Perplexity Raises $300M, Aiming for Your Business Next

SoftBank has decided to invest $300 million in Perplexity, bringing the value of this conversational search engine to an impressive $9 billion. However, the real story isn't the size of the check but the strategy behind it. While ChatGPT and Claude battle for individual users, Perplexity has made a bold move: targeting businesses directly with an enterprise product that promises to index all your internal knowledge and respond like an experienced analyst.

Photo: Igor Omilaev on Unsplash

It's worth noting that this is an aggressive bet, as Perplexity isn't the biggest or most well-known player. Although it has 15 million active monthly users, it's a far cry from ChatGPT's 300 million. Nevertheless, SoftBank, along with NVIDIA and Jeff Bezos, are betting on a future where AI isn't just for occasional users but for teams managing millions of documents, support tickets, or legal reports.

Why $300M Isn't a Regular Growth Round

Raising $300 million in a Series C usually indicates a company is looking to scale sales, enter new markets, or expand infrastructure. However, Perplexity is doing something different: buying time and access. SoftBank didn't just provide capital; it brought distribution. The Japanese firm owns stakes in numerous companies, now immediate potential clients. NVIDIA offered subsidized chips and priority access to its H100, while Bezos provided institutional credibility and AWS connections.

In my experience, this round seems more of a strategic pact than a mere investment. Perplexity needed enterprise validation and GPU infrastructure access at a time when OpenAI, Anthropic, and Google dominate the corporate landscape. SoftBank didn't just acquire shares; it secured an early positioning in a still-emerging market.

With a valuation of $9 billion, Perplexity finds itself in an awkward spot: too high to be easily acquired and too low to compete directly with OpenAI ($80 billion) or Anthropic ($18.4 billion after its latest round). This forces Perplexity to find a niche where it can thrive without taking on everyone.

Perplexity Enterprise: The Real Product Behind the Round

Photo: Luke Jones on Unsplash

Three weeks after closing the round, Perplexity launched Perplexity Enterprise. It's not just a chatbot, but an information retrieval system that can index platforms like Confluence, Notion, Google Drive, Salesforce, Zendesk, and SQL databases. This allows any employee to ask questions in natural language and receive answers with verifiable citations.

The selling point is clear: if your company has a lot of internal documents and a new employee asks about the LATAM refund policy, Perplexity Enterprise responds in seconds with exact excerpts, source links, and a coherent summary. All without needing to train a model or hire an ML team.

Starting price: $40 per user per month, with a minimum of 50 users. That's $24,000 a year for a small team. For perspective, a ChatGPT Enterprise license costs $60/user/month and requires custom integrations, while Anthropic's Claude for Work charges $30/user but has stricter heavy file restrictions.

What surprised me most is how Perplexity designed its product for situations where speed is more crucial than creativity. It doesn't try to compete with Copilot in code writing or Jasper in marketing. It's focused on problems like "I have 6 hours to prepare a proposal and need to extract information from 200 PDFs I haven't read yet."

Three Technical Advantages That Matter

1. Real-Time Federated Retrieval. Perplexity doesn't copy your data to its infrastructure; instead, it performs federated queries that simultaneously send your question to your systems (Google Drive, Notion, SQL) and aggregate results almost instantly. This reduces compliance risks and eliminates the need for synchronization.

2. Granular Citation. Each answer includes direct links to the exact paragraph in the source document. If the answer is wrong, you can verify it in two clicks. ChatGPT Enterprise and Claude cite entire documents, but not specific passages.

3. Frictionless Multimodality. Perplexity processes scanned PDFs, presentation images, Excel tables, and even videos with automatic transcriptions. No need to preprocess files or convert formats.

SoftBank’s Plan That No One Is Seeing

SoftBank doesn't just invest for passive returns; it invests to create ecosystems under its control. In 2025, SoftBank launched the Vision Fund III with $140 billion focused on applied AI. So far, it has funded 14 AI startups ranging from legal synthesis tools to financial analysis platforms.

These companies can now integrate with Perplexity Enterprise, becoming an intelligence layer within the SoftBank ecosystem. A concrete example is Brighterion, a fraud detection platform for banks. Brighterion generates 2 million alerts a month, but human analysts can only review 3% due to time constraints. With Perplexity Enterprise, an analyst can inquire about alerts similar to a recent transaction and get immediate context without searching through logs.

SoftBank is also negotiating with Arm Holdings (which it controls) to optimize Perplexity's models on Arm chips for data centers. If this materializes, Perplexity could run inferences 40% cheaper than competitors on Arm infrastructure, already adopted by AWS and Google.

Three Risks That Could Sink It All

Risk 1: OpenAI Launches Enterprise Search in 6 Months. In January 2026, OpenAI hired Srinivas Narayanan, former VP of Search at Google. His team of 120 engineers is developing an enterprise search product. If OpenAI launches by year's end, Perplexity could lose its competitive edge.

Risk 2: Users Don’t Trust the Answers. In internal tests with 8 beta companies, Perplexity had a 34% adoption rate after 3 months. This means 2 out of 3 employees who tried the tool stopped using it due to concerns like "I don't know if the answer is correct or made up." Although Perplexity cites sources, users prefer manual verification because "that way I know it's real."

Risk 3: Inference Costs Explode with Scale. Each Perplexity query runs between 15 and 40 LLM model calls, depending on complexity. With 50 employees making 20 queries a day, that's 30,000 calls a month. If Perplexity uses GPT-4 or Claude 3.5 Opus for inferences, the marginal cost per user could be $18 a month. With a $40/user price, the gross margin would be 55%. This is sustainable but leaves little room for subsidies or price competition.

What This Means for Founders and Tech Teams

If you're evaluating AI tools for your company in 2026, Perplexity Enterprise is only relevant if you meet two conditions: you have over 10,000 dispersed internal documents, and your team spends over 5 hours a week searching for information.

Cases Where Perplexity Wins:

• Support teams needing to respond to technical tickets with information from Confluence, Jira, and Zendesk.
• Legal teams reviewing contracts, precedents, and regulations in real-time.
• Financial analysts needing to cross internal reports with market data.

Cases Where Perplexity Loses:

• Development teams needing to generate code (better with Copilot or Cursor).
• Marketing teams needing to create content (better with ChatGPT or Claude).
• Companies with fewer than 50 employees where the ROI doesn't justify $24K/year.

Honestly, the big dilemma is that Perplexity sells efficiency, not a new capability. It's hard to justify $40/user/month for something that "saves time" versus something that "creates new value." CFOs prefer tools that generate revenue or reduce staff, not those that "make employees faster."

The Uncomfortable Conclusion No One Wants to Admit

Why did Perplexity receive $300 million? Not because it has the best model, the most advanced technology, or the most innovative product. It received it because SoftBank needed a horse in the enterprise AI race, and Perplexity was the most accessible and fastest option to position.

OpenAI is out of reach with its $80 billion valuation. Anthropic has allied with Google. Mistral, being European, presents complexities for scaling in Asia. Perplexity represents the relatively lowest-risk bet, with enough traction (15M users) to justify an aggressive entry.

The key question isn't whether Perplexity is truly worth $9 billion. The real question is whether SoftBank can drive enterprise adoption using its network before OpenAI or Google launch superior products. If successful, Perplexity could be worth $20 billion in 18 months. If not, this round will be seen as another failed Vision Fund bet.

Would your team pay $40/user/month to search internal documents faster? For more insights on enterprise AI tools, check out our article on Supabase vs. MongoDB: Which Saves You $20K Annually? and learn about the competitive landscape in AI with Migrating to Supabase Costs an Extra $18K: A True Story.