a16z "Big Ideas for 2026: Part Two"

Software has eaten the world. Now, it will drive the world forward.

Written by: a16z New Media

Translated by: Block unicorn

Yesterday, we shared the first part of the "Big Ideas" series, which included the challenges our infrastructure, growth, bio + health, and Speedrun team partners believe startups will face in 2026.

Today, we continue with the second part of the series, featuring contributions from the American Dynamism team (an investment team a16z established in 2021) and the Apps team.

American Dynamism

David Ulevitch: Building an AI-Native Industrial Base

The United States is rebuilding the economic components that truly empower the nation. Energy, manufacturing, logistics, and infrastructure are once again in the spotlight, and the most important shift is the rise of a truly AI-native, software-first industrial base. These companies start with simulation, automated design, and AI-driven operations. They are not modernizing the past—they are building the future.

This is creating enormous opportunities in fields such as advanced energy systems, heavy robotics manufacturing, next-generation mining, and bio- and enzyme-driven processes (producing precursor chemicals relied upon by various industries). AI can design cleaner reactors, optimize extraction, design better enzymes, and coordinate autonomous machine clusters with insights that traditional operators could never achieve.

The same transformation is reshaping the world outside the factory. Autonomous sensors, drones, and modern AI models can now continuously monitor ports, railways, power lines, pipelines, military bases, data centers, and other critical systems that were once massive and difficult to manage comprehensively.

The real world needs new software. The founders who build this software will shape the prosperity of America for the next century.

Erin Price-Wright: The Revival of American Factories

America’s first great century was built on strong industrial power, but as is well known, we have lost much of that strength—partly due to offshoring, and partly due to a deliberate societal lack of constructive building. However, the rusting machines are coming back to life, and we are witnessing a revival of American factories centered on software and AI.

I believe that by 2026, we will see enterprises tackling challenges in energy, mining, construction, and manufacturing with a factory mindset. This means combining AI and automation with skilled technical workers to run complex, customized processes as efficiently as an assembly line. Specifically, this includes:

• Quickly and repeatedly navigating complex regulatory and permitting processes
• Accelerating design cycles and designing for manufacturability from the outset
• Better management of large-scale project coordination
• Deploying autonomous systems to accelerate tasks that are difficult or dangerous for humans

By applying the techniques Henry Ford developed a century ago—planning for scale and repeatability from the start—and integrating the latest advances in AI, we will soon achieve mass production of nuclear reactors, build housing to meet national demand, construct data centers at astonishing speed, and enter a new golden age of industrial strength. As Elon Musk said, "The factory is the product."

Zabie Elmgren: The Next Wave of Observability Will Be Physical, Not Digital

Over the past decade, software observability has changed how we monitor digital systems, making codebases and servers transparent through logs, metrics, and traces. The same transformation is about to sweep the physical world.

As over a billion connected cameras and sensors are deployed across major US cities, physical observability—real-time awareness of the operation of cities, power grids, and other infrastructure—is becoming both urgent and feasible. This new layer of perception will also drive the next frontier of robotics and autonomous technology, as machines rely on a universal framework that makes the physical world as observable as code.

Of course, this shift also carries real risks: tools that can detect wildfires or prevent construction site accidents could also trigger dystopian nightmares. The winners of the next wave will be those who earn public trust, build privacy-protecting, interoperable, and AI-native systems, and enhance societal transparency without compromising civil liberties. Whoever builds this trusted framework will define the direction of observability for the next decade.

Ryan McEntush: The Electronic Industrial Stack Will Change the World

The next industrial revolution will not only happen in factories, but also inside the machines that power those factories.

Software has already revolutionized how we think, design, and communicate. Now, it is changing how we move, build, and produce. Advances in electrification, materials, and AI are converging, enabling software to truly control the physical world. Machines are beginning to sense, learn, and act autonomously.

This is the rise of the electronic industrial stack—a comprehensive technology powering electric vehicles, drones, data centers, and modern manufacturing. It connects the atoms that drive the world with the bits that control it: from minerals refined into components, energy stored in batteries, electricity controlled by electronic devices, to motion achieved through precision motors—all coordinated by software. It is the invisible foundation behind every breakthrough in physical automation; it determines whether software merely hails a taxi or truly takes the wheel.

However, from refining critical materials to manufacturing advanced chips, the ability to build this stack is slipping away. If the US wants to lead the next industrial era, it must manufacture the hardware that underpins it. The countries that master the electronic industrial stack will define the future of industrial and military technology.

Software has eaten the world. Now, it will drive the world forward.

Oliver Hsu: Autonomous Labs Accelerate Scientific Discovery

With advances in multimodal model capabilities and continuous improvements in robotic manipulation, teams will accelerate autonomous scientific discovery. These parallel technologies will give rise to autonomous labs capable of closing the loop on scientific discovery—from hypothesis generation to experiment design and execution, to reasoning, results analysis, and iteration on future research directions. The teams building these labs will be interdisciplinary, integrating expertise in AI, robotics, physical and life sciences, manufacturing, and operations, enabling continuous cross-domain experimentation and discovery in unattended labs.

