Perplexity

Full Report · Nvidia buying Perplexity · 2026-09-08

You are considering acquiring this company outright.

Worth a term sheet with the specific reps and indemnities named here.

What am I actually buying, and what breaks the day I own it?

Signals that fired and how heavily this decision weights each: layoffs (15), litigation active (45). Combined 60/100.

Standing next step: Term sheet, with the risks above written as reps and a holdback sized to them. Do not net them off the price; price them separately so each one is arguable.

The written sections below reach their own conclusion from the full evidence. Where they and this weighted read differ, the sections carry the argument and the citations.

The Short Answer

What the whole report comes to, and what to do about it.

You are buying a fast-growing AI answer engine with a credibly large and accelerating revenue base, but the core product mechanism — crawling and synthesising third-party content at scale — is under active legal attack from multiple named plaintiffs, and an adverse ruling could require fundamental architectural changes to the product on the day you own it.

The revenue trajectory is real in direction if not in precise magnitude: Sacra estimates annualized revenue of $750 million as of August 2026, up from $232 million at end of 2025, and TechCrunch reported ARR approaching $200 million as of September 2025 — all private estimates, none audited, but directionally consistent. The valuation Nvidia is reportedly discussing — above $30 billion per Yahoo Finance in August 2026 — is being set against a cost structure that, at an earlier stage, showed $57 million spent on AI models against $34 million in revenue, with no public profitability figure on record at any point.

The litigation stack is not a tail risk — it is a present operational constraint. Reddit's data-scraping lawsuit survived a dismissal motion in July 2026, Cloudflare independently confirmed that Perplexity deployed stealth crawlers to evade no-crawl directives, and active copyright suits from Dow Jones, Britannica, and CNN each attack the same retrieval mechanism. A CIPA class action filed March 2026 alleges user prompt data was shared with Google and Meta — a direct conflict with Nvidia's enterprise data commitments.

The founder has said, on the record, that he does not want to be acquired by Big Tech. Aravind Srinivas told CNBC in July 2025 that Perplexity plans to remain independent. That statement predates the current discussions, but it is the most recent public position from the person whose retention is most critical to the asset's value.

The strongest argument against this assessment is that the revenue growth rate, if Sacra's estimates are even approximately correct, is exceptional — roughly 3x in under a year — and Nvidia's existing relationship as an investor and infrastructure partner gives it a structural advantage over any other acquirer in negotiating both price and founder retention. Nvidia is not a generic Big Tech buyer; it is already embedded in Perplexity's technical stack through the NeMo partnership, which changes the integration calculus.

The single most important thing that could not be established is the gross margin by product line net of model inference costs. The one internal cost figure in evidence — $57 million on AI models against $34 million in revenue at an undated earlier period — raises the question of whether the business is profitable at the unit level at current scale, or whether the revenue growth is being bought at a loss per query. That figure would be found in Perplexity's audited financial statements or, absent an audit, in the management accounts and revenue recognition policy that any serious data room would contain.

Concretely: before any price is agreed, Nvidia should demand the full audited or management accounts for the last two fiscal years, the complete cap table with liquidation preference terms for all eleven rounds, the text of every material customer and publisher contract with change-of-control flags identified, and a legal opinion on the exposure in each active lawsuit. The Amazon litigation — 3:25-cv-09514 (N.D. Cal.) — deserves specific attention: Nvidia acquiring Perplexity would make Amazon an active litigant against a direct competitor, and any AWS infrastructure agreement almost certainly contains a change-of-control clause that could be triggered on close.

Which Company This Is

The entity this report is about, and how we know it is that one and not another of the same name.

The company under review trades as Perplexity and operates through the domain perplexity.ai, which is the anchor that settles its identity. It is incorporated in the United Kingdom as PERPLEXITY AI LTD, company number 15470377, registered in London, with an incorporation date of 6 February 2024 and a current status of active on Companies House. The operating entity is an American privately held software company headquartered in San Francisco, California, as confirmed by Wikipedia and Yahoo Finance; the UK registration appears to be a subsidiary or holding vehicle rather than the primary operating entity.

The company was founded in August 2022 and is at the late-venture stage, having completed at least a Series D round as of December 2024 per Sacra, with a most recently reported valuation of $22.6 billion per Tracxn and total funding of $1.72 billion across eleven rounds.

One registry match requires explicit qualification. The GLEIF LEI index returns identifier 969500KXRLQTCC2JJ496 registered to the legal name "PERPLEXITY LABS," status active. This was located by searching the index for that exact name and nothing further was verified against it. It is reported here as an identifier registered to a company of that name, not as a confirmed registration of the subject of this report. A direct check of the LEI registration record against the subject's company number 15470377 and registered address would settle whether these are the same legal entity.

Companies House records for PERPLEXITY AI LTD returned no persons-with-significant-control statement in the evidence supplied; that gap is noted and unresolved.

The identification of the operating company is high-confidence on the basis of the domain, the UK registry number, and corroborating coverage across multiple independent sources; the LEI match to "PERPLEXITY LABS" remains unconfirmed and should be resolved before closing.

What They Do, And How The Money Works

The business as it actually operates: what is sold, to whom, on what terms, and where the revenue comes from.

Perplexity sells access to an AI-powered answer engine that retrieves, synthesises, and cites information from the web in real time, returning structured answers rather than a list of links. The core product exists in three commercial layers: a consumer subscription, an enterprise platform, and a developer API.

The consumer tier is called Perplexity Pro, priced at $20 per month or $200 per year. It unlocks higher usage limits, access to premium AI models, a feature called Perplexity Computer, and premium data sources including PitchBook. A free tier exists with limited weekly usage. The buyer here is an individual knowledge worker, researcher, or student who wants cited, real-time answers without managing a separate search workflow.

