Perplexity plans IPO in 2028 regardless of what happens to Anthropic or OpenAI, CEO tells CNBC

Perplexity said it is still planning to go public in 2028, unchanged from prior guidance, even as Anthropic has confidentially filed for an IPO and OpenAI is also reportedly preparing one. CEO Aravind Srinivas said the success or failure of these mega-IPOs, including SpaceX, will be an important signal for investor appetite and could create ripple effects across the AI sector. He also argued that frontier AI valuations remain justified, but future spending will be more disciplined and based on cost-efficient model selection.

Analysis

The real market signal here is not the 2028 timeline itself, but the impending repricing of AI as a cash-flow discipline story rather than a pure capability story. Once the largest private labs are forced into public-market scrutiny, the winners will be the companies that can prove unit economics under varying inference loads, not just benchmark leadership. That shifts capital toward orchestration layers, model-agnostic applications, and infrastructure providers with pricing power, while compressing multiples for any vendor whose demand is driven by “best model at any cost” behavior.

This is also a competitive warning shot to frontier labs: the market is likely to distinguish between research velocity and monetization durability. If model improvements slow for even 1-2 quarters, public investors will likely haircut forward ARR assumptions faster than private rounds have in the past, because listed comps create an observable discount rate for the entire AI stack. In that regime, open-source and smaller models gain share not because they are superior in absolute terms, but because CFOs will finally be able to justify them on payback periods under 12 months.

The second-order effect is a procurement reset. Enterprises will increasingly route low-stakes workloads to cheaper models and reserve frontier models for edge cases, which reduces the “all-in” revenue capture opportunity for premium model providers. That favors vertical software companies that bundle model choice and governance, and it hurts pure-play inference monetizers whose growth depends on expanding token consumption rather than expanding workflow automation. The most important catalyst window is the next 3-9 months: if early mega-IPOs trade poorly or the market senses slowing model cadence, AI spend plans should tighten quickly.

Key Decisions for Investors

• Short basket of frontier-lab-adjacent high-multiple AI names on strength over the next 1-2 quarters; express via long-dated puts or put spreads in QQQ/ARKK proxies most exposed to AI enthusiasm decay. Risk/reward: 2:1 to 3:1 if IPO comps disappoint and multiple compression broadens.
• Go long MSFT against a basket of pure-play model winners over 6-12 months. MSFT benefits if enterprises rationalize AI spend and consolidate around integrated productivity/workflow layers; downside is lower if AI capex slows, upside is cleaner monetization visibility.
• Pair long SNOW or DDOG vs short a high-token-consumption AI beneficiary over 3-6 months. Expect governance, observability, and workflow-enablement software to absorb budget as CFOs force model-agnostic procurement. Target 15-25% relative outperformance if enterprise AI spend becomes ROI-gated.
• Accumulate NVDA only on post-IPO selloffs, not into strength. The stock still has structural demand support, but if model providers signal spending discipline, the next leg is likely more volatile and range-bound. Use call spreads instead of outright longs to cap premium decay.
• Watch for a downgrade window in late 2026-2027 on any company whose growth depends on token expansion rather than workflow automation; if the market begins pricing public AI names on gross margin and retention instead of TAM, those names should be shorted aggressively.