Nothing CEO Predicts AI Agents Replace Apps

Carl Pei's prediction is not just another tech executive speculating about the future — it represents a concrete business thesis backed by $200 million in fresh capital and a planned product roadmap. When the CEO of a $1.3 billion hardware company says apps will disappear, it signals that real R&D dollars are being allocated to build the alternative. The mobile app economy generates roughly $200 billion annually, supporting millions of developers and underpinning the platform power of Apple and Google. A genuine shift from apps to AI agents would redistribute that value in ways that benefit hardware makers, AI model providers, and service backends at the expense of app stores and the current discovery-and-distribution model.

The timing is significant. Pei's remarks came just two weeks after Qualcomm CEO Cristiano Amon independently declared 2026 'the year of agents' at a separate event. When both the chip supplier and the device maker are aligned on the same narrative — and have a financial relationship through Qualcomm Ventures' investment in Nothing — it suggests coordinated supply-chain readiness, not mere thought leadership. The convergence of investor capital (Tiger Global), silicon capability (Qualcomm), and a hardware company willing to bet its product roadmap on the thesis creates a credible, if still speculative, pathway to execution.

Pei's vision centers on a fundamental redesign of the smartphone interface. Today, accomplishing a single real-world goal — say, meeting a friend for coffee — requires opening a messaging app, then Maps, then Uber, then a calendar app. Each app is a discrete silo with its own UI, login, and data model. Pei argues this multi-app friction is an artifact of an interface designed for human fingers, not for intelligent software. His proposed alternative: a single AI agent that understands user intent ('I want to grab coffee with Sarah tomorrow') and orchestrates all necessary services behind the scenes — messaging, routing, ride-hailing, scheduling — without the user ever opening an app.

Critically, Pei distinguishes between two stages of AI evolution. Stage 1 is single-task automation — an AI that books a flight when asked — which he dismisses as 'super boring' and essentially a voice-activated version of existing apps. Stage 2, the transformative phase, involves proactive AI that learns user patterns and intentions over time and acts without explicit prompting. The technical design implication is equally radical: 'The future is not the agent using a human interface. You need to create an interface for the agent to use.' This means building machine-readable APIs and agent-native operating systems rather than wrapping AI around existing touch-based UIs. Nothing's planned AI-native devices for 2026 presumably target at least early prototypes of this architecture.

The financial stakes and scale indicators surrounding this prediction are substantial. Nothing has raised over $450 million in total funding, with the most recent $200M Series C in September 2025 explicitly tied to AI-first device development. The company's $1.3 billion valuation reflects investor confidence in the AI-agent thesis, not just its current hardware sales. Nothing crossed $1 billion in cumulative sales and grew 150% year-over-year in 2024, demonstrating commercial traction that gives Pei a platform from which to make bold predictions.

On the disruption side, the global mobile app economy generates approximately $200 billion in annual revenue across app stores, in-app purchases, and advertising. Apple's App Store alone generated an estimated $85 billion in 2024, with Apple taking a 15-30% commission on transactions. Google Play follows a similar model. If agents intermediate between users and services, bypassing app stores entirely, the commission structures that fund Apple and Google's platform businesses face an existential threat. Pei estimates a 7-10 year full transition timeline, with significant OS changes visible within 5 years — a window that aligns with typical smartphone platform cycles and gives incumbents time to respond but also gives challengers like Nothing time to build.

The most immediate impact is strategic: startup founders and app developers received a direct warning from Pei that app-centric business models face disruption. Companies whose 'core value lies' in their app interface — rather than in the underlying service or data — are most vulnerable. This echoes the web-to-mobile transition of the 2010s, when companies that were slow to build mobile apps lost ground to mobile-native competitors. The next transition may favor companies with strong APIs and service layers over those with polished but siloed app interfaces.

For Apple and Google, the threat is nuanced. Both companies are investing heavily in their own AI agents — Apple Intelligence and Google Gemini — which positions them to control the agent layer just as they control the app layer today. However, Pei's argument is that incumbent platforms have a conflict of interest: they profit from the app store model and are therefore incentivized to layer AI on top of apps rather than replace them. This creates an opening for a company like Nothing, which has no app store revenue to protect. The privacy implications are also significant: a proactive AI agent that anticipates user needs requires deep behavioral profiling and continuous data access, raising trust and regulatory questions that could slow adoption or advantage companies that can credibly promise on-device processing.

Pei's prediction fits into a broader industry shift that has been building throughout 2025 and into 2026. The emergence of capable AI agents — from OpenAI's operator-style tools to Anthropic's computer-use capabilities to Google's Gemini actions — has made multi-step autonomous task execution technically credible for the first time. What was a research demo in 2024 is becoming a product category in 2026. The question is no longer whether AI agents can orchestrate tasks across services, but who controls the agent layer and how value flows through it.

The social media response to Pei's prediction was notably strong. A 9to5Google tweet about the prediction garnered over 2,200 likes, suggesting the idea resonates with a tech-savvy audience even if it remains speculative. An EO YouTube video featuring Pei accumulated 773,000 views with a dedicated AI chapter, indicating sustained audience interest in Nothing's AI vision beyond the news cycle. Reddit discussion was absent, likely due to the story's freshness (it broke March 18-19, 2026), but the X and YouTube engagement suggests the prediction will continue generating discourse. The fundamental tension — between incumbents who profit from the app paradigm and challengers who see an opening in its dissolution — will define the next decade of mobile computing, regardless of whether Pei's specific timeline proves accurate.