The enterprise tier, Perplexity Enterprise Pro, is sold to organisations and is priced at up to $325 per user per month depending on configuration, according to published pricing guidance. The company claims more than 50,000 enterprise customers. Named use cases include finance teams, legal research, public safety organisations, and product teams. The buyer is typically a department head or IT decision-maker at a mid-to-large organisation. Perplexity has also extended this tier specifically to law enforcement and public safety agencies, which it describes as a first for a frontier AI company.

The developer API is sold on a pay-as-you-go basis, with no flat monthly fee; buyers are charged per token consumed. This channel is aimed at software teams embedding Perplexity's search and answer capabilities into their own products. Apollo's sales platform is a cited example, where API integration is reported to have driven a 46% increase in booked meetings for Apollo users.

Revenue is predominantly recurring. Consumer Pro subscriptions are annual or monthly contracts. Enterprise Pro is a per-seat subscription. The API is transactional but generates predictable volume from embedded integrations. Sacra estimates annualised revenue reached $750 million in August 2026, up from $232 million at end of 2025. TechCrunch reported in September 2025 that annual recurring revenue was approaching $200 million at that point, corroborating the direction of growth. An earlier data point from a LinkedIn post citing Perplexity's own financial disclosures showed $34 million in revenue at an earlier undated period, with $57 million spent on AI models — a cost structure that signals the business was running at a significant loss at that stage and that model inference costs are the dominant operating expense.

One element of the model that is not fully visible from outside is the advertising component. A class action filed in March 2026 alleges that user prompts were shared with advertisers including Google and Meta, which implies an advertising or data-monetisation layer is either live or was tested. Perplexity has also announced a publisher revenue-sharing programme, suggesting advertising revenue is being built into the product alongside subscriptions. The precise split between subscription and advertising revenue is not publicly disclosed.

For Nvidia considering an outright acquisition: the money comes primarily from recurring per-seat and per-token subscriptions, but the cost of serving those subscriptions — dominated by AI model inference — has historically exceeded revenue, and the advertising layer that could change that unit economics picture is currently the subject of active litigation.

Who Runs It, And How To Reach Them

The named people who decide things here, what is known about them, and where a route to them exists.

Aravind Srinivas is the co-founder, President, and CEO of Perplexity, and is the person any acquirer would negotiate with directly. His LinkedIn profile records his tenure at Perplexity from August 2022 to the present, making him one of the company's original founders with just over four years in the role as of this report date. Before Perplexity, he held research positions at OpenAI and Google DeepMind, according to Wikipedia's entry on Perplexity AI. He completed a PhD at UC Berkeley, as confirmed by Stanford GSB coverage. His public LinkedIn profile is at linkedin.com/in/aravind-srinivas-16051987. As recently as June 2026 he was giving on-record interviews to CNBC and Fortune, so his presence in the role is current across multiple independent sources.

One material point for this decision: in July 2025, Srinivas stated publicly that Perplexity plans to remain independent and that he is not interested in a Big Tech acquisition. Nvidia is, by any reasonable characterisation, a Big Tech acquirer. That stated position is on the record and should be treated as a negotiating constraint, not a formality.

Denis Yarats is co-founder and CTO, also present since August 2022 per his LinkedIn profile. He is the technical counterpart to Srinivas and would be central to any assessment of the engineering organisation and model infrastructure. No departure or role change was found in the web index search for leadership changes.

Johnny Ho is listed as co-founder and Chief Strategy Officer across Craft.co and The Org. These are secondary aggregator sources; no primary confirmation of his current role was found in the evidence, so his status should be verified directly.

Andy Konwinski is listed as a co-founder across Tracxn and Crunchbase, and also appears as a board member per Crunchbase. No current operational title is confirmed in the evidence; his role appears to be that of a founding-era board presence rather than a day-to-day executive.

Dmitry Shevelenko holds the title of Chief Business Officer and was appointed to the board of directors of Lazard effective September 2, 2025, per Lazard's own announcement. He is the most senior commercial executive named in the evidence and would be relevant to any revenue and partnership due diligence.

The Companies House record for PERPLEXITY AI LTD (company number 15470377, incorporated 2024-02-06) returned a persons-with-significant-control statement rather than named individuals, meaning no PSC is formally registered at Companies House as of the search date. This is a gap: it does not establish that no controlling person exists, only that none is named on the UK register. For an outright acquisition, the beneficial ownership structure of the UK entity would need to be resolved separately.

The Money: Funding, Valuation, Runway

Every financial figure on the record, dated and sourced, and what the shape of it implies.

Perplexity has raised capital across multiple rounds since its founding in 2022. The earliest round on record is a Seed, followed by a Series A of $25.6 million closed in March 2023, led by Peter Sonsini of New Enterprise Associates. A Series B brought cumulative funding to $100 million by January 2024, with IVP leading and prior investors continuing. Tracxn records a Series C in April 2024 and a Series D in December 2024, though neither the amounts nor lead investors for those rounds are established in the evidence available. CNBC reported in November 2024 that Perplexity was in the final stages of raising $500 million at a $9 billion valuation. TechCrunch reported in September 2025 that the company raised $200 million at a $20 billion valuation. A LinkedIn post dated February 2026 references a $1.5 billion raise at a $14 billion valuation, which conflicts with the September 2025 TechCrunch figure of $200 million at $20 billion — these two reports disagree on both amount and valuation and cannot be reconciled on the available evidence. Tracxn states total funding of $1.72 billion across eleven rounds and a current valuation of $22.6 billion, though this is a third-party tracker figure, not a filed or confirmed disclosure.

The valuation trajectory is unambiguously upward in reported terms: from $9 billion in late 2024 to $18 billion by Bloomberg's July 2025 report, to $20 billion per TechCrunch in September 2025, to a reported $22.6 billion per Tracxn, to a figure above $30 billion now under discussion. Specifically, Yahoo Finance reported on August 23, 2026 that Nvidia is in discussions about a Perplexity investment at a valuation exceeding $30 billion. All of these are reported figures from press and tracker sources; none is a filed or audited valuation. Nvidia's own involvement as a prospective investor at this stage is directly relevant to this acquisition review.

