously prompted OpenAI to shift more towards a new type of AI called reasoning models, which use more processing power to generate better answers.

According to previous reports from The Information, before OpenAI launched the GPT-5 model this summer, employees found that the adjustments made to the model during pre-training were effective when the model was smaller but ceased to work as the model grew. This indicates that OpenAI needs to address these pre-training issues to catch up with Google in this field.

Last month, Altman assured employees that OpenAI would make progress in the coming months, including the launch of a new large language model codenamed Shallotpeat. According to insiders, OpenAI aims to fix the bugs encountered during the pre-training process while developing this model

Competitors Break Through on Multiple Fronts

The report also points out that OpenAI's dominance is being shaken. Anthropic was founded by former OpenAI employees, and its models specialize in generating computer code based on customer needs, from developing new applications to updating existing code.

According to a report from The Information earlier this month, this four-year-old company is expected to surpass OpenAI in revenue generated from selling AI through application programming interfaces to software developers and businesses this year.

Meanwhile, Google continues to promote its Gemini chatbot through its search applications and other products, competing with OpenAI's ChatGPT. Although ChatGPT still significantly leads Gemini in terms of usage and revenue, the gap has been narrowing. Altman stated in a memo:

"For most people, ChatGPT is AI, and I expect that to continue."

The report also noted that another advantage for Google is its financial strength. OpenAI is one of the fastest-growing companies in history, growing from nearly zero revenue in 2022 to an expected $13 billion this year. However, the company also anticipates that in its pursuit of human-level AI, it will consume over $100 billion in the coming years, while needing to spend hundreds of billions to rent servers, which means it may need to raise an equivalent amount of additional capital.

In contrast, Google, valued at $3.5 trillion, generated over $70 billion in free cash flow in just the past four quarters. While ChatGPT seems to be eating into Google's search market share, Google's financial performance has improved, partly due to its thriving cloud business, which rents servers to large clients including OpenAI and Anthropic.

The financial gap between OpenAI and established companies like Google has led public market investors to question whether the unprecedented revenue growth of this startup (including expected growth) is sufficient to alleviate concerns about its future cash consumption.

Focus on Long-Term "Ambitious Bets"

Despite the pressure, Altman expressed hope to focus on "very ambitious bets" technologically, even if it means OpenAI may "temporarily lag under the current regime."

These bets include making progress in using AI-generated data to train new AIs, as well as "post-training" techniques (such as reinforcement learning), which essentially provide positive or negative evaluations of model answers, allowing them to learn and improve.

Altman has privately and publicly discussed the company's bets on automating AI research itself to accelerate breakthroughs, including enabling AI to surpass human capabilities across various fields from energy and biotechnology research to healthcare. Altman stated in the memo:

"We need to stay focused under short-term competitive pressure; we as a company have accumulated enough strength to withstand excellent models launched elsewhere. But it is crucial that most of our research teams focus on truly achieving superintelligence." He added, "I don't want this letter to be discouraging—our company has performed exceptionally well... I expect this situation to continue."

At the end of the memo, he wrote:

"We have to do so many difficult things at the same time—the best research labs, the best AI infrastructure companies, and the best AI platform/product companies—it's tough, but that's our fate. And I wouldn't trade places with any other company."