Perplexity AI

  Executive Briefing

Perplexity AI is a San Francisco-based AI search company founded in August 2022 that has built what it calls an "answer engine" — a system that retrieves live information from the web and synthesizes it into concise, cited responses rather than returning lists of hyperlinks. In less than four years it has grown from a research prototype into one of the fastest-scaling AI product companies in the world, reporting an annualized revenue run rate of approximately $500 million by April 2026 and roughly 780 million monthly queries by mid-2025, representing 335% year-over-year growth.¹²

The company is led by its four co-founders: Aravind Srinivas (CEO), a former OpenAI research scientist and UC Berkeley PhD who conceived the product as a realization of Larry Page's original vision for direct-answer search; Denis Yarats (CTO), a former Meta AI researcher who connected with Srinivas after they independently published nearly identical research papers; Johnny Ho (Chief Strategy Officer), formerly an engineer at Quora; and Andy Konwinski (President), a co-founder of Databricks and UC Berkeley alumnus. Chief Business Officer Dmitry Shevelenko, who joined in October 2023 from a career spanning Uber, LinkedIn, and Meta, has led the company's commercial and partnership operations and joined the board of Lazard in September 2025.³

Perplexity occupies an unusual position in the AI landscape: it is not primarily a model developer but rather an inference-and-retrieval orchestration company that combines proprietary fine-tuned models (the Sonar family) with third-party frontier models and real-time web grounding to deliver a differentiated search experience. Its strategy has evolved rapidly from pure consumer search into a broader platform play — encompassing an agentic AI product (Perplexity Computer), an AI-powered browser (Comet), a developer API, and deep distribution partnerships with telecoms and device OEMs. The company is targeting a 2028 IPO, as confirmed by Srinivas in a June 2026 CNBC interview.⁴

  At a Glance

  Origins & Founding

Perplexity AI was incorporated in August 2022 in San Francisco, with the founding team formally joining forces in July 2022 around a shared thesis: that large language models combined with real-time web retrieval could replace keyword search with a fundamentally superior mode of information access. Aravind Srinivas, who grew up in Chennai, India, earned a master's degree from IIT Madras, and completed a PhD in computer science at UC Berkeley focusing on reinforcement learning, had spent formative research stints at DeepMind, Google Brain, and OpenAI before concluding that the AI field's most consequential near-term application was search. He was reportedly inspired by the belief that Larry Page's original ambition for Google — delivering direct answers to users — had never been fully realized, and that AI could finally accomplish it.⁵

Denis Yarats brought complementary technical depth as a PhD-level AI researcher from Facebook/Meta and NYU; the two connected after independently publishing nearly identical research papers, which they took as a sign of intellectual alignment. Johnny Ho, a Quora engineer, contributed product and engineering experience building at the intersection of knowledge and community. Andy Konwinski brought institutional credibility and capital connections as a co-founder of Databricks, one of Silicon Valley's most successful data infrastructure companies. Together they launched Perplexity Ask — the core conversational search product — in December 2022 and quickly attracted early interest from prominent angel investors including Jeff Bezos, Andrej Karpathy, Nat Friedman, and Naval Ravikant, raising a seed round of approximately $25.6 million in early 2023.⁵⁶

  History & Timeline

    2022–2023: Launch and early product discovery

Perplexity launched its first public product on 7 December 2022 with Perplexity Ask, followed eight days later by Bird SQL — a Twitter/X search interface that translated natural language queries into SQL using OpenAI Codex. Bird SQL served as the company's most visible early demo and established its style: using AI to make existing information systems dramatically more accessible. When Twitter ended free API access in February 2023, the Bird SQL experiment closed, and the team pivoted fully to AI search.⁵

By March 2023 the product had reached approximately 2 million monthly active users, and the company closed its seed round. The early product confirmed the core insight: users preferred cited, conversational answers to blue-link results, and willingness to pay was detectable even at this stage. The company introduced a paid Pro subscription tier to begin converting the early adopter base.

    2024: Unicorn, enterprise, and the search war begins

January 2024 brought a $73.6 million Series B at an approximately $520 million valuation, followed in April 2024 by a $165 million Series C that pushed Perplexity past a $1 billion valuation — achieving unicorn status in under two years.¹ The company launched Enterprise Pro at $40 per seat per month, targeting organizations that wanted internal knowledge search capabilities alongside the public web product.