On revenue, the evidence carries several signals but none is a filed number. TechCrunch reported in September 2025 that ARR was "approaching $200 million," attributed to a source familiar with the company, not to management on the record. A Reddit post from March 2026 cites "several investor press outlets" as the basis for the same $200 million ARR figure as of September 2025. Sacra, a private-company research service, estimates annualized revenue of $750 million as of August 2026, up from $232 million at end of 2025, and attributes growth to a February event not further specified in the evidence. Sacra's figures are estimates, not disclosures. A LinkedIn post from May 2025 references Perplexity releasing financial data showing $34 million in revenue and $57 million spent on AI models, though the period covered is not stated in the evidence. No profitability figure has been put on the record by anyone.

The last confirmed raise appears to be in late 2025, making the gap to the current date approximately nine to twelve months. That gap is not long by private-company standards and, given the revenue trajectory reported by third parties and the active discussions around a new round at over $30 billion, points toward growth rather than distress — though the evidence does not include audited financials, so that reading rests on unverified third-party estimates. The more pointed observation for an acquirer is that if Nvidia closes a minority investment at a $30-billion-plus valuation before an acquisition is agreed, the acquisition price floor moves materially upward and Nvidia's own negotiating position becomes complicated by its status as both prospective investor and prospective buyer.

Who Pays Them, And Who They Are Up Against

Named customers, apparent concentration, and the competitive field as the record shows it.

Perplexity's own customer page claims the product is "[t]rusted by 50,000+ [enterprises]" but names no individual enterprise customers on that page itself. Named relationships do appear in case study and partner materials, though each of these is a claim by the company or a third party about a relationship that existed when the material was written.

The most concrete named relationship in the evidence is Apollo, the sales platform, where a Perplexity API case study claims Apollo users drove 46% more booked meetings using Perplexity's API. The United States Anti-Doping Agency (USADA) appears in a third-party case study roundup dated 2026 as a named user for research and operations. Ramp, the corporate finance platform, published a customer story in April 2026 in which Perplexity itself is the customer rather than the vendor — it describes Perplexity's own ten-person finance team using Ramp's tools, which is a useful data point on Perplexity's internal scale but not evidence of an enterprise sale. The Perplexity for Public Safety announcement from January 2026 positions the company as the first frontier AI firm to offer Enterprise Pro to law enforcement agencies, but no specific agency is named as a signed customer.

The evidence does not establish revenue concentration in any single account or sector. The named references span sales technology (Apollo), anti-doping compliance (USADA), and a government/public safety vertical that is still being built out. This breadth is consistent with early enterprise land-and-expand motion rather than dependence on a single anchor customer, though the evidence is too thin to rule out concentration.

On competitive positioning, the evidence consistently places Perplexity against ChatGPT, Google Gemini, Microsoft Copilot, and Anthropic Claude. Multiple comparison pieces frame Perplexity as citation-first and real-time-search-first, where ChatGPT is described as chat-first. G2 review data, as of February 2026, gives the product high marks for ease of use (95%) and ease of setup (97%). The G2 review page highlights accurate AI responses, source citations, and web research as recurring positives.

The complaint-site picture is less clean. The Better Business Bureau profile records complaints alleging that Perplexity charges full annual subscription fees for services that are subsequently degraded, and that the refund policy does not accommodate this. A Reddit post from April 2026 describes a $200 billing dispute in which the user could not reach a human support agent. These complaints cluster around billing and subscription management rather than product quality, which is a support-infrastructure problem that an acquirer would inherit on day one. For Nvidia, which would be acquiring a consumer-facing subscription business alongside the underlying technology, the gap between the product's strong usability scores and its weak billing-support infrastructure is a material operational risk that would require immediate remediation.

Warning Bells

What is actually wrong or worth worrying about here, what the evidence for it is, and what it does to this specific decision.

Perplexity AI, Inc. carries an active and growing litigation portfolio that represents the single most consequential risk for an outright acquirer. As of the report date, at least six distinct legal actions are on record, spanning copyright infringement, data-scraping abuse, and consumer privacy violations under California law.

The copyright exposure is the most structurally threatening. Dow Jones and the New York Post sued Perplexity for using copyrighted news content in its retrieval-augmented generation pipeline, as documented by BakerHostetler's case summary. Encyclopaedia Britannica filed a separate copyright and trademark infringement action in September 2025, alleging widespread, knowing, and illegal use of Britannica's content. CNN filed suit in the Southern District of New York on 28 May 2026, case 1:26-cv-04427. These are three independent, named plaintiffs with filed dockets — corroborated across primary court records and authoritative legal coverage. The concern is not speculative: Perplexity's core product depends on crawling and synthesising third-party content at scale, and the legal theory in each case directly attacks that mechanism. For Nvidia, acquiring Perplexity means acquiring these open liabilities and, more importantly, acquiring a product architecture that multiple courts are being asked to declare unlawful. If any of these cases produces an adverse ruling or injunction against Perplexity's crawling and synthesis pipeline, the product stops working in its current form on day one of ownership.

The Reddit data-scraping lawsuit survived a dismissal attempt. A Manhattan federal judge rejected most of Perplexity's motion to dismiss in July 2026, per Reuters. The underlying allegation, corroborated by a Cloudflare investigation published in August 2025, is that Perplexity deployed stealth, undeclared crawlers to evade website no-crawl directives. Amazon filed a separate action in the Northern District of California on 4 November 2025, case 3:25-cv-09514, with Perplexity's own blog post from the same date characterising Amazon's legal threat as bullying. The failed dismissal motion is a primary court record; the Cloudflare findings are an independent technical investigation. Together they establish that the scraping conduct is not merely alleged — it has been technically confirmed by a third party and has now survived initial judicial scrutiny. For Nvidia, this matters because the scraping behaviour is not a historical incident: it is the data-acquisition method underlying the product. Remediation would require either licensing agreements with a large number of content owners or a fundamental change to how the product retrieves information.

