Editor's Note: This article summarizes a16z's key trends for 2026—from multimodal data and creative tools to AI, native universities, privacy chains, and vertical integration. From AI collaboration to the resurgence of American industry, a multi-layered, panoramic vision of the future is presented. Infrastructure: Multimodal Data: Additional structure from documents, images, and videos for enterprise scenarios. AI creative tools will become fully multimodal: users can provide any reference content (such as videos, images, and sounds) to allow models to generate new scenes, continue stories, edit existing clips, or reshoot from different angles, thus providing creators from meme makers to Hollywood directors with traditional director-level fine-grained control and spawning multiple successful products. **Agent-Native Infrastructure: Supporting Large-Scale "Agent-Speed" Workloads** Enterprise backend infrastructure will face a significant challenge from internal AI agents: shifting from low-concurrency workloads at human speeds to recursive, bursty, and massively parallel tasks at agent speeds. This necessitates building "agent-native" infrastructure, adopting thunder cluster mode as the default, significantly increasing concurrency limits, and optimizing coordination mechanisms; otherwise, legacy systems may misinterpret agent behavior as DDoS attacks. Multimodal Creative Tools: Provide any form of reference content to the model to generate new works or edit existing scenes. AI creative tools will truly move towards multimodal support, allowing users to provide any form of reference content (such as video, images, and sound) to generate new scenes, continue stories, edit existing clips, or reshoot angles, thus providing creators of all kinds, from meme makers to Hollywood directors, with near-traditional directorial precision. AI Native Data Stack: How AI Agents Handle Contextual Issues and the Development of Business Intelligence Tools While the modern data stack has matured and integrated over the past year through mergers and acquisitions (such as the merger of Fivetran and dbt) and unified platforms (such as Databricks), truly AI-native data architecture is still in its early stages. In the future, AI will profoundly transform data flow, vector databases, proxy-based contextual access, and automated workflows, enabling data infrastructure to... AI infrastructure is inextricably integrated. Interactive Video: Beginning to create immersive experiences that truly feel like you can "walk into" the world. Video will transform from a passive viewing medium into an interactive space that can be truly "stepped into": video models can finally understand time, maintain consistency, remember history, and respond to user behavior in real time, thereby continuously maintaining the rules of people, objects, and physics. This allows generated videos to become buildable "living environments" for robot practice, game evolution, design prototyping, and agent learning, enabling, for the first time, humans to inhabit AI-generated videos. Consumer Side: AI Native University: Professors Become Learning Architects, Planning Data and Adjusting Models. The first AI Native University will be established, an adaptive academic institution built from scratch and operating around intelligent systems. Curriculum, instruction, research, and operations will be continuously optimized through real-time data feedback. Professors will become learning architects, and assessments will focus on students' transparency and skillful use of AI, thereby cultivating top talent proficient in AI orchestration and governance for the new economy. The Emergence of Health MAUs (Active Users): Users are not patients, but they want to monitor their health. Medical technology will see a new core user group: "Healthy MAUs"—consumers who are not yet sick but want to regularly monitor and understand their health. Leveraging AI to reduce care costs, new prevention-oriented insurance products, and the rise of subscription models, a wave of AI-native startups and existing companies transforming will target this largest potential group, providing continuously interactive, data-driven prevention services, thus shifting from a treatment-oriented to a prevention-oriented approach. ChatGPT as an AI App Store: Developers can reach ChatGPT's 900 million users. With the OpenAI Apps SDK, Apple's support for mini-apps, and the launch of ChatGPT's group chat functionality, ChatGPT, with its 900 million users, will become the "app store" of the AI era, providing a brand-new native distribution channel for consumer AI products, thereby triggering a once-in-a-decade golden wave of consumer technology together with new technologies and new consumer behaviors. World Models in Storytelling: Generating Complete 3D Environments from Text Prompts AI world models (such as Marble and Genie 3) will dominate the narrative field, blurring the lines between players and creators by generating interactive 3D virtual worlds from text. This will foster co-creation of multiverses like "Generative Minecraft," new story formats, and the digital economy, while also becoming a training ground for AI agents, robots, and even... AGI's powerful simulation environment opens up new frontiers for creative media and the economy. Privacy creates network effects: Privacy differentiates chains and creates chain lock-in. Privacy will become the most important moat in the cryptocurrency field: it is not only a key feature for global financial on-chaining, but also creates a powerful chain lock-in effect and network effect through the difficulty of "bridging secrets". This makes it difficult for users to migrate once they join a privacy chain, thus forming a winner-takes-all pattern in an environment of homogeneous performance competition, where a few privacy chains dominate most of the crypto market. Prediction Market Expansion: Becoming Larger, Broader, and Smarter Prediction markets will become larger, broader, and smarter by: listing more complex event contracts and increasing transparency through cryptography; introducing decentralized governance and LLM oracles to resolve disputed cases; leveraging AI trading agents to uncover new signals and reveal deep predictive factors; and complementing, rather than replacing, the polling ecosystem improved by AI and cryptography. Understanding Your Agent: Agents Require Cryptographic Signatures for Transactions. The bottleneck in the agent economy will shift from intelligence to identity verification: Non-human identities have far surpassed those of humans, yet remain a "bankless ghost." There is an urgent need to build the "Know Your Agent" (KYA) primitive, binding agents to their principals, constraints, and responsibilities through cryptographic signatures. Otherwise, merchants will continue to block agent transactions, and the financial industry has only months left to establish this critical infrastructure.