June 2024 marked a significant commercial inflection: a $250 million Series D led by SoftBank Vision Fund 2 at a $3 billion valuation, combined with reported monthly query volume of 169 million, signaled that Perplexity had escaped the niche AI-curiosity category and was competing for mainstream search traffic.¹ The company also began attracting legal and reputational scrutiny: Dow Jones and the New York Post filed a copyright infringement lawsuit in June 2024, and a Forbes editorial accused Perplexity of plagiarism; a contemporaneous Wired investigation documented that Perplexity had used undisclosed IP addresses to crawl sites that had opted out via robots.txt.⁷

In July 2024, the company launched a publisher revenue-sharing program in an attempt to reframe the content relationship. November 2024 saw the introduction of Perplexity Shopping Hub and the company's first advertising product: Sponsored Follow-Up Questions, which placed paid units inside the answer flow rather than as traditional display ads.

    2025: Scale, agentic expansion, and global distribution

December 2024 closed out one of the most active growth years in recent AI history: Perplexity raised $500 million at a $9 billion valuation in a round led by IVP, crossing approximately $1 billion in cumulative capital raised, and acquired Carbon, an enterprise data connector startup, to strengthen its internal knowledge search capabilities.¹

January 2025 brought two audacious strategic moves: Perplexity submitted an unsolicited bid to merge with TikTok US ahead of a threatened ban — later revised to offer the US government up to 50% ownership in a new entity valued above $300 billion at IPO — and launched Perplexity Assistant, a multimodal AI assistant product.⁸ The TikTok bid was widely interpreted as a publicity play rather than a financially credible acquisition, and it did not succeed.

In February 2025, Perplexity released R1 1776, an open-source post-trained version of DeepSeek R1 with censorship removed, publishing the weights on HuggingFace and framing the release as a statement about information freedom.⁹ The same month, the company launched the Sonar model family via API, built atop Meta Llama 3.3 70B and fine-tuned for real-time web-grounded generation. May 2025 saw reported monthly query volume reach 780 million.²

June 2025 brought a $500 million Series E led by Accel at a $14 billion valuation, Andy Konwinski pledged $100 million toward a new AI research institute, and the copyright litigation wave intensified with suits from the BBC, Nikkei, and Asahi Shimbun.⁶¹⁰ July 2025 marked the launch of Comet, an AI-powered Chromium-based browser with an agentic sidebar, initially available to Max subscribers at $200 per month. A landmark distribution deal with Bharti Airtel offered Perplexity Pro free to Airtel's reported 360 million customers in India for 12 months.¹¹ A further $100 million was raised at an $18 billion valuation from investors including Nvidia, SoftBank Vision Fund 2, NEA, and IVP.¹²

August 2025 brought one of the more attention-generating strategic gestures of the year: Perplexity submitted an unsolicited all-cash bid of $34.5 billion for Google's Chrome browser amid DOJ antitrust proceedings requiring Google to consider divesting Chrome, pledging $3 billion in Chrome/Chromium investment over two years. The bid was generally received as strategically audacious and brand-building rather than immediately financially credible.¹³ Comet was made available worldwide free of charge in October 2025, and the company paused onboarding of new advertising clients, signaling a pivot toward subscription and enterprise revenue.

December 2025 saw Cristiano Ronaldo join as an investor (stake undisclosed) and The New York Times copyright suit escalate. In January 2026, Perplexity signed a $750 million, three-year commitment to Microsoft Azure, gaining access to OpenAI, Anthropic, and xAI models via the Microsoft Foundry program — a significant infrastructure and capability alignment.¹

    2026: Agentic AI and the path to IPO

February 25, 2026 marked the launch of Perplexity Computer, an agentic AI product that orchestrates 19 AI models to execute multi-step tasks — writing, coding, email, and integrations with Gmail, Slack, GitHub, Notion, and Salesforce — initially for Max subscribers before expanding rapidly to enterprise.² The company simultaneously discontinued AI-integrated advertising, extended free Pro access to students, military veterans, and government employees, and confirmed a 2028 IPO target. By April 2026, annualized revenue was reported at $500 million, representing approximately 335% year-over-year growth from $232 million in full-year 2025.² CNN filed a copyright suit in May 2026.⁷

  Mission, Philosophy & Research Agenda

Perplexity's stated mission is to build the world's best answer engine — replacing the traditional keyword search-and-links paradigm with AI-synthesized responses that cite their sources in real time. CEO Aravind Srinivas has articulated this as the realization of Larry Page's original Google vision: a system that directly answers questions rather than generating lists of links for users to investigate themselves. The company frames AI search as an antidote to SEO-spam-polluted results and positions itself as a tool for genuinely curious users who value factual accuracy and source transparency over engagement-optimized ranking.