The California Invasion of Privacy Act class action filed on 31 March 2026 alleges that Perplexity unlawfully shared users' private conversation data — including sensitive personal queries — with Google and Meta for advertising purposes, as reported by Law360 and covered independently by Privado AI. Labaton's Lantern case tracker also lists a separate tracking-related lawsuit alleging Perplexity embedded tracking without disclosure. These are allegations at the class-action investigation or early filing stage — not yet adjudicated — and should be treated as unverified pending discovery. However, for Nvidia, the specific concern is that an acquisition would inherit any regulatory exposure that follows from these allegations, including potential FTC scrutiny, at a moment when Nvidia itself is under close antitrust attention.

The Ninth Circuit vacated an injunction against Perplexity's AI agents in August 2026, per the Knight Columbia Institute, which is a partial legal win. However, the existence of the injunction in the first place — and the need to appeal it — confirms that Perplexity's agentic product line has already attracted court-ordered restraint at some stage of litigation. The evidence does not establish the full procedural history of that injunction, and what could not be established from the record searched is which underlying case generated it or what conduct it originally targeted.

Consumer billing complaints are a secondary but operationally relevant concern. The Better Business Bureau complaint record includes an allegation that Perplexity knowingly charges a full-year subscription for services it intentionally degrades after purchase. A Reddit post from April 2026 describes an unauthorised charge of $200 with no path to human support, single source, unverified. These are self-selected consumer complaints and do not individually establish a pattern, but the BBB record adds a second independent signal. For Nvidia, the practical issue is that a consumer-facing subscription business with documented billing friction and an AI-only support channel creates reputational and regulatory surface area that would need immediate remediation post-close.

The Persons with Significant Control register for PERPLEXITY AI LTD (Companies House number 15470377) returned a persons-with-significant-control-statement rather than named individuals. This means the UK register, as searched, does not name any individual with significant control over the UK entity. This is a gap, not a clearance: it does not establish that no such persons exist, only that the register entry does not name them. An acquirer would need to resolve this through direct disclosure from the company and legal review of the UK entity's relationship to the US operating entity, Perplexity AI, Inc., before close.

The LEI index entry 969500KXRLQTCC2JJ496 is registered to a legal entity named "PERPLEXITY LABS" with active status. This was matched on legal name alone and was not cross-verified against the Companies House number, registered address, or any officer record. It cannot be confirmed as the same entity as PERPLEXITY AI LTD or the US operating company without checking the LEI registration's underlying entity data — specifically its registered address, jurisdiction, and associated identifiers — against the confirmed registry records.

---

THINGS THAT ARE NOT WARNINGS HERE

The Snap layoff story is not a Perplexity workforce reduction. The evidence shows that Snap announced layoffs after a deal with Perplexity collapsed — meaning Snap's headcount was affected, not Perplexity's. No layoff tracker entry for Perplexity itself was found in the layoff and workforce tracker records searched. The Blind layoff tracker page for Perplexity AI returned no populated entries in the evidence supplied. The scored signal for layoffs fired at a weight of 15/100, but the evidence does not support a finding that Perplexity has conducted internal workforce reductions.

CEO Aravind Srinivas's public remarks in March 2026 that AI layoffs are not so bad because most people dislike their jobs generated press backlash. This is a reputational signal about public communication style, not evidence of internal management failure or workforce instability at Perplexity itself.

Srinivas stated in July 2025 that Perplexity plans to remain independent and is not looking to be acquired by large technology companies. This is a stated preference, not a legal barrier. It is relevant context for deal dynamics and founder retention negotiations, but it does not constitute a structural obstacle to an acquisition.

The existing Nvidia relationship — Perplexity uses NVIDIA NeMo to fine-tune retrieval models, and as of August 2026 Nvidia was in discussions about a Perplexity investment at a valuation above $30 billion — means Nvidia already has partial visibility into the company's technical stack. This reduces some due-diligence uncertainty but does not resolve the legal exposure described above.

Corporate Standing

Whether the company legally exists, is in good standing, and is filing what it must.

Two distinct legal entities appear in the registry evidence, and Nvidia must understand both before closing.

The first is PERPLEXITY AI LTD, company number 15470377, incorporated in England and Wales on 6 February 2024, registered in London, and currently active on Companies House. This is a private limited company. The operating business that Nvidia is considering acquiring is a US-incorporated entity, Perplexity AI, Inc., which is the entity named in US court filings, funding announcements, and regulatory coverage throughout this report. The UK entity, PERPLEXITY AI LTD, appears to be a subsidiary or holding vehicle; its relationship to the US operating company — whether it holds IP, employs staff, or serves a tax or regulatory purpose — could not be established from the evidence available. Before closing, Nvidia's counsel must map the full corporate group structure to confirm what PERPLEXITY AI LTD owns and whether it forms part of the acquisition perimeter.

The second identifier is LEI 969500KXRLQTCC2JJ496, registered in the GLEIF index to a legal entity named "PERPLEXITY LABS," with status listed as active. This was found by searching the index for that exact name and nothing further was verified. It is reported here as an identifier registered to a company of that name, not as a confirmed registration of either PERPLEXITY AI LTD or Perplexity AI, Inc. What would settle it is a direct comparison of the LEI registration record's legal address, jurisdiction, and registration number against the Companies House filing for PERPLEXITY AI LTD and against any US state incorporation documents for the operating entity.

On the UK entity, the persons-with-significant-control record returned a statement rather than named individuals, meaning the register does not currently disclose who ultimately controls PERPLEXITY AI LTD. For an acquirer, this is a gap that must be closed: beneficial ownership of the UK entity needs to be confirmed and reconciled with the cap table of the US operating company before any transaction proceeds.

Perplexity AI, Inc. is a private company and files no public accounts with the SEC. No insolvency proceedings, strike-off notices, or dissolution filings were found in the record sets searched, but the absence of such findings in a web search is not a clearance — a formal search of the relevant US state registry (Delaware, where most US AI companies incorporate) and a full Companies House filing review for the UK entity would be required to confirm clean standing on both sides of the Atlantic.

