Jun 15, 2026, 10:04 AM ETInterview

Perplexity CEO: Micron Will Be More Valuable Than Meta & How Export Controls Helped Not Hurt China

Perplexity has built a $20 billion valuation and 45 million users with over 1 billion monthly searches in just three years.
The company operates with a team of 400 people, a scale Aravind Srinivas views as a benchmark for future "efficient" tech companies.
Perplexity's revenue has more than tripled since the beginning of the year, driven by advanced "Deep Research" and "Computer" products.
The company has reduced its burn rate by over 50% while competing with OpenAI, which has driven down costs for the industry.
Srinivas predicts Perplexity could reach a $200 billion valuation with roughly 10,000 employees, or even $2 trillion with fewer than 100,000 total employees.
Perplexity is positioning its "Computer" product as an orchestrator rather than just a model builder, competing across different models (e.g., GPT-5.5 and Claude Opus) to maximize value.

Srinivas claims Perplexity has changed Google.com more than any product manager in Google's history by forcing a redesign of their homepage.
Google's new "AI Mode" interface closely mimics Perplexity's design, including font, citations, inline hyperlinks, and suggested follow-ups.
Srinivas asserts that Google bid $34 billion to buy Chrome, a valuation exceeding Perplexity's current market cap at the time of the statement.
The company argues that while Google dominates search, it lacks a dominant presence in the "frontier" category of agent-based work (coding, deep research).
Srinivas believes OpenAI is not financially ready for an IPO despite its consumer dominance, citing a lack of profitability in non-advertising revenue streams.
Srinivas disputes the viability of a $100–$200 billion advertising market for chat interfaces like OpenAI, arguing that discovery and subjective browsing (Meta/Instagram) drive ad revenue, not objective Q&A.

Srinivas posits that the "frontier" in AI is shifting from answering questions to "doing work" via autonomous agents.
He argues that the true value in AI lies in "orchestration" (harnesses, tools, connectors) rather than the model itself, stating "the model is no longer the product."
Perplexity differentiates itself by orchestrating across multiple models and tools, whereas competitors like Anthropic and OpenAI lock their models into single-product harnesses.
The core metric for AI success, according to Srinivas, is "token value per watt per user," prioritizing efficiency and output over raw model parameters.
Srinivas predicts that power users (e.g., single engineers or finance teams) will drive revenue growth exceeding the ad revenue of Google or Meta, with some users spending $10,000+ monthly on agent loops.
He argues that open-source models will commoditize the base layer, but companies will still pay a premium for "frontier" intelligence via specialized harnesses and local compute.

Srinivas identifies a shortage of power as the single biggest bottleneck for AI data center build-outs, surpassing hardware availability.
He estimates that 40% of proposed data center projects in the US are stalled due to public resistance, which he attributes to fear of job loss and wealth inequality rather than actual water/power consumption.
The company predicts that High Bandwidth Memory (HBM) suppliers like Micron could become more valuable than Meta in the next 6–12 months due to memory bottlenecks in agent loops.
Srinivas believes the US has a competitive advantage due to export controls, which have forced Chinese competitors like DeepSeek to develop more efficient, vertically integrated architectures that bypass the NVIDIA stack.
He warns that reliance on "server-side" frontier models is unsustainable for 24/7 agents; the future requires a hybrid approach using local models for privacy and cost efficiency.
Data center companies (e.g., CoreWeave, Nebius) must offer software orchestration on top of hardware to be sustainable, rather than simply renting GPUs.

Srinivas predicts that in two years, agent traffic on platforms like Cloudflare will surpass human traffic.
He forecasts that advertising models will survive only in "subjective" sectors (fashion, travel), while "objective" transactions will be fully automated by agents.
The CEO predicts a shift toward companies with fewer employees but higher output, noting that his own company (Perplexity) achieved multi-billion dollar status with just 400 people.
Srinivas believes the US must invest heavily in physical infrastructure (fabs, data centers, energy) to maintain competitiveness against China, which is building its own AI stack.
He anticipates an IPO for Perplexity could happen as early as 2028, though he notes they are currently growing revenue faster than profitability is the primary focus.
Srinivas expresses a strong bullish view on the long-term value of SpaceX, labeling it the only "end-game" company building space infrastructure for connectivity, unlike AI labs that are substitutable.

Srinivas rejects motivation derived from wealth, stating he is driven by "impact" and the fear of missing out on solving difficult problems.
He maintains an "offense-only" strategy, advising against defense or fear of failure, citing his background of coming from a lower-middle-class family in India where "nothing to lose" is a strategic advantage.
Srinivas critiques the "doom and gloom" narrative around AI job displacement, arguing that the real story is the ability to build billion-dollar companies with small teams.
He highlights the "10x engineer" concept, suggesting that agents will allow single individuals to perform the work of hundreds, but only if they possess the right "harness" and agency.
Srinivas identifies "asking better questions" as the defining skill of the AI era, urging entrepreneurs to assume they have unlimited agency and plan accordingly.
He cites Elon Musk's ability to "zone out" all distractions to focus on the single limiting problem as the most valuable entrepreneurial trait he has observed.
Srinivas describes Jensen Huang's mindset as operating with the urgency of a company 30 days from bankruptcy, despite having a $5 trillion market cap.