Will Bitsky: The Data Journey of Critical Industries

In 2025, the zeitgeist of AI will be defined by the limits of compute resources and data center construction. By 2026, it will be defined by the limits of data resources and the next frontier of the data journey—our critical industries.

Our critical industries remain treasure troves of latent, unstructured data. Every truck dispatch, every meter reading, every maintenance job, every production run, every assembly, every test drive is material for model training. However, terms like data collection, labeling, and model training are not commonly used in industry.

The demand for this type of data is endless. Companies like Scale, Mercor, and AI research labs are tirelessly collecting process data (not just "what was done," but "how it was done"). They pay high prices for every bit of "sweatshop data."

Industrial enterprises with existing physical infrastructure and workforces have a comparative advantage in data collection and will begin to leverage it. Their operations generate massive amounts of data, which can be captured at almost zero marginal cost and used to train proprietary models or licensed to third parties.

We should also expect startups to emerge and offer help. Startups will provide the coordination stack: software tools for collection, labeling, and licensing; sensor hardware and software development kits (SDKs); reinforcement learning (RL) environments and training pipelines; and eventually, their own intelligent machines.

Apps Team

David Haber: AI-Enhanced Business Models

The best AI startups are not just automating tasks; they are amplifying customers’ economic value. For example, in contingency-based law, firms only get paid if they win. Companies like Eve use proprietary outcome data to predict case success rates, helping firms select better cases, serve more clients, and improve win rates.

AI itself can enhance business models. It not only reduces costs but also generates more revenue. By 2026, we will see this logic extend across industries, as AI systems align more deeply with customer incentives and create compound advantages that traditional software cannot match.

Anish Acharya: ChatGPT Will Become the AI App Store

Consumer product cycles require three elements for success: new technology, new consumer behaviors, and new distribution channels.

Until recently, the AI wave met the first two conditions but lacked a new native distribution channel. Most products grew by relying on existing networks like X or word of mouth.

However, with the release of the OpenAI Apps SDK, Apple’s support for mini-programs, and ChatGPT’s group chat feature, consumer developers can now directly tap into ChatGPT’s 900 million user base and leverage new mini-program networks like Wabi for growth. As the final link in the consumer product lifecycle, this new distribution channel is poised to spark a once-in-a-decade consumer tech gold rush by 2026. Ignore it at your own risk.

Olivia Moore: Voice Agents Begin to Take Hold

In the past 18 months, the vision of AI agents handling real interactions for businesses has gone from science fiction to reality. Thousands of companies, from small businesses to large enterprises, are using voice AI to schedule appointments, complete bookings, conduct surveys, collect customer information, and more. These agents not only save costs and generate additional revenue for businesses but also free up employees to focus on more valuable—and more interesting—work.

But since this field is still in its early stages, many companies are still at the "voice as an entry point" stage, offering only one or a few types of calls as a single solution. I’m excited to see voice assistants expand to handle entire workflows (possibly multimodal) and even manage complete customer relationship cycles.

This will likely mean agents will be more deeply integrated into business systems and empowered to handle more complex types of interactions. As underlying models continue to improve—agents can now call tools and operate across different systems—every company should deploy voice-first AI products and use them to optimize key business processes.

Marc Andrusko: Proactive, Promptless Apps Are Coming

By 2026, mainstream users will say goodbye to prompt boxes. The next generation of AI apps will display no prompts at all—they will observe your actions and proactively offer suggestions for you to consider. Your integrated development environment (IDE) will suggest refactoring before you even ask. Your customer relationship management (CRM) system will automatically generate follow-up emails after you finish a call. Your design tools will generate options as you work. The chat interface is just an auxiliary tool. Now, AI will become the invisible scaffolding running through every workflow, activated by user intent rather than commands.

Angela Strange: AI Will Ultimately Upgrade Banking and Insurance Infrastructure

Many banks and insurance companies have integrated AI features such as document ingestion and AI voice agents into their legacy systems, but only by rebuilding the infrastructure that supports AI can AI truly transform financial services.

By 2026, the risk of failing to modernize and fully leverage AI will outweigh the risk of failure, and we will see large financial institutions abandon contracts with legacy vendors in favor of newer, more AI-native alternatives. These companies will break free from the constraints of past classifications and become platforms that can centralize, standardize, and enrich underlying data from legacy systems and external sources.

What will be the result?

