SANTA CLARA, 27 AUG 2026 — Nvidia is reportedly in talks to invest in Perplexity in a round that would value the search company above US$30bn, up from about US$20bn a year ago. Perplexity's annualised revenue has climbed past US$750m, from under US$250m at the start of this year.
Put those two numbers together and something counterintuitive falls out. The valuation rose by half, and the multiple it represents fell by a lot more than that.
The multiple compressed while the price went up
At more than US$30bn against annualised revenue above US$750m, the company is being discussed at roughly forty times revenue. That is a high multiple by any ordinary standard and a considerably lower one than it carried before.
A year ago the valuation was around US$20bn, struck when revenue was materially below where it now sits — under US$250m as recently as January. On those figures the earlier round implied a multiple somewhere near double the current one.
The company is more expensive in absolute dollars and cheaper as a multiple of its revenue, which simply outgrew the valuation. An investor entering now pays more dollars for a business that is, on this measure, cheaper than it was.
That does not make it cheap. Forty times revenue prices a great deal of future growth, and revenue growing three-fold in eight months is a rate that arithmetic will eventually stop.
Annualised is not the same as earned
The revenue figure deserves the same scepticism as the valuation.
Annualised revenue is a projection. It takes a recent period — often a single month — and multiplies it out, which is a reasonable way to describe a fast-growing subscription business and a poor way to describe what the company actually collected over the past twelve months. A business that tripled during the year has earned considerably less than its annualised figure implies, because most of those months were smaller.
That matters for the multiple. Forty times an annualised number is a larger multiple of trailing revenue, and how much larger depends entirely on the growth curve nobody outside has seen.
None of this makes the figure misleading. It is the standard measure in this category and it is what an investor in a growth round is buying. The two quantities are different, and they are often reported as though they were the same.
Nvidia keeps investing in its own customers
Nvidia has been an investor in Perplexity since 2023. What is new is that this follow-on sits inside a larger pattern.
The company has guaranteed up to US$105bn of OpenAI's Ohio data centre lease. It has taken positions across the model and application layers. And the businesses receiving that support are, without exception, businesses whose growth is bounded by how many accelerators they can afford.
The logic is straightforward, and not necessarily sinister. A supplier whose customers are capital-constrained can convert its own balance sheet into their purchasing power, and the returns arrive twice — once as an investment gain, once as hardware revenue.
The principle is not new. The scale is. When the largest supplier in a market is also a significant investor in a large share of its customers, the demand signal it reports becomes harder to read as independent. Nobody has suggested any impropriety, and none is implied here. The point is narrower: an outside observer trying to judge how much of the demand is organic has fewer clean data points than they did two years ago.
The shape of the deal changed, and that is informative
The most revealing detail in the reporting is what Nvidia considered before equity.
It reportedly weighed a technology licensing arrangement, and separately the possibility of hiring some of Perplexity's team, before the conversation moved to taking a stake.
Those three options accomplish different things. A licence buys the technology, an acqui-hire buys the team, and an equity stake buys exposure to the business as a going concern.
Moving from the first two to the third suggests a judgement that Perplexity is worth more running than dismantled, which is a compliment paid in a form the company would not have enjoyed receiving in the first two versions.
What Perplexity actually sells is contested
Revenue tripling in eight months is the strongest fact here and it deserves a caveat about what is being counted.
Consumer subscriptions, enterprise agreements and any advertising or commerce arrangements all land in the same annualised figure, and they are not equally durable. Subscription revenue from individuals is famously churn-prone in this category, where the switching cost is a browser tab.
No breakdown has been published. A company being valued at forty times revenue by investors who have seen the composition is a different proposition from the same number read by people who have not, and everyone outside the round is in the second group.
The competitive position is the harder question
Perplexity competes with the search incumbent and with the model developers, both of which give away a version of what it sells.
That is survivable — plenty of businesses live alongside a free alternative by being better at something specific — and it does mean the company's defensibility rests on product judgement rather than on a technical moat. The models it runs on are largely other people's.
An investment from Nvidia does not change that. It lowers the cost of the compute underneath — a significant advantage where inference is the largest variable cost — and it says nothing about whether users stay.
What it means from here
The transferable point here is a general one about reading AI valuations.
A valuation headline gives you the price. The multiple tells you what an investor is assuming, and the two can move in opposite directions. A company whose valuation doubles while its revenue triples has become cheaper, and a company whose valuation holds while revenue stalls has become dearer. Almost no coverage reports the second number, and it is the one that tells you what the investor believes.
The narrower consequence is competitive. If this closes, a well-funded search alternative with subsidised compute becomes more viable in markets where the incumbent's dominance is habitual rather than contractual — which describes most of Southeast Asia. That is a slower story than the valuation, and a more consequential one.