Philosophically, Perplexity sits closer to the product-first, rapid-iteration end of the AI spectrum than the research-publication-first mode of labs like Anthropic or OpenAI. Its research agenda is focused and applied: (1) retrieval-augmented generation (RAG) at scale — building systems that ground language model outputs in real-time web retrieval and produce consistently cited answers; (2) proprietary model fine-tuning — developing the Sonar family specifically optimized for web-grounded factuality, citation quality, and speed rather than general capability; (3) agentic workflows — extending the answer-engine paradigm from answering questions to taking multi-step actions on behalf of users, realized in Perplexity Computer; and (4) information freedom — demonstrated by the R1 1776 open-source release, which removed censorship from DeepSeek R1 and published weights publicly. The company does not maintain a large-scale academic publishing program but releases model weights when it serves the open-source narrative and publishes technical detail through the Perplexity blog.

  Research & Publications

Perplexity is primarily a product-driven organization rather than a fundamental research institution; it does not compete with frontier labs on pretraining scale or publish extensively in academic venues. Its technical contributions concentrate in three areas.

The Sonar model family (launched February 2025) represents Perplexity's core proprietary research output: models fine-tuned on Meta Llama 3.3 70B using RLHF on user interaction data, optimized specifically for real-time search-augmented generation. Serving Sonar through Cerebras inference infrastructure, Perplexity has demonstrated approximately 1,200 tokens per second — positioning the product on speed as well as quality.² The family includes Sonar, Sonar Pro (multi-step search with extended context), Sonar Reasoning Pro (integrating DeepSeek R1 capabilities for logical inference), and Sonar Deep Research (a multi-step agentic research model capable of generating structured deliverables including spreadsheets, dashboards, and presentation decks, significantly updated in early 2026).

R1 1776 (released 18 February 2025) is the company's most prominent open-source contribution: a post-trained version of DeepSeek R1 with CCP censorship removed across approximately 300 topic categories. The model was trained using Nvidia NeMo 2.0 on a dataset of approximately 40,000 multilingual prompts with a multilingual censorship classifier, and its weights are publicly available on HuggingFace.⁹ The release was framed as demonstrating that uncensored, open research AI is achievable and was timed to coincide with significant public debate about DeepSeek's origin and possible data restrictions.

Perplexity Computer (February 2026) introduced a hybrid local-cloud inference orchestrator — a system capable of automatically routing tasks between on-device and cloud inference — unveiled in technical detail at Computex 2026 in June 2026. This architecture is significant as a model for privacy-preserving agentic AI that reduces dependence on cloud round-trips for sensitive tasks.

  Models & Products

Perplexity ships its technology across several distinct product surfaces: a consumer web and mobile search interface, a developer API, an enterprise subscription tier, an agentic computer-use product, and a browser. The underlying model family and API serve as both the internal inference backbone and an external revenue line.

Sonar model family (API, launched February 2025):

• Sonar — real-time search-augmented generation model; the base offering for developers, built on Meta Llama 3.3 70B and optimized for web-grounded factuality and citation quality.
• Sonar Pro — balanced professional-tier model with multi-step search, extended context retention, and higher throughput for complex queries.
• Sonar Reasoning Pro — reasoning model integrating DeepSeek R1 capabilities for logical inference and technical problem-solving, grounded in live web retrieval.
• Sonar Deep Research — exhaustive multi-step research model capable of generating structured deliverables; significantly updated in early 2026.
• R1 1776 — open-source post-trained model (February 2025); censorship-removed version of DeepSeek R1 available on HuggingFace.

Consumer and enterprise products:

• Perplexity Search — the core conversational answer engine (launched December 2022); available free with rate limits.
• Perplexity Pro ($20/month) — paid subscription with access to advanced models and 600 Pro searches; extended free to students, military veterans, and government employees from February 2026.
• Enterprise Pro ($40/seat/month) — B2B offering with internal knowledge search capabilities; Carbon integration for enterprise data connectors.
• Perplexity Max ($200/month) — top-tier subscription with frontier model access and Perplexity Computer included.
• Perplexity Assistant (January 2025) — multimodal AI assistant product.
• Perplexity Computer (February 2026) — agentic AI product orchestrating 19 AI models to execute multi-step tasks with integrations into Gmail, Slack, GitHub, Notion, and Salesforce; launched for Max subscribers, expanded to enterprise.
• Comet Browser (July 2025; free worldwide October 2025) — AI-powered Chromium-based browser with an agentic sidebar for multi-step research and task workflows.
• Perplexity Finance (October 2024) — module for stock prices, earnings data, and financial queries.
• Perplexity Shopping Hub (November 2024) — product search and commerce features with PayPal/Venmo integration.