Who Controls It

Registered owners and persons with significant control.

The UK Companies House register for PERPLEXITY AI LTD (company number 15470377) returns a persons-with-significant-control statement rather than a named individual or corporate entity. That means the register records that a PSC statement has been filed, but does not publicly name who holds significant control of the UK entity. This is a material gap for an acquirer: under UK law, a PSC statement is typically filed when the company has not yet identified its registrable persons with significant control, or when control is exercised through a chain that has not been fully traced to a natural person or relevant legal entity. It does not confirm the absence of a controlling party — it confirms only that one has not been named on the public register as of the date searched.

The operational parent is Perplexity AI, Inc., a US-incorporated private company. The UK entity appears to be a subsidiary or holding vehicle incorporated in February 2024, and the PSC filing at Companies House tells Nvidia nothing about who ultimately controls the US parent. That question — who holds a controlling stake in Perplexity AI, Inc. — is not answered by any filing in the evidence available here. The company has raised a total of approximately $1.72 billion across multiple rounds, with investors including Nvidia itself, Jeff Bezos's Bezos Expeditions, SoftBank, IVP, and Accel, at a reported valuation of $22.6 billion as of the most recent data available. Tracxn Co-founder and CEO Aravind Srinivas has been in post since the company's founding in August 2022, LinkedIn but his precise equity stake, and those of co-founders Denis Yarats, Johnny Ho, and Andy Konwinski, are not disclosed in any source available to this report.

For an outright acquisition, Nvidia needs three things the current evidence does not supply: the cap table of Perplexity AI, Inc. showing each shareholder's percentage and class of shares; the identity and consent requirements of any investor holding blocking rights or board approval rights under the US company's shareholder agreement; and a reconciliation of the UK PSC register to confirm whether PERPLEXITY AI LTD is wholly owned by the US parent or involves a separate ownership structure. The LEI record registered to "PERPLEXITY LABS" (LEI 969500KXRLQTCC2JJ496, status active) adds a further unresolved identity question: whether that entity is a distinct legal person with its own ownership chain, or simply an alternative registered name for the same group, cannot be determined from the index match alone and would require direct verification against the issuing authority's records.

Until the US cap table and the UK PSC register are resolved, Nvidia cannot confirm with certainty what it is buying or who has the legal authority to sell it.

How Complicated The Cap Table Is

Preference stacks, secondary sales and option overhang — the things that decide who actually gets paid at close.

Perplexity AI, Inc. is a US-incorporated private company that has raised a total of $1.72 billion across at least eleven funding rounds, reaching a reported valuation of $22.6 billion as of the most recent data available from Tracxn. Nvidia is already named as an investor in the company at the $20 billion valuation stage, per Yahoo Finance reporting from August 2026, which means Nvidia's acquisition team must account for its own existing position when modelling the cap table — a complication that is not resolved by the evidence available here.

The investor roster includes Bezos Expeditions, IVP, SoftBank, Accel, and others across seed through Series D rounds, per Tracxn and Crunchbase. Eleven rounds over roughly three years, with each round carrying its own liquidation preferences, anti-dilution provisions, and pro-rata rights, is a structurally complex stack. At a $22.6 billion valuation, the gap between headline price and what actually flows to common shareholders — founders, employees, and option holders — depends entirely on the preference terms negotiated in each round, none of which are publicly disclosed.

The founders — Aravind Srinivas, Denis Yarats, Johnny Ho, and Andy Konwinski — hold equity that has been diluted across those eleven rounds, but their precise remaining ownership is not established in any evidence available here. No secondary sales by founders or early investors have been reported in the evidence, but absence in the web index is not confirmation that none occurred.

Employee option overhang is similarly opaque. The company is actively hiring at compensation packages that include equity, with base salaries for engineering roles ranging from $220,000 to $405,000 per year plus equity per NEA's job board, and an Associate Product Manager base of $210,000 plus equity per Perplexity's own careers page. The size of the option pool, the strike prices across grant cohorts, and the vesting schedules are not in the public record.

The Companies House register for PERPLEXITY AI LTD (UK entity, company number 15470377) returned no persons-with-significant-control data in the search conducted. That absence covers only the UK-registered entity and says nothing about the US operating company's cap table.

What would settle this for acquisition purposes: the company's cap table in full, the certificate of incorporation and all amended articles, each round's term sheet or stock purchase agreement showing liquidation preference multiples and participation rights, a current option pool schedule with strike prices and vesting status, and confirmation of any secondary transactions. None of those documents are in the evidence supplied, and none can be inferred from what is available.

Is The Revenue Real, And Does It Recur

Whether what this company calls revenue is contracted and repeating, or one-off work counted once.

Perplexity's revenue is structured around two recurring streams: a consumer subscription (Perplexity Pro at $20 per month or $200 per year) and an enterprise tier (Enterprise Pro, priced up to $325 per user per month), plus a pay-as-you-go API platform. The subscription architecture is designed for recurrence — annual billing locks in consumer revenue, and enterprise contracts imply multi-seat commitments. That structural design is real. What is not established is whether customers actually renew.

The revenue numbers in circulation are analyst estimates and investor-sourced claims, not audited figures. Sacra estimates annualized revenue of $750 million as of August 2026, up from an estimated $232 million at end of 2025. TechCrunch reported in September 2025 that ARR was approaching $200 million, citing a source familiar with the company. A Reddit post from March 2026 cited investor press outlets placing ARR at "approaching $200 million" as of September 2025. These figures are consistent in direction but are not independently verified, and the growth rate implied between them — roughly 3.5x in under a year — has not been corroborated by any filed document.

