Sequoia Capital: AI is leading a $10 trillion revolution, grander than the Industrial Revolution

Sequoia Capital defines the current wave of artificial intelligence as a profound "cognitive revolution," believing that its transformative power will rival or even surpass that of the Industrial Revolution, and harbors a massive business opportunity valued at $10 trillion.

Recently, Sequoia Capital partner Konstantine Buhler delivered a speech titled "The AI Revolution: A $10 Trillion Wave, Greater than the Industrial Revolution."

He explicitly pointed out that the next battleground for AI is the vast service industry market. Similar to how SaaS software not only eats into traditional software market share but also greatly expands the entire market boundaries, AI will also undergo disruptive reshaping and expansion in the service industry. Likewise, AI will disrupt and reshape traditional service industries, such as law, accounting, and healthcare, which have been dominated by large partnerships, giving rise to new AI-driven public companies and even changing the S&P 500 index list as we know it today.

In the dawn of the "cognitive revolution," Sequoia Capital is actively seeking and investing in startups that can "specialize" general AI technology. These companies, akin to Rockefeller and Carnegie of the past, bear the mission of building future market leaders, refining general AI models into "cognitive assembly lines" that address specific industry pain points.

To guide investment directions, Buhler shared five major investment trends and five investment themes that Sequoia Capital is closely monitoring for the next 12 to 18 months. These trends and themes serve as a treasure map for gold diggers in the AI era.

Key Points of the Speech:

• Core Argument: Sequoia Capital defines artificial intelligence (AI) as a "cognitive revolution," whose scale and impact will rival or even surpass the Industrial Revolution.
• Market Opportunity: The core business opportunity of AI lies in the $10 trillion U.S. service industry market, where AI will not only capture market share but also, like SaaS reshaping the software market, greatly expand the service industry market itself.
• Historical Analogy: The development of AI is likened to the "specialization" process of the Industrial Revolution, transitioning from general technologies (like steam engines/GPUs) to highly specialized applications (like factory assembly lines/dedicated AI applications), with startups being the driving force behind this process.
• Five Investment Trends: Sequoia Capital has observed five ongoing trends: 1) Work models shifting to "high leverage, high uncertainty"; 2) Measurement standards shifting from academic benchmarks to "real-world validation"; 3) Reinforcement learning moving from theory to practice; 4) AI penetrating the physical world, beyond the realm of robotics; 5) Computing power becoming a new productivity, with per capita computing power consumption expected to grow 10 to 1,000 times.
• Five Investment Themes: In the next 12-18 months, Sequoia will focus on five themes: 1) Persistent memory; 2) Seamless communication protocols between AIs; 3) AI voice; 4) AI security throughout the entire chain; 5) Open-source AI.
• Ultimate Goal: The aim is to accelerate the development of the above themes, significantly compressing the construction time of the "cognitive assembly line" from years to months, thereby hastening the arrival of the entire cognitive revolution

Cognitive Revolution: Learning from History, Specialization is Key

Sequoia Capital believes that to understand the future of the AI revolution, one must learn from history. Buhler compares the development of AI to three key milestones of the Industrial Revolution: the first GPU (graphics processing unit) in 1999 was like the "steam engine" of that time; the AI systems that emerged in 2016 were akin to the first "factory"; and future AI applications are equivalent to the "assembly line" of factories.

It is noteworthy that it took 144 years for the Industrial Revolution to evolve from the first factory to a mature assembly line. Buhler emphasizes that the core driving force behind this long cycle is "specialization"—transforming general technologies and labor into highly specialized components and labor capable of achieving specific outputs.

Today, AI is at a similar historical juncture. General large models provide foundational capabilities, but the true value release relies on startups "specializing" them to solve problems in specific industries and scenarios. In Sequoia's view, today's AI startups are the key forces playing the roles of Rockefeller and Carnegie in this cognitive revolution, building the great applications of the future through specialization.

Trillion-Dollar Opportunity: AI Reshaping the Service Industry Landscape

Commercial value is at the core of investment. Buhler points out that the opportunities in AI far exceed those in the software market. He focuses on the $10 trillion U.S. service industry market, where the current penetration of AI may be only around $20 billion. This represents the "10 to the 13th power" dollar-level opportunity for AI.

An internal memo from Sequoia shows that the market sizes for positions such as registered nurses, software developers, and legal services are all extremely large. The firm has already made investments in these areas, backing startups like Open Evidence, Factory, and Harvey.

Buhler emphasizes that just as traditional law firms or accounting firms do not appear on the market capitalization rankings of the S&P 500 index, the cognitive revolution will create a historic opportunity for AI-driven service companies to grow into new public giants, thereby greatly expanding the landscape of capital markets.