➡️ See individual model cards for specifications, pricing, and benchmark results.

  People

Perplexity's executive team is notably compact relative to its valuation and query volume, reflecting the company's emphasis on lean product-engineering operations. Aravind Srinivas (CEO) is the company's most public voice and primary technical and strategic authority; his background at OpenAI, Google Brain, and DeepMind gives him credibility across the model and product communities. Denis Yarats (CTO) oversees the core model and inference systems that underpin every product; his research background at Meta AI grounds the Sonar fine-tuning program. Andy Konwinski (President), while maintaining external ventures — he launched the venture fund Laude in 2024 and pledged $100 million toward a new AI research institute in June 2025 — remains publicly associated with Perplexity's leadership, though his day-to-day operational involvement is not fully confirmed.⁶ Johnny Ho (Chief Strategy Officer) leads strategic planning. Dmitry Shevelenko (Chief Business Officer, from October 2023) has been the primary driver of commercial partnerships, enterprise deals, and distribution agreements; his appointment to Lazard's Board of Directors in September 2025 reflects his standing in the broader business community.³

Notable investors with prior AI research credentials who have provided informal advisory value include Andrej Karpathy (seed investor; former Tesla and OpenAI), Nat Friedman (seed investor; former GitHub CEO), and Naval Ravikant (seed investor; AngelList).

  Funding, Ownership & Business

Perplexity has raised approximately $1.5–1.72 billion across its full funding history (figures vary by source; treat as reported/estimated), growing from a $25.6 million seed round to a reported valuation of approximately $21–22.6 billion by January 2026 in under four years.¹ The funding trajectory is one of the steeper in AI-sector history.

Business model: Perplexity operates a tiered freemium model across consumer, professional, and enterprise segments. Consumer subscriptions (Pro at $20/month, Max at $200/month) generate the largest volume of recurring revenue. Enterprise Pro ($40/seat/month) targets knowledge workers and teams. The Sonar API serves developers and other AI products on a consumption basis. Distribution partnerships — Airtel, SoftBank Corp, Deutsche Telekom, Samsung, Motorola — generate revenue through commercial terms with OEMs and telecoms rather than direct user billing. The company launched advertising (Sponsored Follow-Up Questions) in November 2024 but paused new client onboarding in October 2025 and formally discontinued AI-integrated advertising in February 2026, indicating a strategic preference for subscription and enterprise revenue quality over advertising volume. A $750 million, three-year commitment to Microsoft Azure signed in January 2026 provides infrastructure and access to OpenAI, Anthropic, and xAI models via the Foundry program.¹

Annual revenue has grown sharply: approximately $35 million ARR in August 2023, $150 million ARR by July 2025, $232 million full-year revenue in 2025, and approximately $500 million annualized by April 2026 — the last being a company-reported figure, not independently audited.²

  Partnerships & Ecosystem

Perplexity has built an unusually broad distribution and infrastructure partnership network for a company of its age, reflecting a deliberate strategy of embedding its product at the device and network layer rather than relying solely on direct-to-consumer acquisition.

Telecom and device OEM partnerships provide Perplexity with hundreds of millions of potential end-users through pre-installation and bundling agreements:

• Bharti Airtel (July 2025) — Airtel's reported 360 million Indian customers offered free 12-month Perplexity Pro subscriptions; significantly boosted the Indian user base and positioned Perplexity for international growth.¹¹
• SoftBank Corp (Japan) — distribution partnership delivering Perplexity to SoftBank's mobile customers; SoftBank Vision Fund 2 is also a financial investor.
• Deutsche Telekom — integration announced at Mobile World Congress 2025 into Deutsche Telekom's AI-powered phone.
• Motorola (April 2025) — Perplexity assistant pre-installed on Motorola smartphones.
• Samsung — Samsung NEXT investment (2024); reported advanced discussions for pre-installation on Galaxy S26 as a default assistant option (commercial terms unconfirmed); Samsung Perplexity TV App available on 2025 Samsung TVs.