Staking Media: Not only does it embody the concept of "investment", but it also provides verifiable evidence
“staked media will emerge: Media creators will use cryptographic tools such as token assets, programmable lock-up, prediction markets, and on-chain history to publicly and verifiably prove that they are "staking their money on their words", thereby replacing the pseudo-neutrality or empty declarations of traditional media with transparent interest binding and risk commitment, providing a new and strong signal for information credibility, and supplementing rather than replacing existing media forms. Enterprise and Fintech The real disruption of enterprise software lies in the fact that the "system of record" will begin to lose its dominant position: AI, through direct reading, writing, and reasoning about operational data, will transform ITSM, ... Passive databases such as CRM are transformed into autonomous workflow engines. With the advancement of inference models and agent-based processes, prediction, coordination, and end-to-end execution are achieved, making the dynamic agent layer the main interface, while traditional record systems are relegated to a commercialized persistence layer. Their strategic value gives way to players who control the intelligent execution environment that employees actually use. Vertical AI Evolves into Multi-User Collaboration: The Collaboration Layer Becomes a Moat. Vertical AI will evolve from information retrieval and single-user reasoning to a "multi-user mode": AI agents represent various stakeholders (such as buyers and sellers, tenants and suppliers, CFOs and lawyers) to collaborate and coordinate within domain-specific permissions, workflows, and compliance frameworks. This ensures contextual synchronization, automatic negotiation within parameters, and anomaly flagging, significantly improving task success rates. Through multi-user and multi-agent collaboration, value is amplified, ultimately forming a powerful network effect and a moat against switching costs.
Voice Intelligence Agent Extension: Handling Complete Workflows and Customer Relationship Lifecycles
AI Voice agents will evolve from single-call "entry point solutions" to core products that are deeply integrated into enterprise systems, handling complete multimodal workflows and even the entire customer relationship lifecycle. With advancements in underlying models and improved tool invocation capabilities, every company will have voice-first AI agents to optimize key business processes. AI Native Banking Infrastructure: Smoother, Parallel Workflows AI will truly transform the banking and insurance industry: Large financial institutions will abandon legacy systems and turn to AI native infrastructure platforms to centralize and standardize data, thereby significantly simplifying parallel workflows, merging traditional categories to form a larger market, and creating new winners 10 times larger than the old giants. This is because the future of financial services lies in rebuilding a completely new operating system based on AI, rather than simply adding AI to old systems. American Innovation Powerhouse: The U.S. will rebuild an AI-native, software-first industrial base: Starting with simulation, automated design, and AI-driven operations in the energy, manufacturing, logistics, and infrastructure sectors, building the future directly rather than modernizing the past. This will unlock significant opportunities in advanced energy systems, robotic heavy manufacturing, next-generation mining, bio-enzyme processes, and autonomous sensor/drone real-time monitoring systems, shaping the next century of American prosperity. Factory Thinking: Modular Deployment of AI, Combined with Skilled Workers, Enables Complex Processes to Operate Like an Assembly Line. American factories are poised for a resurgence, centered on software and AI, employing a "factory thinking" approach to address challenges in energy, mining, construction, and manufacturing. By modularly deploying AI and autonomous systems alongside skilled workers, complex, customized processes are transformed into assembly line operations. From day one planning for scale and repeatability, this enables rapid navigation and oversight, accelerated design cycles, optimized coordination of large projects, and autonomous handling of hazardous tasks. Ultimately, this will lead to the large-scale production of nuclear reactors, housing, and data centers, ushering in a new golden age of industry. "The factory is the product." Physical Observability: Real-time Understanding of the Operation of Systems like Cities and Power Grids. The observation revolution will extend from the digital world to the physical world: leveraging billions of deployed, connected cameras and sensors to achieve real-time understanding of infrastructure such as cities and power grids, providing a shared physical perception texture for robots and autonomous systems. The winners will be those who build privacy-preserving, interoperable, and AI-native trusted systems, improving social readability without sacrificing freedom. Autonomous Science Labs: A Closed-Loop Operation from Hypothesis to Experimental Design With advancements in multimodal models and robotic manipulation capabilities, interdisciplinary teams will build autonomous laboratories to achieve a closed loop in scientific discovery: from hypothesis generation, experimental design and execution, to result reasoning and iterative research. By integrating expertise in AI, robotics, physical/life sciences, manufacturing and operations, and utilizing "lights-out labs," continuous experiments can be conducted across multiple fields, significantly accelerating scientific discovery.