The one internal financial data point in evidence is a LinkedIn post referencing $34 million in revenue against $57 million spent on AI models, with no date attached to the revenue figure. That cost structure — where model inference spend exceeds revenue — is the central question for an acquirer. Perplexity's product depends on third-party model APIs and its own inference infrastructure. If the $34 million figure predates the ARR growth claimed by Sacra, the unit economics may have improved; if it is more recent, the business is still burning on every query.

The enterprise customer base is described as 50,000-plus businesses, but named case studies — USADA, Apollo, Ramp — do not disclose contract values, durations, or renewal terms. The customer story page is a marketing surface, not a contract schedule.

For Nvidia, the specific documents that would settle this are: the contract schedule showing average enterprise contract length and renewal rates by cohort; the revenue recognition policy distinguishing subscription from usage-based API income; and the gross margin by product line net of model inference costs. None of these are public. Ask for them before any price is agreed.

Can They Fund The Term

Funding, financial standing, and how long the money in public view lasts.

Perplexity has raised a total of $1.72 billion across eleven funding rounds, reaching a reported valuation of $22.6 billion as of mid-2026, per Tracxn. The most recent publicly reported round — $200 million at a $20 billion valuation — was reported by TechCrunch in September 2025. As of August 2026, Yahoo Finance reported that Nvidia itself was in discussions to invest at a valuation exceeding $30 billion, which is directly relevant: Nvidia would be acquiring a company it was simultaneously evaluating as a portfolio target, and any such investment discussions would need to be unwound or restructured on close.

On revenue, Sacra estimates annualized revenue reached $750 million in August 2026, up from $232 million at end of 2025. As of September 2025, TechCrunch reported ARR approaching $200 million from a source familiar with the company. These are private estimates and analyst projections, not audited figures — no public filing or audited financial statement has been identified in the record sets searched.

The cost structure is material to runway. A LinkedIn post from May 2025 references financial data showing $34 million in revenue against $57 million spent on AI models alone — a figure that, if representative of the cost base relative to earlier revenue, implies significant operating losses at scale. Whether that ratio has improved as revenue has grown cannot be established from the evidence available.

What cannot be established from the record sets searched: cash on hand, burn rate, debt obligations, the terms of any credit facilities, and whether the $1.72 billion raised has been substantially consumed. No balance sheet, income statement, or audited accounts have been filed publicly — Perplexity is a private company with no SEC reporting obligation. The specific documents that would answer runway directly are audited financial statements, a current cap table with liquidation preferences, and the terms of the most recent funding round. These would need to be obtained through the acquisition data room. Without them, the gap between a $750 million revenue estimate and an unknown cost base leaves the question of how long the money lasts genuinely open.

If The Founder Leaves

How much of this company's ability to deliver sits with one or two named people.

Aravind Srinivas is the co-founder, President, and CEO of Perplexity, and the evidence makes clear that he is the company's dominant public face, strategic voice, and external representative. His LinkedIn profile shows he has held the CEO role since the company's founding in August 2022. Denis Yarats is co-founder and CTO, confirmed on LinkedIn and corroborated by Tracxn. Andy Konwinski and Johnny Ho are also named co-founders, with Ho listed as CSO by Craft.co. The four co-founders together constitute the founding technical and strategic core.

The concentration risk sits most visibly with Srinivas. Every major external statement on product direction, AI economics, and competitive positioning has come from him — the CNBC appearances on token value per watt, the Fortune interview on fear of failure, the Reddit post addressing user frustration directly. He was named to TIME's 100 Most Influential People in AI for 2024. The company's public identity is substantially his identity.

Yarats carries the technical counterpart risk. As CTO and co-founder, he holds the institutional knowledge of the retrieval and inference architecture that differentiates the product. No evidence in the record identifies a depth of engineering leadership below him that would be visible to an acquirer without internal access.

What the evidence cannot settle is the degree to which either man's departure would be contractually constrained post-acquisition. Vesting schedules, retention agreements, and any existing change-of-control provisions are not in the public record and were not returned by the web index search. The persons-with-significant-control register for the UK entity PERPLEXITY AI LTD (company number 15470377) returned no named individuals in the evidence provided. The specific documents that would answer the retention question are: employment agreements and equity vesting schedules for Srinivas and Yarats, any existing change-of-control or key-man clauses in investor agreements, and direct conversation with both founders about post-acquisition intent. Srinivas stated publicly in July 2025 that the company plans to remain independent, which is a relevant prior position for Nvidia to address directly in any acquisition negotiation.

Who Is Already Looking

Public signals that people here are moving on before you have made an offer.

Dmitry Shevelenko, Perplexity's Chief Business Officer, accepted a seat on Lazard's board of directors in September 2025 while remaining at Perplexity. An executive at his level taking an outside board role at a major financial institution is a public signal worth noting: it diversifies his professional standing independently of any acquisition outcome, and it means he arrives at a deal table with options that did not exist before.

CEO Aravind Srinivas stated publicly in July 2025 that Perplexity plans to remain independent and that he is not interested in a Big Tech acquisition. That statement predates the current Nvidia discussions by over a year, but it is the most recent on-record position from the person whose retention is most critical to the asset's value. A founder who has said he does not want to be acquired, and who has since raised at a $20 billion valuation and is now in discussions at a $30 billion-plus valuation, has had his outside options improve materially since making that statement. The gap between what he said and what the current deal implies about his willingness to engage does not resolve the retention question — it sharpens it.

Co-founder and CTO Denis Yarats has been in the role since August 2022 with no public signal of departure. Co-founder Andy Konwinski is listed on record but carries no recent public signal either way.

What the evidence does not show is any internal attrition data: voluntary departure rates, equity vesting schedules, or whether key technical staff below the co-founder level are interviewing elsewhere. The Blind layoff tracker lists Perplexity AI but returns no confirmed layoff events for the company itself, which means the workforce signal is absence of a public event, not a clean bill of health.

The specific documents that would answer this section are: current vesting cliff and acceleration terms for each co-founder and named executive, any retention agreements already in place, and direct conversation with Srinivas on whether his July 2025 independence statement still reflects his position. None of those are in the public record as of this report date.