Five Investment Trends: What is Happening Now

Buhler shared five key trends that Sequoia Capital has observed, revealing how AI is changing the real world:

1. Leverage over Certainty: The working model is shifting from "low leverage, high certainty" to "ultra-high leverage, uncertain outcomes." For example, salespeople can use AI agents to manage hundreds of clients, transforming their own roles into correcting and guiding the AI, thus achieving over 1000% leverage effect.

2. Real-World Validation: The gold standard for measuring AI capabilities has shifted from academic benchmarks to real-world performance. For instance, the AI security company Expo demonstrates its technological excellence by competing and winning against top human hackers on the Hacker One platform.

3. Reinforcement Learning Goes Practical: Reinforcement learning, which has long remained at the theoretical level, has become a practical tool for many startups to gain competitive advantages over the past year, especially in fields like programming.

4. AI Deepens into the Physical World: The influence of AI extends far beyond humanoid robots; it is changing the physical world by optimizing processes and accelerating hardware manufacturing. For example, Nominal uses AI to accelerate hardware manufacturing and quality assurance.

5. Computing Power as New Productivity: The computing power consumed by each knowledge worker (FLOPS) is expected to increase at least tenfold, and optimistically, it could reach up to 1000 times. This presents significant opportunities for computing power suppliers and companies that arm employees with AI.

Five Investment Themes: Layout for the Next 18 Months

Looking ahead to the next 12 to 18 months, Sequoia Capital will focus on investing in the following five themes to address the core bottlenecks in current AI development:

1. Persistent Memory: AI needs to have long-term memory (remembering long-term context) and identity persistence (maintaining a unique personality) to handle more complex productivity tasks. Currently, there is no "scale rule" for model training in this field, presenting a huge opportunity.

2. Seamless Communication Protocols: Just as the TCP/IP protocol ignited the internet revolution, if AIs can establish seamless communication protocols, it will give rise to disruptive applications, such as AIs autonomously completing the entire shopping process of price comparison, ordering, and payment across the web.

3. Explosion of AI Voice: AI voice has many exciting applications, including AI friends, AI companions, and AI therapists, all of which are consumer-facing applications. On the enterprise side, AI voice can be used for automating logistics coordination, financial transactions, and other scenarios.

4. End-to-End AI Security: AI security covers the entire chain from model development and distribution to end users. In the future, each user, or even each AI agent, may have hundreds of dedicated AI security guardians

5. The Crossroads of Open Source AI: Sequoia believes that ensuring the open-source community can compete with top proprietary models is crucial, as it relates to whether the AI ecosystem can remain open and free, avoiding future monopolization by a few well-funded tech giants.

The full speech is as follows (translated by AI tools):

Cognitive Revolution

00:00 At Sequoia Capital, we firmly believe that artificial intelligence is leading a revolution, and we think this transformation will be as grand as the Industrial Revolution, if not grander. In today’s brief presentation, we will share the entirety of this "Cognitive Revolution" and why it presents a massive opportunity valued at 10 to the power of 13 (10 trillion dollars).

00:22 We have four parts. First is our argument around artificial intelligence. Next is the business opportunities. Then we will delve into some investment trends. These are the dynamics we are currently observing in the field of artificial intelligence. Finally, we will discuss some investment themes, which are our outlook for the next 12 to 18 months.

00:42 We believe that artificial intelligence has parallels with the Industrial Revolution. Here, we want to highlight three key moments in the Industrial Revolution. The invention of the steam engine that started it all; the first factory system that brought all necessary components under one roof; and finally, the birth of the first assembly line as we know it today. The interesting aspect of this slide is the time span between these moments: it took 67 years from the first steam engine to the first factory. Moreover, that first factory didn’t even use a steam engine but was powered by water. Then, it took another 144 years from the first factory to the assembly line we know today. The question is, why did it take so long? Particularly, what happened during those 144 years? We believe the reason it took so long is the inevitability of "specialization." For a complex system, after surpassing a certain scale to achieve maturity, it must combine general components and labor with highly specialized components and labor. In other words, what happened at that time was the specialization of these general technologies to produce specific outputs.

01:58 Now we are living in a cognitive revolution. You could say that the "steam engine" of that time was the first GPU in 1999, the GeForce 256, and the system that emerged in 2016, capable of integrating all necessary components to produce AI capabilities, is the first "AI factory." So the question arises: who will become the John Rockefeller, Andrew Carnegie, Westinghouse, or Wedgwood of this cognitive revolution? We believe the answer lies in today's startups—these are the ones practicing this specialization, along with those that have yet to be established but will build these applications.

