Perplexity plans to go public in 2028 regardless of how Anthropic and OpenAI fare in their own market debuts, CEO Aravind Srinivas told CNBC in an interview aired Tuesday. The timeline lands as Anthropic confidentially filed for an IPO last week at a prior valuation of nearly $1 trillion, and OpenAI filed its own confidential paperwork on Monday. SpaceX is also listing this week, stacking three of the largest tech IPOs in history into a single window.
Srinivas had previously said Perplexity would not go public before 2028. His latest comments harden that into a date the company intends to hold, even if the listings ahead of it stumble.
“Agnostic of these two companies, we were planning for something in 2028 so that still remains the case”— Aravind Srinivas, CEO of Perplexity
The CEO acknowledged the dependency on what happens in the next several months. He said the SpaceX IPO this week will function as the leading indicator for how investors price Anthropic and OpenAI, and that ripple effects from a weak reception would be unavoidable. He added that he expects the listings to go well because the underlying businesses are performing.
Key facts
- 01Perplexity CEO Aravind Srinivas confirmed a 2028 IPO target, independent of Anthropic and OpenAI's listing outcomes.
- 02Anthropic confidentially filed for an IPO last week at a prior valuation near $1 trillion; OpenAI filed Monday.
- 03Srinivas said a six-month gap without model capability advances would dent frontier-lab valuations.
- 04Perplexity routes tasks to open-source models when they hit 90% reliability at 10–20x lower cost than frontier models.
- 05SpaceX's IPO this week is the leading indicator for how the Anthropic and OpenAI listings will land.
Anthropic's near-$1 trillion private mark and OpenAI's pending valuation are both under scrutiny. Srinivas argued both deserve their pricing because they sit at the frontier of model capability, but he tied that defense to a specific condition: continued progress on model quality.
He drew the line at six months. If either lab goes that long without shipping a meaningful capability advance, the valuations come under pressure. He said there is no sign of that slowdown happening now, but the metric is concrete enough to be measured against future releases of Claude and GPT.
The conversation pivoted to AI spending, which has become a board-level question after OpenAI CEO Sam Altman reportedly called AI costs a "huge issue" on a recent company livestream. One emerging pattern Srinivas flagged is "tokenmaxxing," where employees crank up AI usage to signal productivity rather than to get better outputs.
Perplexity's own product routes queries across multiple model providers, picking the best fit per task with cost factored in. Srinivas described a clear threshold for when frontier intelligence loses the argument on a given workload.
That stance has direct revenue implications for the labs Perplexity depends on. If a meaningful share of enterprise tasks can be served by open-source models at a tenth to a twentieth of frontier pricing, the API revenue base that justifies trillion-dollar valuations narrows to the workloads where frontier capability is genuinely required. Srinivas framed the future of frontier intelligence as still strong, but no longer the recipient of "mindless spending."
Perplexity itself sits in an unusual position heading into a 2028 listing. It does not train its own frontier models, instead orchestrating across third-party providers, which keeps its compute bill lower than a lab's but ties its margins to whatever pricing the labs set. A scenario in which Anthropic and OpenAI lose pricing power to open-source alternatives is, mechanically, good for Perplexity's gross margin.
The risk on the other side is timing. A cold reception for the SpaceX, Anthropic, or OpenAI offerings would reset the comparable set Perplexity will be priced against in 2028, and the company has roughly two and a half years to grow into whatever multiple the public market is willing to pay by then. Srinivas's bet is that the labs ahead of him land well enough to keep the window open.
Perplexity's 2028 commitment is the more interesting signal in the interview than the commentary on its larger peers. It tells the market the company believes it can grow on its current architecture — a router sitting on top of other people's models — for long enough to reach public-market scale, and that it does not need to become a frontier lab to get there. If that thesis holds, the AI stack splits cleanly into model providers and model orchestrators, and the orchestrator layer turns out to be a viable standalone business. If it doesn't, 2028 is a long time to wait to find out.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
