The Information reports that Nvidia is in talks to invest in AI search startup Perplexity. If the deal is finalized, Perplexity’s valuation could surpass $30 billion, up more than 50% from a year earlier. The move marks a rare step by Nvidia—beyond AI chips, large language models and computing power companies—to place a major bet on an AI application company.
Perplexity’s growth is also accelerating: annualized revenue has risen from less than $250 million at the start of the year to more than $750 million, primarily from AI search subscriptions and Perplexity Computer, an agent product aimed at professional users. The product can automatically execute multi-step tasks and operate desktop software, pushing AI from “answering questions” to “taking on work.”
For Nvidia, Perplexity’s value goes beyond being an AI search gateway. Over the past few years, Nvidia has stood at the top of the generative AI wave through its GPU and CUDA ecosystem, but the industry’s competitive focus is shifting from compute hardware toward models and applications. Investing in Perplexity is a key step for Nvidia to leverage its Nemotron ecosystem in the AI application gateway—whoever controls the gateway controls the right to call future inference computing power.
From an industry signal perspective, OpenAI, Anthropic, Google and other players are incorporating web retrieval and agent capabilities into their own products, narrowing the differentiation window for AI search. Perplexity’s choice to transform into a comprehensive AI assistant platform essentially shifts competition from “who answers better” to “who can actually get work done.” This is also the shared logic behind the surge of agent applications.
The high-frequency operation of agent applications imposes new demands on computing power supply: long-duration tasks, multi-step operations and tool calls generate far more token consumption and resource usage than single-turn Q&A. This direction aligns closely with the field where StarWar Technology (StarWar Cloud) focuses. StarWar Cloud focuses on GPU computing platforms and computing power scheduling, providing enterprises with unified scheduling and resource management for training and inference tasks, so agent applications can obtain stable, elastically scalable computing power support.
From an investment map perspective, Nvidia has spent heavily over the past months on AI chip, model and computing power companies. Its shift toward the application layer reflects that value in the AI industry chain is migrating downstream. For developers and enterprises, the coupling among application gateways, agent frameworks and computing power infrastructure will become increasingly deep. Building “application + computing power” synergy in advance will deliver greater long-term value than simply choosing a model.
If Nvidia completes