Trillion-Dollar Service Opportunity

02:35 We are Sequoia Capital, not the Sequoia Historical Research Institute. So, let's talk about real money. You may have seen this slide before. We used it at the AI Ascent conference. The $350 billion circle on the left represents spending on software during the early stages of cloud computing transformation. Those six small slices represent $6 billion in Software as a Service (SaaS) spending.

03:01 What happened later is that SaaS not only increased its share in the local software market, but it actually expanded the entire market, growing it to over $650 billion today. We believe similar things will happen in the field of artificial intelligence, and the opportunities will be even greater. This is a $10 trillion opportunity in the U.S. service market, of which only about $20 billion is currently automated by AI. This is a 10 to the power of 13 (trillion-dollar) opportunity that can not only expand AI's share in this market but also expand the market itself.

03:36 You may have seen the previous slide, but this one is being shown for the first time. It is taken from an internal memo at Sequoia, showing service industry jobs sorted by the rightmost column. That column is the number of employees in those jobs multiplied by the annual average wage published by the U.S. Census Bureau. You will notice that they are all very large markets. You will also notice that Sequoia has already invested in these areas. Consider Open Evidence and Freed in the registered nurse field, or Factory and Reflection in the software developer field, or Harvey, Crosby, and Finch in the legal field.

04:18 At Sequoia, we like to think about Total Addressable Market (TAM), especially in the market. Our founder, Don Valentine, always emphasized the importance of the market. This chart shows the market capitalization ranking of the S&P 500. You will notice several very large companies. On the far left is NVIDIA, with a market capitalization exceeding $4 trillion, and the percentage is its stock price performance over the year. Kirkland and Ellis law firm and Baker Tilly accounting firm are not on this slide, even though these companies generate billions of dollars in revenue each year.

04:54 We believe the cognitive revolution presents an opportunity to expand the market and to expand this slide to include many large, independent publicly traded companies established in the service sector driven by AI.

Investment Trend #1: Leverage Over Uncertainty

05:13 Next, let's talk about the five major investment trends we are focusing on in the artificial intelligence cognitive revolution. The first is to embrace leverage and manage uncertainty. We have noticed that work is shifting from a past model of low leverage and 100% certainty of results for tasks to a new model of over 100% leverage and lower uncertainty in the precise performance of results for tasks If you are a salesperson, your job is to manage a series of clients, namely potential clients. Today, you may need to manage these clients yourself, monitoring every opportunity. But in the future empowered by AI agents, you can use tools like ROCS, having hundreds of AI agents, one for each client, to track their progress, see the changes happening, and show you opportunities to re-engage and expand relationships with clients. Of course, this AI agent won't do everything exactly your way; it may miss certain things or make mistakes. That's where humans come in—to make corrections. In this case, we see leverage of over 100%, and it could even reach 1000%, albeit with more uncertainty. This is not exactly the same as the work you are doing.

Investment Trend #2: Real-World Validation

06:35 Secondly, we notice that the metrics have shifted to the real world. For most of AI's development history, we used academic benchmarks. Over a decade ago, when I was still an AI engineer, we used ImageNet as a benchmark for computer vision research. But today, if you want to prove excellence, you must demonstrate it in the real world. Take Expo as an example; they wanted to prove that their AI is the top AI hacker in the world. They didn't just rely on academic benchmarks to prove this; they entered the real world and competed with all registered hackers globally on the Hacker One platform to find vulnerabilities. They were able to demonstrate that they could compete and win on real-world data, becoming the world's number one hacker. We notice that this is the new gold standard—not just academic benchmarks, but real-world measurements.

07:30 Thirdly, reinforcement learning is moving from theory to practice. The AI industry has been talking about reinforcement learning for a long time. Over the past year, we have seen it truly come into focus. Not only have large inference labs benefited from it, but we have also seen many of our portfolio companies benefit from it. Consider Reflection, which uses reinforcement learning to train some of the best open-source models in the coding field.

Investment Trend #3: AI Deepening into the Physical World

07:55 The fourth trend we notice is that AI is deepening into the physical world. This wave is genuinely happening, and its implications go far beyond humanoid robots. It is also creating new processes and hardware through artificial intelligence. Take Nominal as an example; it uses AI to accelerate the hardware manufacturing process and also employs AI for quality assurance after product deployment.

