Poke Brings AI Agents to Everyday Users via Text Message

Poke raised an additional $10M on top of a $15M seed round, putting the company at a $300M post-money valuation and launched publicly in March 2026. The Palo Alto startup offers a text-based AI agent accessible via iMessage/SMS/Telegram without downloads, is backed by Spark Capital and General Catalyst, and plans to scale via user-created automations and creator partnerships to drive mainstream adoption.

Analysis

A lightweight SMS-first agent meaningfully lowers distribution friction versus app-based assistants and therefore converts a different user cohort: low-tech, high-frequency utility users (medication reminders, simple automations) who will not tolerate app installs or complex onboarding. If even 10–20% of the US smartphone base adopts such an interface within 12–24 months, it creates a steady, low-ARPU but huge-volume stream of inference requests and third-party integrations that monetizes differently (creator recipes, paid automations, API access) than advertising or premium LLM subscriptions.

Because Poke orchestrates across multiple model providers and lives outside any single app ecosystem, it accelerates a two-track market: winners selling cheap, high-throughput inference (commodity GPUs, optimized stacks) and middleware aggregators that sit between carriers, LLMs, and creators. That favors infrastructure suppliers over walled-platform ad sellers in the near-to-medium term and increases downside risk for businesses that rely on lock-in to extract high CPMs from assistant-like features.

Key tail risks are non-linear: platform-level throttling (Apple/Google), carrier filtering or pricing changes for SMS-heavy flows, and privacy/regulatory constraints that force on-device compute or blunt personalization. Catalysts to watch in the next 3–18 months are explosive creator/recipe growth (user engagement multiplier), any telco/aggregator commercial deals, and potential acquihires by large incumbents — which could either validate the model and compress exit multiples or eliminate a future competitor quickly.

Key Decisions for Investors

• Long NVDA (6–12 month call position): buy delta-0.30–0.40 calls expiring 9–12 months out to express rising inference demand from high-volume, low-ARPU agents. Position size 2–4% notional; stop-loss 25% of option premium. R/R: asymmetric — modest premium loss vs multi-33–50% equity upside if cloud/edge spend accelerates.
• Short META (6–12 month put or equity hedge): initiate a protective short/put sized to 50–70% of NVDA long notional to express risk to feed-based monetization from cross-platform agents. Target 15–25% downside capture; time horizon 6–12 months; cut if Meta announces a compelling native agent strategy or acquires middleware.
• Pair trade — long NVDA / short META (6–18 months): 1.0 NVDA exposure vs 0.6 META exposure to play infrastructure upside vs platform monetization pressure. This isolates model-inference demand from advertising cyclicality; rebalance on material M&A or platform policy moves.
• Watchlist & optionality: allocate a small carry bucket (0.5–1% NAV) to private or small-cap aggregator/telecom middleware names and short-dated event-driven puts around any announced platform restrictions. These are high-conviction informational triggers to scale or reduce the public trades above.