Nvidia is in discussions to invest in Perplexity as part of an equity round that would value the AI search company at more than $30 billion, The Information reported Sunday. Reuters picked up the story Monday and was explicit that it could not independently verify the details. Neither company has confirmed a thing. Benzinga said Nvidia and Perplexity did not immediately respond to requests for comment.
Hold onto that. This is a reported negotiation, not a closed deal, and the terms — size of the round, Nvidia's share of it, whether any compute commitment rides along — are not public.
What is reasonably well established is the markup. Perplexity finalized a round at roughly $20 billion last September. A $30 billion-plus round would be more than 50% higher in about twelve months. Zoom out and the curve gets steeper: Perplexity was valued at $121 million in April 2023. Thirty months later it is negotiating at 250 times that.
The revenue number, and the asterisk
The bull case rests on a specific figure. Perplexity's annualized revenue has climbed above $750 million, up from less than $250 million at the start of the year — roughly a threefold jump in eight months, according to the reporting. Much of the acceleration is attributed to Perplexity Computer, the cloud-based agent that automates desktop work for professionals by orchestrating 19 separate models behind one interface.
That figure deserves scrutiny, and at least one outlet has applied some. Writing for Forkast on Yahoo Finance, Lena Park noted that third-party trackers Sacra and Tracxn put Perplexity's ARR at roughly $450 million to $500 million as of April 2026, and argued: "While some external reporting has cited a $750 million ARR figure, this appears to be a conflation with the $750 million Microsoft Azure infrastructure commitment signed by the company in January 2026."
There is a $750 million Microsoft number, and it is not revenue. It is a three-year spending commitment to access models through Azure's Foundry program — a cost line, not a top line, with AWS remaining Perplexity's primary cloud. Whether the ARR figure is real or an artifact of two identical numbers is the most load-bearing question in this story, and nobody outside the cap table can answer it.
Why Perplexity specifically
Nvidia has been on Perplexity's cap table since 2023 and led a Series E extension in July 2025. Ties tightened this year: in July, Perplexity committed to running its agent workloads on Nvidia's Vera CPUs. And in March, at GTC in San Jose, Nvidia named Perplexity an inaugural member of the Nemotron Coalition alongside Black Forest Labs, Cursor, LangChain, Mistral AI, Reflection AI, Sarvam and Thinking Machines Lab — a group pooling data, expertise and DGX Cloud compute to train an open model that will underpin the Nemotron 4 family.
The coalition is Nvidia's answer to an uncomfortable fact: Chinese open-weight models went from about 1.2% of global open-source usage in late 2024 to roughly 30% by the end of 2025, and Alibaba's Qwen displaced Meta's Llama as the most-used self-hosted model in the world. Nvidia would rather the default open model be one trained on its silicon.
Note also what Nvidia reportedly considered before pivoting to equity: a licensing arrangement, possibly paired with hiring some of Perplexity's team. That is the structure it used on Poolside this month — about $1 billion of equity at a $12 billion pre-money valuation plus roughly $6 billion for a technology license, with more than 100 Poolside engineers moving to the Nemotron program — and on Groq in December 2025.
The circularity problem
Nvidia's portfolio now includes OpenAI, xAI, Safe Superintelligence, Poolside, CoreWeave and Nebius. In February it agreed to invest $30 billion in OpenAI. Separately it put $1.5 billion into SoftBank-backed SB Energy, which is building the PORTS-Pike campus in Ohio that OpenAI will fill exclusively with Nvidia hardware. Nvidia's credit support there was initially floated near $250 billion before being scaled back to the first 4.25 gigawatts of an 8 GW commitment, with Nvidia absorbing up to $105 billion of residual value loss if OpenAI walks.
Critics call this circular financing: Nvidia funds the buyers, the buyers buy Nvidia. Bernstein's Stacy Rasgon — no bear, he has repeatedly argued the stock is cheap and once wrote that "there's no such thing as too many chips" — has flagged the pattern as a persistent source of investor unease around Nvidia's largest deals.
Jensen Huang rejects the framing outright. Asked directly whether the Ohio arrangement was circular financing, he wrote: "No. OpenAI will pay the lease." He argues Nvidia is securing scarce infrastructure rather than manufacturing demand, and that frontier labs are simply "growing faster than their balance sheets and long-term credit profiles can support." His summary of the whole strategy: "In the AI economy, compute is revenue."
Both things can be true. Perplexity's compute spend is real. So is the fact that a supplier taking equity in its customers makes demand signals harder for outsiders to read — which is precisely why the ARR discrepancy matters.
What to watch
Nvidia reports fiscal Q2 on August 26, and analysts are broadly constructive: Bank of America's Vivek Arya carries a $350 price target, Oppenheimer an Outperform at $265. Nvidia closed Friday at $214.72, down 0.98%.
Three things matter more than the headline number. Whether the round closes at all and at what size — a $30 billion valuation only means something once someone wires money. Whether Nvidia's participation comes bundled with a compute commitment, as the Vera CPU deal did; that says something different about strategy than a passive check. And whether anyone puts an audited number next to that $750 million before Perplexity's stated 2028 IPO. As Aravind Srinivas told CNBC in June, "Agnostic of these two companies, we were planning for something in 2028 so that still remains the case" — referring to OpenAI's and Anthropic's listings. That gives the market two years to find out which revenue figure was the right one.
“No. OpenAI will pay the lease.”— Jensen Huang, CEO, Nvidia