Investment Trend #4: Computing Power as New Productivity

08:21 Finally, we notice that the new production function is computing power. That is, the number of floating-point operations (FLOPS) consumed by each knowledge worker. If you survey our portfolio companies, they will say they predict that the computing power consumed by each knowledge worker will increase at least tenfold. This means that the computing power consumption for each knowledge worker will increase by at least ten times. As mentioned earlier, knowledge workers may use one, dozens, hundreds, or even thousands of AI agents In a more optimistic scenario, we see that the computational power consumption of every knowledge worker in the future will increase by 1,000 times or even 10,000 times. This is very powerful for companies that reason, protect reasoning, and leverage this new production function to empower more employees.

Investment Theme #1: Persistent Memory

09:10 Next, let's talk about five investment themes. These are the directions we will focus on investing in over the next year. The first theme is persistent memory. “Persistent memory” can refer to at least two things. The first is “long-term memory,” which means AI can remember contextual information shared with it over a long period. The second is “identity persistence,” which means AI agents can consistently maintain their unique personality and style. These two points are crucial for AI to take on an increasing number of work functions. An AI entering the productivity space must have long-term memory to understand the context and functions of the entire organization.

09:54 We note that in the field of persistent memory, there has not yet been a “scaling law” similar to that in model training. Whether through vector databases and RAG or increasingly longer context windows, many attempts have been made, but this fundamental issue remains unresolved, which also indicates a huge opportunity ahead.

Investment Theme #2: Communication Protocols

10:09 The next theme is seamless communication protocols. The emergence of multimodal communication protocols (MCP) has understandably ignited market enthusiasm. However, let's revisit the internet revolution: the birth of TCP/IP was not the end, but a starting gun.

10:23 In this communication revolution, we have the opportunity to enable seamless communication between AIs. We believe this will give rise to many disruptive major applications. One scenario worth considering is shopping. Currently, if you want to purchase a product using AI, you might first do some research with AI and then execute the order through your preferred provider with a one-click checkout. But in the future, we see that AI will be able to do this because of seamless communication protocols. AI will be able to complete the entire process, find you the best price, execute the purchase, and finalize it, while reducing the distractions from merchants that make using them easier than other providers.

Investment Theme #3: AI Voice

11:09 The third trend is AI voice. You might be surprised that I haven't mentioned AI video, but this is intentional. I believe AI video may arrive within a year, but AI voice is already here. This is because not only has its fidelity and voice quality significantly improved to the level of everyday use, but the latency has also been greatly reduced, allowing for real-time conversations with AI voice.

11:35 AI voice has many exciting applications. These include AI friends, AI companions, AI therapists, all of which are consumer-facing applications. I personally also have high hopes for other applications of AI voice on the enterprise side If you are a business that needs to transport goods and you are dealing with logistics issues, you will find that many logistics coordination tasks are still done via voice communication. Now, you can work towards a future that automates many of these logistics coordination tasks. If you want to buy and sell bulk fixed income products, you are likely still communicating with over-the-counter trading desks via phone. All of this can be accelerated in the business sector by using AI voice.

Investment Theme #4: AI Security

12:22 Next is AI security. We believe there are absolutely huge opportunities in the field of AI security that span the entire chain from the development layer to the end consumer. At the development layer, we see opportunities to help large foundational model labs and AI labs develop technology safely. Then there is distribution, ensuring safe distribution and preventing bad actors from intervening in the process. Then there are the users themselves, ensuring that when they use products or write new applications, they do not inadvertently introduce vulnerabilities. A specific example is: a consumer may be instructed by their AI to download a software through the terminal. They may not be familiar with terminal operations. And the AI may not know that this software could introduce vulnerabilities into the consumer's environment.

13:15 We are entering a world where AI can protect both individuals and AI agents. In fact, we see a future that is completely different from the physical world: in this digital world, you can have hundreds of AI security agents for each individual, even for each agent. Unlike the physical world, you are not limited by physical space or even the same costs. You can deploy a large number of AI security agents for every person and every agent.

Investment Theme #5: Open Source AI

13:44 Finally, we find that open source AI is at a critical moment in the development of AI. If you had asked us two years ago, we would have said that it seemed open source models had the opportunity to compete with state-of-the-art foundational models and might even surpass them. Today, that position seems more unstable. We believe that it is very important for the open source community to compete and provide some of the most advanced foundational models. We think this is crucial for a freer and more open future, where anyone can build. We want to help build that future, making open source models available for everyone to create excellent products, and the future of AI should not be limited to well-funded giants. These are the five investment themes we are thinking about. Now the question becomes, what will happen if we can translate these investment themes into progress? We believe that this progress will be able to significantly compress the time required for that cognitive assembly line from many years to just a few years. Thank you very much for your attention, and we hope we can build this great cognitive revolution together