Nvidia in Talks to Back Perplexity at $30B Valuation

Nvidia is in discussions to invest in Perplexity as part of an equity funding round that would value the AI search startup at more than $30 billion, according to a report from The Information published August 23, 2026. Neither company has confirmed the talks, which remain preliminary and could still fall apart. But the reported deal would mark another escalation in Nvidia’s strategy of taking direct stakes in the companies that consume its chips — extending the world’s most valuable semiconductor maker deeper into the software layer that runs on top of its hardware.

The round would represent a jump of more than 50% from Perplexity’s last private valuation. The San Francisco-based company was valued at roughly $20 billion in a financing round in September 2025. A $30 billion-plus mark would place it among the most valuable AI application companies in the world, behind only a handful of foundation-model labs.

The revenue behind the number

What separates this reported round from the froth of the broader AI funding market is the revenue curve underneath it. Perplexity’s annualized revenue has climbed above $750 million, up from less than $250 million at the start of 2026, according to figures cited in the reporting. That is roughly a tripling of the run-rate in under eight months — the kind of growth that justifies, at least on paper, a valuation step-up of the size being discussed.

Much of that acceleration has come from products that push Perplexity beyond its original answer-engine. The company, led by CEO Aravind Srinivas, has expanded from AI-powered search into an AI browser called Comet and an agent layer, Comet Assistant, that can carry out multi-step tasks on a user’s behalf. It has also leaned into Perplexity Computer, a cloud-based AI agent aimed at professionals who want to automate routine computer work — filling forms, pulling data, navigating web apps — without writing code.

Those agentic products matter for the economics. Search queries are cheap to monetize but hard to charge for directly; agents that complete billable work are a more natural fit for subscription and usage pricing. The revenue mix appears to be shifting toward the latter, which is part of why investors are willing to underwrite a higher multiple.

Nvidia is already on the cap table

Nvidia is not a new backer. The chipmaker participated in earlier Perplexity rounds alongside a roster of high-profile investors including Jeff Bezos, SoftBank, Accel, and IVP. A fresh check at a higher valuation would deepen an existing relationship rather than open a new one.

That distinction is important. When a strategic investor doubles down, it signals conviction that the earlier bet is working — and it gives Nvidia a larger claim on a company whose growth is directly tied to demand for inference compute. Every Comet session, every Perplexity Computer task, and every search answer runs on GPUs. Backing Perplexity is, in effect, Nvidia financing its own future order book.

The compute-landlord playbook

The reported talks fit a pattern Nvidia has been running all year. The company has committed capital to model labs, cloud providers, and application startups across the stack, from its reported $5 billion investment in Safe Superintelligence to a $6 billion model-factory license deal with Poolside. In each case the logic is the same: Nvidia supplies the hardware, then takes an equity or licensing position in the company buying it, capturing upside beyond the one-time chip sale.

Critics call it circular — Nvidia funding customers to buy Nvidia chips, inflating demand that flatters its own results. Supporters call it a rational way for the dominant compute supplier to internalize the value it creates. Either way, a Perplexity stake would push the strategy into a new segment: consumer- and prosumer-facing AI search and agents, a market where Nvidia has had less direct exposure than in data-center training.

There is a competitive angle too. Perplexity joined Nvidia’s Nemotron Coalition earlier in 2026, an effort to promote open AI models as a counterweight to fast-moving Chinese labs. Nvidia has separately been pushing hard on open-weight models, releasing systems like Nemotron 3.5 Lightning to seed an ecosystem that runs best on its silicon. A larger stake in a marquee application company that champions open models advances that agenda.

Perplexity’s legal and competitive backdrop

The reported round arrives as Perplexity navigates a thornier operating environment than a pure growth story would suggest. The company’s aggressive browser and agent products have drawn scrutiny over how they access third-party websites and data. Perplexity has been at the center of a closely watched dispute over automated browsing and the Computer Fraud and Abuse Act, a case tracked in our coverage of the Comet–Amazon Ninth Circuit ruling. How courts treat AI agents that act on a user’s behalf will shape the addressable market for exactly the products driving Perplexity’s revenue.

On the competitive front, Perplexity sits between two kinds of rivals. Above it are the foundation labs — OpenAI, Anthropic, Google — that increasingly bundle search and agentic features directly into their assistants. Below it are a swarm of AI-search challengers. Perplexity’s bet is that a focused, fast-moving product company can out-execute the labs on user experience while riding their models where useful. A deep-pocketed strategic backer like Nvidia strengthens that position, both financially and in access to compute.

What it means

If the round closes near the reported terms, it tells you three things about where the AI market is in late 2026.

First, revenue is finally the story. A $30 billion valuation on $750 million of annualized revenue is a rich multiple, but it is anchored to a real, fast-growing top line rather than to model benchmarks or promise. The market is rewarding AI companies that can convert usage into recurring dollars — and Perplexity’s shift toward agents and its Computer product is the mechanism. Watch whether that revenue holds its growth rate as the base gets larger; tripling from $250 million is easier than tripling from $750 million.

Second, Nvidia’s strategy is hardening into something structural. The compute-landlord thesis — supply the chips, then own a slice of the demand — is no longer opportunistic. It is a deliberate, repeated pattern spanning labs, clouds, and now consumer AI search. That gives Nvidia extraordinary leverage over the AI economy, but it also concentrates risk: if application-layer growth disappoints, Nvidia absorbs the loss twice, once as an investor and once as a supplier. Regulators and rival chipmakers are watching the circularity closely.

Third, the inference layer is where the next fight is. Training runs made the headlines and the first fortunes, but the durable, recurring revenue lives in inference — the compute burned every time a user asks a question or dispatches an agent. Nvidia backing a leading inference-heavy application company is a bet that this layer, not model training, is where sustained demand will accumulate. Perplexity’s numbers are early evidence for that view.

The immediate caution is that the talks are unconfirmed and preliminary; The Information’s report describes discussions, not a signed term sheet, and either side could walk. But the direction of travel is unmistakable. The winners take stakes in each other, the compute supplier finances its own customers, and the valuations keep climbing on the strength of revenue that — for now — is actually showing up. Whether that reflexivity is a healthy flywheel or a warning sign is the question hanging over the entire AI trade heading into the fall.