Infrastructure and model access:

• Microsoft Azure — $750 million, three-year commitment (January 2026) covering compute and access to frontier models via Microsoft Foundry.¹
• Cerebras — inference infrastructure powering Sonar's approximately 1,200 tokens-per-second throughput.
• Nvidia — both a financial investor and technical partner; NeMo 2.0 framework used for R1 1776 training.
• Brave Search — Perplexity licenses Brave's search index as a core web retrieval data source.

Commerce and content:

• PayPal / Venmo — commerce integration enabling in-answer purchasing within Shopping Hub.
• TripAdvisor — travel module partnership (January 2025).
• Le Monde — early publisher partner in the revenue-sharing program launched July 2024.

  Compute & Infrastructure

Perplexity's compute strategy is differentiated from frontier pretraining labs in that its primary requirement is inference throughput rather than training scale. The company does not pretrain foundation models from scratch; its computational investment is concentrated in fine-tuning (the Sonar family) and real-time inference at scale. The Cerebras partnership is central to achieving the approximately 1,200 tokens-per-second inference speeds that differentiate the Sonar API from competing offerings on latency-sensitive search use cases.²

For training and fine-tuning, Perplexity uses Nvidia NeMo 2.0 — as demonstrated in the R1 1776 post-training process — and the Microsoft Azure commitment provides the long-term cloud infrastructure backbone.¹⁹ The hybrid local-cloud inference orchestrator introduced with Perplexity Computer (Computex 2026 technical disclosure) represents a further evolution: the system automatically routes tasks between on-device compute and cloud inference based on privacy requirements, latency, and task complexity.

  Notable Events & Controversies

Copyright litigation wave is the company's most significant ongoing legal exposure. Multiple major publishers have filed suits alleging unauthorized scraping, reproduction of paywalled content, and disregard for robots.txt directives:⁷

• Dow Jones / New York Post — filed June 2024
• The New York Times — filed October 2024; escalated December 2025
• BBC — filed June 2025
• Nikkei / Asahi Shimbun — filed August–September 2025
• Reddit — filed October 2025
• CNN — filed May 2026⁷

A Wired investigation (June 2024) documented that Perplexity used undisclosed IP addresses to crawl sites that had blocked its known crawlers via robots.txt. Cloudflare's CEO publicly accused Perplexity in August 2025 of using crawler tactics resembling those of bad actors; Perplexity responded by calling this a "charlatan publicity stunt." Amazon obtained a preliminary injunction blocking Perplexity Computer's agent from accessing Amazon properties over alleged unauthorized automated access. Most publisher cases remain in active litigation as of June 2026; outcomes are uncertain. The company has attempted to address the underlying tension through a publisher revenue-sharing program (launched July 2024) and commercial licensing discussions, but the gap between its legal posture and its operational content practices has not been fully resolved.

TikTok bid (January 2025): Perplexity submitted an unsolicited proposal to merge with TikTok US ahead of a threatened ban, later revised to offer the US government up to 50% ownership in a new entity at IPO above $300 billion. The bid did not succeed and was widely read as a high-profile publicity maneuver that succeeded in generating significant coverage of the Perplexity brand.⁸

Chrome acquisition bid (August 2025): Perplexity submitted an unsolicited $34.5 billion all-cash bid for Google's Chrome browser amid DOJ antitrust proceedings that required Google to consider Chrome divestiture. The company pledged $3 billion in Chrome/Chromium investment over two years. The bid was seen as strategically bold but financially speculative given Perplexity's own valuation at the time.¹³

  Competitive Position

Perplexity competes across two distinct layers: as an AI search product against Google Search (approximately 90% global search market share), Microsoft Bing (with ChatGPT integration), and smaller AI search rivals such as You.com; and as an AI assistant and agent against OpenAI ChatGPT (approximately 700 million weekly users), Anthropic Claude, xAI Grok, and the broader agentic AI market.

Its core differentiators include: real-time web grounding with citations on every answer (a structural advantage over context-window-limited general assistants); inference speed (Sonar at approximately 1,200 tokens per second via Cerebras); source transparency as a deliberate design principle; OEM and telecom distribution embedding Perplexity into devices and networks reaching hundreds of millions of users; the Comet browser as a moat-building play to control the browsing layer; and Perplexity Computer as an early agentic capability with deep enterprise integrations. At approximately 100 million reported monthly active users as of 2026, it operates at a significant scale disadvantage compared to Google's billions but a meaningful scale advantage over most AI-native search competitors.²

Key risks include: the threat of foundational model commoditization as GPT-5, Gemini 2.5, and others improve directly at question answering; copyright litigation creating potential injunctive or operational risk; dependence on third-party model providers for frontier capability; and Google's distribution dominance through Android, Chrome, and the Safari default deal.

