Perplexity CEO Aravind Srinivas indicates on Reddit that his company is not in financial trouble, but is not considering an IPO for the time being. He is responding to rumors on the same platform that the situation at Perplexity is in a precarious state.
Recently, an anonymous questioner on Reddit wrote that based on rumors and his own research, he has five indicators that should show that Perplexity is in trouble. First, the company has apparently stopped all marketing activities and collaborations.
He also claims that Perplexity’s models no longer work properly and that follow-up questions keep switching to an automatically selected LLM. This is instead of an LLM of their own choosing. Certain LLM’s have also disappeared from the list of LLM’s available for use. This would allow Perplexity to save money, according to the questioner.
In addition, Perplexity would have switched server infrastructure and no longer use AWS. This would mean that a cheaper server network would now be used. The company would also consider a rapid initial public offering for a necessary quick cash injection.
Furthermore, the questioner’s own research on LinkedIn would show that many Perplexity employees had been fired or had changed jobs.
Response from Perplexity CEO
Perplexity CEO Aravind Srinivas responded to these allegations on the same platform and provided further clarification.
As for the ‘auto mode’, the reason is that Perplexity now releases many products non-stop. Too many buttons and dropdown menus would only cause chaos for users. According to Srinivas, this is no longer tenable. With ‘auto mode’, the AI itself determines whether end users want a quick answer to a search query or want to ask a real deep research query. According to the CEO, this is also the long-term future for AI chatbots.
For the selection of LLMs, specific LLMs are more suitable for each type of search question than others. Especially when it concerns deep search questions. That is why there is currently no model selection tool available for this and deep search questions are now handled with a combination of Open AI 4o, Claude Sonnet, DeepSeek R1 and the Sonar LLM from Perplexity itself.
According to the Perplexity CEO, the server infrastructure is fine. To further optimize the infrastructure, the logging and lookup activities are being revised, and Python code is being converted to GoLang code. The CEO apologizes for the impact on end users that was not anticipated.
Deep research follow-up questions do indeed go back to auto because research has shown that between 15 and 20 percent of the results to these types of questions were not seen by end users. Perplexity concluded that users want faster and simpler answers, and these can be better provided in auto.
Sufficient funds and no quick IPO
However, the most important question for Perplexity was whether it has sufficient financial resources to continue to exist. CEO Aravind Srinivas’ answer is very simple: There are certainly still sufficient financial resources available, and Perplexity’s sales are growing. Measures are being taken to improve functionality rather than cut costs.
In addition, an IPO is not expected before 2028.
Read also: Perplexity seeks growth capital to strengthen position as AI search engine