Litigation and Regulatory Record

Court and regulator activity on the public record — and, explicitly, which of those records we searched.

The web search restricted to court and legal indexes, the web search restricted to regulator and enforcement pages, and the web search restricted to review and complaint sites were all searched for this report. The company profile store was also searched and returned nothing, which is an absence in that record, not a clearance.

The active litigation load is the most consequential single risk in this acquisition. Five distinct legal actions are on the public record, spanning copyright, data privacy, and data-scraping claims.

Reddit sued Perplexity alleging it used shell companies to circumvent technical access barriers to scrape Reddit's data. A Manhattan federal judge rejected most of Perplexity's motion to dismiss that case on 31 July 2026, meaning the lawsuit proceeds toward discovery and potential trial. Perplexity AI loses bid to toss Reddit lawsuit over data scraping Perplexity's own public response characterised the core legal issue as whether it used shell companies to circumvent technical barriers, not merely a content-rights dispute. Our Response to Reddit's Lawsuit

Encyclopaedia Britannica filed a copyright and trademark infringement lawsuit in September 2025, alleging widespread, knowing, and illegal use of its content. Britannica Files Copyright and Trademark Infringement Lawsuit Dow Jones and the New York Post have separately sued over Perplexity's use of their copyrighted news content in its retrieval-augmented generation pipeline. Dow Jones & Company, Inc. v. Perplexity AI, Inc. CNN filed its own action in the Southern District of New York on 28 May 2026. Cable News Network Inc v. Perplexity AI, Inc., 1:26-cv-04427

A proposed class action filed on 31 March 2026 alleges Perplexity unlawfully shared users' private conversations and prompts with advertisers, including Google and Meta, in violation of the California Invasion of Privacy Act. Perplexity AI CIPA Lawsuit: User Prompt-Sharing Case If certified, this class action carries statutory damages per violation that could scale with user volume.

For Nvidia specifically, the copyright cases are not merely financial contingencies — they go to whether the core product can legally continue to ingest and synthesise third-party content at scale. A court order restricting Perplexity's crawling or RAG practices would require architectural changes to the product Nvidia is buying. The CIPA class action introduces a separate regulatory surface: a finding that user prompts were shared with advertisers without consent would draw FTC attention and could conflict with Nvidia's own enterprise data commitments to customers.

What could not be established from the records searched: the current status of the Amazon.com Services LLC v. Perplexity AI, Inc. action (filed 4 November 2025 in the Northern District of California, docket 3:25-cv-09514) beyond the initial filing and service date. Full docket searches on each active case, and any settlement discussions or licensing negotiations with the plaintiff publishers, would be required before close.

What Breaks On Close

Change-of-control clauses, key contracts and licences that do not survive a change of owner.

The most consequential break-on-close risk visible in the evidence is the Amazon dispute. In November 2025, Perplexity published a post titled "Bullying is Not Innovation" describing an aggressive legal threat from Amazon demanding that Perplexity prohibit Comet users from running AI assistants on Amazon infrastructure. A formal lawsuit followed: Amazon.com Services LLC v. Perplexity AI, Inc., 3:25-cv-09514 (N.D. Cal.), filed November 4, 2025. Nvidia acquiring Perplexity would place Amazon in the position of active litigant against a direct competitor. Whether any underlying infrastructure or API agreement between Perplexity and AWS contains a change-of-control termination right — or whether Amazon would accelerate its legal posture on close — cannot be determined from public filings. That answer requires production of every AWS service agreement and any side letter tied to the Comet product.

The content licensing picture is equally exposed. Perplexity faces active copyright suits from Dow Jones and the New York Post, Britannica, CNN, and Reddit — the last of which survived a dismissal motion in July 2026. None of these plaintiffs have settled on terms that are public. Any content access arrangement — formal licence or informal tolerance — that is personal to Perplexity as an independent company could be renegotiated or terminated the moment a strategic acquirer, particularly one already competing in AI infrastructure, takes ownership. Publishers and data owners routinely insert change-of-control consent clauses precisely to preserve that leverage.

The CIPA class action filed March 31, 2026 alleges that Perplexity shared user prompt data with advertisers including Google and Meta. If Perplexity holds advertising or data-sharing agreements with those parties, those agreements almost certainly contain change-of-control provisions that would require consent — or allow termination — on acquisition by Nvidia, a direct competitor to both.

What cannot be established from any searched record: the actual text of any customer enterprise contract, any model or API licence from third-party providers, any publisher revenue-sharing agreement (Perplexity announced a publisher revenue-share programme in August 2025 but terms are not public), and whether the Persons with Significant Control register for UK entity PERPLEXITY AI LTD (Companies House number 15470377) contains any shareholder consent thresholds triggered by a full acquisition. The specific documents required before close are: a full contract schedule with change-of-control flags, every infrastructure and model licence, and the cap table consent mechanics for both the US operating entity and the UK registered entity.

What Happens To You If They Are Bought

Whether this company looks like an acquisition target, and what a change of control would do to your price, your data and your contract.

Aravind Srinivas stated publicly in July 2025 that Perplexity plans to remain independent and that he is not interested in a Big Tech acquisition. That statement was made before Nvidia was reported in August 2026 to be discussing a Perplexity investment at a $30 billion-plus valuation. The gap between those two data points matters: a founder who resists acquisition while accepting a strategic investor from the same acquirer class is a different posture than one who has ruled out a transaction entirely. Whether Srinivas's stated independence preference extends to a full acquisition by Nvidia specifically could not be established from the evidence available as of 2026-09-08.

The UK Companies House register for PERPLEXITY AI LTD (company number 15470377) shows one person with significant control on record. A single concentrated holding is the ownership shape that can execute a sale quickly, without needing to assemble a coalition of shareholders. It also means the price of the company is, in practice, the price of one person's consent. If that person's view of independence shifts — whether because of valuation, personal liquidity, or competitive pressure — the transaction can move fast. The flip side is that the same concentration gives that person the ability to block a deal indefinitely regardless of what other investors want.