  Outlook & Roadmap

Perplexity is targeting a 2028 IPO, confirmed by CEO Aravind Srinivas in a June 2026 CNBC interview, explicitly framed as independent of the listing timelines of OpenAI or Anthropic.⁴ Chief Business Officer Dmitry Shevelenko has stated publicly that the 2028 target has allowed the company to prioritize structural growth over short-term metric optimization.

Strategic priorities for the remainder of 2026 and the path to IPO include:

1. Perplexity Computer expansion — scaling the agentic AI product from Max subscriber early adoption to broad enterprise deployment, with the hybrid local-cloud inference architecture enabling privacy-preserving enterprise use cases.
2. Device and OEM distribution — completing and expanding deals with Samsung (Galaxy S26 pre-installation, commercial terms unconfirmed), Motorola, and other Android OEM partners to embed Perplexity at the OS level.
3. India and international growth — the Airtel partnership has positioned Perplexity to compete for the Indian internet market as it moves toward AI-first search; Perplexity operates in 238 countries across 46 languages.
4. Comet browser — building long-term distribution and data moat through browser-layer control, with agentic capabilities deepening the value proposition beyond simple search.
5. Healthcare expansion — Perplexity Health is a named strategic priority, applying the answer-engine model to medical information.
6. Publisher relationship resolution — ongoing litigation creates both financial and reputational risk; expanded revenue-sharing and licensing programs are the most likely resolution pathway.
7. Proprietary model development — reducing dependence on third-party foundation model providers through continued Sonar family development and fine-tuning.
8. Revenue quality — the discontinuation of advertising and the focus on subscriptions and enterprise contracts reflects a deliberate choice to build toward a higher-multiple IPO profile.

The competitive and legal environment remains fluid. The company's rapid ARR growth, deep distribution partnerships, and expanding product surface make it one of the most closely watched AI companies ahead of a potential 2028 public offering; the copyright litigation exposure and dependence on third-party models are the two most significant structural uncertainties that will shape its trajectory.

---

  References

1. Perplexity AI — Wikipedia
2. Perplexity — Sacra Research
3. Lazard Appoints Dmitry Shevelenko — BusinessWire
4. Perplexity CEO confirms 2028 IPO — CNBC
5. Perplexity — Contrary Research
6. Databricks / Perplexity co-founder pledges $100M — TechCrunch
7. CNN sues Perplexity AI — CNN
8. Perplexity revised TikTok merger proposal — CNBC
9. Perplexity releases censorship-free DeepSeek R1 — WinBuzzer
10. Perplexity $18B valuation funding — TechFundingNews
11. Airtel Perplexity deal — StartupNews
12. Perplexity sees India as shortcut vs OpenAI — TechCrunch
13. Perplexity $34.5B bid for Chrome — SiliconAngle

  References

1. Funding rounds, valuation history, and total raised figures vary across sources; treated as reported/estimated — Wikipedia: Perplexity AI; Contrary Research. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹

2. ARR, query volume, and revenue growth figures are company-reported or press-reported, not independently audited — Sacra Research. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹

3. Dmitry Shevelenko background and Lazard appointment — BusinessWire, Sep 2025. ↩ ↩²

4. 2028 IPO confirmation — CNBC, Jun 2026. ↩ ↩²

5. Founding story and seed round — Contrary Research. ↩ ↩² ↩³

6. Andy Konwinski $100M pledge and seed investor details — TechCrunch, Jun 2025. ↩ ↩² ↩³

7. Copyright litigation — CNN suit, May 2026; Wikipedia: Perplexity AI. ↩ ↩² ↩³ ↩⁴

8. TikTok bid — CNBC, Jan 2025. ↩ ↩²

9. R1 1776 release — WinBuzzer, Feb 2025. ↩ ↩² ↩³

10. Series E-3 ($18B valuation) — TechFundingNews. ↩

11. Airtel partnership — StartupNews, Jul 2025. ↩ ↩²

12. India strategy — TechCrunch, Jul 2025. ↩

13. Chrome acquisition bid — SiliconAngle, Aug 2025. ↩ ↩²