• Workflows will be significantly simplified and parallelized. No more switching between different systems and screens. Imagine: you can view and process hundreds of pending tasks in parallel in the loan origination system (LOS), and agents can even handle the more tedious parts.
• The classifications we know will merge into larger categories. For example, customer KYC, account opening, and transaction monitoring data can now all be unified on a single risk platform.
• The winners in these new categories will be 10 times the size of legacy companies: the scope is larger, and the software market is eating the workforce.

The future of financial services is not about applying AI on top of old systems, but about building an entirely new AI-based operating system.

Joe Schmidt: Forward-Looking Strategies Bring AI to 99% of Enterprises

AI is the most exciting technological breakthrough of our lifetime. Yet, so far, most of the benefits for new startups have gone to the 1% of companies in Silicon Valley—either those physically located in the Bay Area or part of its vast network. This is understandable: founders want to sell products to companies they know and can easily reach, whether by visiting their offices or through boardroom connections with VCs.

By 2026, this will change completely. Enterprises will realize that the vast majority of AI opportunities exist outside Silicon Valley, and we will see new startups using forward-looking strategies to uncover more opportunities hidden in large, traditional vertical industries. In traditional consulting and services (such as system integrators and implementation firms) and slower-moving industries like manufacturing, AI holds enormous potential.

Seema Amble: AI Creates New Coordination Layers and Roles in Fortune 500 Companies

By 2026, enterprises will move further from isolated AI tools to multi-agent systems that need to operate like coordinated digital teams. As agents begin to manage complex, interdependent workflows (such as joint planning, analysis, and execution), companies will need to rethink how work is structured and how context flows between systems. We are already seeing companies like AskLio and HappyRobot undergoing this shift, deploying agents across entire processes rather than single tasks.

Fortune 500 companies will feel this shift most profoundly: they hold the largest reserves of siloed data, institutional knowledge, and operational complexity, much of which resides in employees’ minds. Turning this information into a shared foundation for autonomous workers will unlock faster decision-making, shorter cycles, and end-to-end processes that no longer rely on constant human micromanagement.

This shift will also force leaders to rethink roles and software. New functions will emerge, such as AI workflow designers, agent supervisors, and governance leads responsible for coordinating and auditing collaborative digital workers. In addition to existing systems of record, companies will need coordination systems: new layers to manage multi-agent interactions, judge context, and ensure the reliability of autonomous workflows. Humans will focus on edge cases and the most complex situations. The rise of multi-agent systems is not just another step in automation; it represents a restructuring of how enterprises operate, make decisions, and ultimately create value.

Bryan Kim: Consumer AI Shifts from "Help Me" to "Know Me"

2026 marks the point where mainstream consumer AI products will shift from boosting productivity to enhancing human connection. AI will no longer just help you get things done—it will help you understand yourself more clearly and build stronger relationships.

To be clear: this is not easy. Many social AI products have launched and ultimately failed. However, thanks to multimodal context windows and falling inference costs, AI products can now learn from every aspect of your life—not just what you tell a chatbot. Imagine your photo album surfacing genuine emotional moments, one-on-one and group chat modes shifting based on who you’re talking to, and your daily habits changing under stress.

Once these products truly arrive, they will become part of our daily lives. Generally, "know me" products have better user retention mechanisms than "help me" products. "Help me" products monetize through high willingness to pay for specific tasks and strive to improve retention. "Know me" products monetize through ongoing daily engagement: users are less willing to pay, but retention is higher.

People have always traded data for value: the question is whether the return is worth it. The answer will soon be revealed.

Kimberly Tan: New Model Primitives Give Rise to Unprecedented Companies

By 2026, we will witness the rise of companies that simply could not have existed before breakthroughs in reasoning, multimodality, and computer applications. So far, many industries (such as legal or customer service) have used improved reasoning to enhance existing products. But we are only now beginning to see companies whose core product features fundamentally depend on these new model primitives.

Advances in reasoning can enable new capabilities, such as evaluating complex financial claims or acting on dense academic or analyst research (e.g., adjudicating billing disputes). Multimodal models make it possible to extract latent video data from the physical world (e.g., cameras on manufacturing floors). Computer applications enable automation in large industries whose value was previously constrained by desktop software, poor APIs, and fragmented workflows.

James da Costa: AI Startups Scale by Selling to Other AI Startups

We are in the midst of an unprecedented wave of company creation, driven mainly by the current AI product cycle. But unlike previous cycles, incumbents are not standing still; they are actively adopting AI. So how can startups win?

One of the most effective and underrated ways for startups to outcompete incumbents in distribution is to serve new companies from day one: that is, to serve greenfield companies (entirely new enterprises). If you can attract all the newly founded companies and grow with them, you will become a large company as your customers scale. Stripe, Deel, Mercury, Ramp, and others have followed this strategy. In fact, many of Stripe’s customers didn’t even exist when Stripe was founded.

In 2026, we will see startups founded from scratch scale up across many enterprise software domains. They just need to build better products and go all-in on developing new customers who aren’t already locked in by incumbents.