OpenAI under siege: Sam Altman's lead is no longer a given

Remember the glory days when every OpenAI update sent competitors scrambling to rewrite their roadmaps? That period of dominance appears to be over. Sam Altman's company now finds itself double-encircled: on one side the resource-heavy behemoth Google, on the other Anthropic's finely tuned engineering. What once looked like unconditional leadership now looks like a desperate attempt to stay afloat in a sea of growing competition.

The "life or death" situation for OpenAI is not a question of bankruptcy, since Microsoft has enough money for years to come. It is a question of relevance. While OpenAI was busy with internal reshuffles, the departure of Ilya Sutskever and the disbanding of its safety teams, its rivals were not sitting idle. Google has finally gotten its neural networks in order, rolling out Gemini 1.5 Pro with a context window of two million tokens. That literally changed the rules of the game: you can now feed the model an entire library or hours of video, and it will not "forget" the beginning of the conversation. Meanwhile, GPT-4o still stumbles in long conversations, leaving users feeling its limits.

But the real blow came from a direction few expected on this scale. Anthropic released Claude 3.5 Sonnet, which on many measures, especially writing code and grasping the nuances of human language, simply flattened OpenAI's current offerings. Developers are switching to the "Anthropic" side en masse because the model is faster, more accurate and, importantly, cheaper. OpenAI's magic has started to wear off, exposing an ordinary corporate structure that seems to have become too fond of PR stunts like the "voice from the movie Her" instead of real architectural innovation.

The company's inner workings do not add much optimism either. A mass exodus of top researchers to startups such as Safe Superintelligence (SSI) suggests an ideological crisis has been brewing inside OpenAI. When the best minds leave to "do AI right," that is a bad signal for investors and the market. OpenAI has turned from a research lab into a commercial machine that has to ship products in a hurry to meet expectations and an endless series of funding rounds. That rush leads to mistakes and to the loss of the very "secret sauce" that made the company first.

What next? Altman still has the Strawberry project and the mythical GPT-5 up his sleeve. But the industry no longer waits for just a "bigger model." The market demands efficiency, agentic capabilities and reliability. If OpenAI's next iteration does not show a qualitative leap in reasoning, rather than just in data volume, the "double encirclement" will turn into full capitulation. We are watching the end of the single-leader era and the shift to a harsh oligopoly, where OpenAI will have to fight for every percentage point of market share.

Bottom line: Can OpenAI regain its status as an innovator, or will it finally become "Microsoft 2.0," a reliable but boring corporation chasing bolder competitors?