Source: a16z
Compiled by: Felix, PANews
At the a16z Fintech Connect conference, Alex Rampell discussed with Ben Horowitz, co-founder and general partner of a16z, how AI is rewriting the fundamental rules of software competition, why crypto infrastructure is crucial in an AI-dominated world, and the development trends of venture capital.
PANews has compiled the highlights of the conversation.
Host: You have been deeply involved in this industry for many years. I’d like to start with the book you previously wrote about the difficult situations CEOs face, back when you experienced market crashes and company transformations at LoudCloud and Opsware. Nowadays, most startups are "AI-first," but for CEOs of traditional companies that are 5 to 10 years old and from the pre-AI era, facing a situation where even the financial markets are bearish on them, what should they do?
Ben Horowitz: I believe that in the face of massive industry changes, you must recognize that some of the most basic "physical laws" have changed. Compared to the way tech companies were built in the past, AI brings two fundamental differences: First, it used to be common knowledge that "throwing money at software problems doesn’t solve them" (for example, being two years behind couldn’t be fixed by hiring a thousand engineers in a month), but that’s no longer the case. If you have enough money and good data, you can buy enough GPUs to solve almost any software problem. Second, the software industry once believed that "market dominance grants you 90% of the rights" (such as data lock-in, migration cost lock-in, user interface lock-in), but these moats have largely disappeared. Because code is easy to replicate, data is easy to transfer, and in the future, it won’t even be humans interacting with software but AI, which is very flexible in using user interfaces.
As a CEO, you must first recognize that these advantages are fading. So, what’s next? Where does your value lie? What do you offer? It turns out that many things still hold value. However, if you try to rely on old advantages for good pricing, you’ll face immense pressure. Your pricing must be based on other, more unique value you provide.
Host: We often discuss internally that in the past, if you had a good product, you might have a ten-year development红利期, or at least five years; but now, it might be just five weeks. We also talked about companies going public. Companies are staying private much longer now. If you’re facing an existential crisis, it’s definitely better to handle it as a private company than as a public one. But the "doomsday" in the SaaS (Software as a Service) field arrived because people began to question the ultimate value of companies. Everyone starts a business to create economic value and try to gain financially. But if you wait too long, maybe your company becomes worthless. This is very scary. If facing the risk of being disrupted, what should CEOs do differently now?
Ben Horowitz: Yes, I think you have to be honest with yourself about what you truly have. Some companies deserve to be淘汰, others don’t. If you push many ideas to their logical extreme, you might feel that everything is worthless, because if there are no employees in companies, who will buy your糟糕 software? The evolution of these things is often more subtle than we imagine and takes longer. So the question is, during this transition period, are you getting stronger or are you declining? If no one is buying your product, if customer funds are shifting to buy something else, then you have a huge problem, and you might have to lay off大量 employees and undergo a transformation.
On the other hand, some companies are being ruthlessly slaughtered in valuation, but their own business remains strong. For example, I’m on the board of Navan, which is in the travel business. Obviously, under the "SaaS doomsday" narrative, people think they are doomed, that there’s no future in travel. But when you look deeper, the situation is actually much more complex. In the travel business, you need well-defined relationship networks. If your company is of a certain size and requires global travel, then I need relationships with every airline in the world, every hotel, every train. You have to handle all these things, and you also have to connect the systems back to the customer's budgeting system, etc.
Secondly, giants like OpenAI or Anthropic have absolutely no desire to sell products to travel managers. No one has a channel to directly reach travel managers; you can't even imagine it being a good idea. So you want to move forward, doing what Intuit is doing, transforming yourself into a more AI-oriented company, and then retaining customers. By the way, the AI agent travel experience is actually more complex than people imagine. I don’t know if this will always be the case, but this is the current state. Therefore, I think it depends on the specific company. I don’t think all companies are in the same situation, but I do believe this is a "brave new world"; if you look at it with an old-world perspective, thinking it still follows the old physical laws, then you are dead.
Host: The boundaries are very blurry now. Creating a feature has become easy, but a feature does not equal a product or a company. Since you founded the venture capital firm in 2009 during the global financial crisis, the world has changed dramatically. How is the venture capital world today different from back then?
Ben Horowitz: It’s very different. Our first fund was $300 million, with capital from all traditional limited partners (LPs), endowments, charitable foundations, funds of funds, etc. And we just raised $15 billion for four of our seven funds. And we raised it from different types of investors. When we started, we had almost no international investors; now about 35% of the capital comes from all over the world. Technology has become increasingly important. I think we have to think about the world in a way we never had to before.