What the evidence does not resolve is the full cap table. The Companies House filing covers the UK entity incorporated in February 2024; the operating entity is a US company. The persons-with-significant-control statement for the UK entity does not establish who holds what in the Delaware parent, and no cap table, shareholder agreement, or investor rights agreement was available in the record sets searched. That document set — specifically the investor rights agreement and any right-of-first-refusal or co-sale provisions — is what would determine whether existing investors such as SoftBank, IVP, or Bezos Expeditions hold blocking rights, drag-along rights, or information rights that survive a change of control.

For Nvidia, the practical exposure on a change-of-control scenario runs in two directions. First, Perplexity's active litigation stack — including the Reddit data-scraping case that survived dismissal in July 2026 and multiple copyright suits — likely contains indemnification and assignment provisions that would need to be reviewed for change-of-control triggers. Second, enterprise and publisher contracts, including the revenue-sharing arrangement with publishers announced in August 2025, may carry termination rights on acquisition. The specific documents needed to answer both questions are the material customer contracts and the publisher revenue-sharing agreements, neither of which is in the public record.

What Has Happened Lately

Dated public events, most recent first.

The most consequential recent development for Nvidia as a prospective acquirer is the valuation trajectory. Nvidia is reported to be in discussions about a Perplexity investment at a valuation above $30 billion as of August 2026, up from the $20 billion valuation at which Perplexity reportedly raised $200 million in September 2025. Tracxn records total funding of $1.72 billion across eleven rounds and a current valuation of $22.6 billion. The gap between those figures and the $30 billion-plus discussion number is material to any acquisition price negotiation.

On revenue, Sacra estimates Perplexity reached $750 million in annualized revenue in August 2026, up from $232 million at the end of 2025. That growth rate is the primary justification for the valuation step-up, but Sacra is a single secondary source and no audited figures are available in the evidence.

The CEO's public posture on acquisition is directly relevant. In July 2025, Aravind Srinivas told CNBC that Perplexity plans to remain independent and that he is not interested in a Big Tech acquisition. That statement predates the reported Nvidia investment discussions by over a year, but it establishes a baseline position Nvidia would need to work against or around in any deal negotiation.

In March 2026, Srinivas drew sustained public criticism after remarks to Fortune that AI-driven layoffs are not harmful because most people dislike their jobs. The backlash was covered across multiple outlets. For Nvidia, the reputational exposure here is the post-acquisition integration risk: a founder whose public statements on workforce reduction generated significant negative press is a liability if Nvidia needs to manage headcount changes following a close.

On the legal front, a Manhattan federal judge in late July 2026 rejected most of Perplexity's bid to dismiss Reddit's data-scraping lawsuit, meaning that case proceeds to discovery. A CIPA class action alleging unauthorized sharing of user prompts with advertisers was filed on March 31, 2026. CNN filed suit in May 2026. These are active, not resolved.

In September 2025, Dmitry Shevelenko, Perplexity's Chief Business Officer, was appointed to Lazard's board of directors, giving a key commercial executive an external board seat that would require unwinding or managing on acquisition.

What This Report Does Not Cover

The record sets that were not searched, named, so no absence here reads as a clearance.

Searched for this report: Web index (whatever the index currently holds — no fixed period); Web search for leadership changes (whatever the index currently holds); Web search restricted to status pages and breach coverage (current status-page history and press coverage); Web search restricted to review and complaint sites (current review-site content — self-selected by the people who wrote it); Web search restricted to court and legal indexes (whatever those indexes currently expose to search — not a docket search); Web search restricted to layoff and workforce trackers (tracker coverage, typically 2020 onward); Web search for funding, runway and insolvency events (whatever the index currently holds); Web search restricted to regulator and enforcement pages (current regulator publications reachable by search); Web search restricted to filing registries (current registry pages reachable by search); GLEIF LEI index (point-in-time register, current filings only); Our own company profile store (whatever we have previously gathered for this company); Your own contacts (your account); Companies House (UK register) (current register — UK-registered entities only); Web search for named leadership (whatever the index currently holds — a current page may name a person who has since left); Web search for funding, valuation and revenue on the record (whatever the index currently holds — private figures are reported, not filed); Web search for named customers and case studies (current pages — a case study outlives the contract it describes); Web search for what they sell and what it costs (current pages — pricing changes without notice); Web search for the field they compete in (whatever the index currently holds); Web search for open roles and headcount (current job-board pages); Contact records for this domain (current records — a person who has left may still appear). Where a section above says nothing was found, that is an absence in these records over these periods. It is not a finding that there is nothing to find.

NOT searched, and therefore unchecked: Company enrichment index — this record type is not part of this report yet, so nothing was searched for it. Anything held only in those records is outside what this report can speak for.

Sources

Every source this report rests on, so any sentence in it can be checked.

Every source cited above, deduplicated and grouped. Each was read as it stood on 2026-09-08; a page can change after that date.

perplexity.ai

reddit.com

finance.yahoo.com

linkedin.com

courtlistener.com

tracxn.com

bloomberg.com

learn.g2.com

fortune.com

sacra.com

techcrunch.com

reuters.com

blog.cloudflare.com

bakerlaw.com

corporate.britannica.com

privado.ai

nvidia.com

find-and-update.company-information.service.gov.uk

en.wikipedia.org

techjacksolutions.com

docs.perplexity.ai

gsb.stanford.edu

craft.co

theorg.com

lazard.com

digitaldefynd.com

ramp.com

zapier.com

bairesdev.com

g2.com

bbb.org

law360.com

lantern.labaton.com

knightcolumbia.org

mashable.com

crunchbase.com

careers.nea.com

finout.io

cnbc.com

time.com

teamblind.com

61 distinct sources across 41 domains.

Confidential — prepared for due diligence. Generated from public sources. Use only where the subject has consented or a legitimate diligence interest applies.

MentionFox