The reason we raised so much money is that the US must immediately rebuild its entire infrastructure: we don’t have enough rare earth minerals, not enough power resources, not enough manufacturing capacity, and the existing chips are extremely power-hungry and were initially designed for gaming. This kind of investment for the future world requires huge amounts of capital. It’s a completely new challenge and extremely important because the US power resources are almost exhausted now, while China’s capacity chart shows vertical growth. We even invested in a physical transformer company because we need more efficient, easier-to-manufacture power transformers. Transformers haven’t really changed since we invented electricity.
Host: There’s an old saying that the cure for high prices is high prices themselves, but the problem is there’s a significant delay in between. So now some computers are delivered without memory (RAM). You go to Dell to buy a server, and they say there’s no memory for sale because it’s all sold out. You can build new factories, but that takes 5 years. In 1999 we were building fiber optics, but most of the fiber was unused back then. Now all GPUs are running at full capacity. How do we solve these bottlenecks?
Ben Horowitz: Yes, there were bottlenecks when laying fiber optics back then too, just the bottlenecks were in different places. At that time, server transmission speeds weren’t even enough to support video playback. There were no load balancers, no application services, nothing. So you had fiber and bandwidth, but you couldn’t build applications, and most end-users weren’t connected to the network. So the system couldn’t function, and then we experienced the dot-com bubble burst. But the situation now is somewhat different because almost every link is a bottleneck.
I think what might happen in the future is that we will have enough chips before we have sufficient power resources. NVIDIA will produce enough chips, but then we will face shortages of memory and lack of electricity. Therefore, you have to carefully study where we are in each环节 of the supply chain and find ways to alleviate the bottlenecks. By the way, God bless Musk’s "Terrafab" plan. This is his philosophy; he will personally go and solve all the bottleneck problems. This is how he operates, and that’s why we need him.
Host: Absolutely. You are an expert in both AI and crypto. Now the scariest thing is that anyone can use Claude or ChatGPT to generate highly sophisticated personalized scam emails or phone calls. All means of communication seem to become unusable. I don’t know if you agree?
Ben Horowitz: One hundred percent agree.
Host: Because usually, when I receive an email like "Dear Allen from Index Ventures" with the wrong name, I can just delete it, and I’m庆幸 he got the name wrong. But what if anyone can perfectly forge content with personalization? The best way to look at an email inbox now is that it’s a to-do list with write permissions for the public. What should we do? This brings us back to encryption technology; this is why I mentioned hashing, because its original purpose was to stop spam. Do you think there is overlap between AI and cryptocurrency?
Ben Horowitz: Yes, I think everything starts with the problems created by AI. I woke up in the middle of the night one day thinking if someone used AI to fake me on Zoom, making my finance team wire $500 million to Nigeria, that would be a huge problem. We are facing several core issues: First, on social media, dating apps, or Zoom meetings, how do you prove you are a real human and not a bot? Second, I often receive AI-forged videos from family members who believe they are real. In the future, even AI won’t be able to tell what is AI-generated, so we need cryptographically strong signed content to prove its authenticity (e.g., this is really someone’s speech and not a fake). For this kind of provenance, you shouldn’t trust Google, Meta, or the US government, but rather trust the mathematical game-theoretic properties of the blockchain. Third, if we are to implement Universal Basic Income (UBI), governments are extremely inefficient at distributing money (about $450 billion was stolen in the pandemic stimulus plans), so everyone needs a secure address to receive funds; this is a cryptocurrency problem. Finally, how can AI become an economic agent (e.g., as a merchant making money, receiving money)? They need internet money (cryptocurrency) as infrastructure. Therefore, AI has催生ed many opportunities in the cryptocurrency field.
Host: So where is venture capital heading? Some say white-collar jobs will disappear, leaving only venture capital as a profession, probably because investing is betting on entrepreneurs for non-deterministic problems, and interpersonal relationships might survive in the AI era. Considering the potential changes in white-collar jobs in the next 5 to 10 years, what do you think the future venture capital world will look like?
Ben Horowitz: This is hard to predict. If we look back at the Industrial Revolution transition, the venture capitalists in railroads and cars back then eventually became banks like J.P. Morgan, Goldman Sachs. There could be many scenarios: One设想 is that the industry sees a few超大型 companies monopolizing everything, and venture capital firms move upstream; another设想 is that large AI labs reach the limits of intelligence and are nationalized, becoming public utilities like electricity, and everyone builds on this infrastructure, which would be a completely different venture capital world. Power shortages could also lead to large companies controlling all GPUs, or conversely, push computing power towards edge devices (like phones running small models). Future venture capital might become larger and more exciting because everyone in the world becomes an entrepreneur, or it might become more difficult to build new companies, like in the later stages of the Industrial Revolution.
Host: With 8 billion people now able to turn the ideas into reality (whether it’s writing code, composing music, or making movies), this is actually very exciting. How can we make this technological transition seem less scary and reverse that dystopian narrative?
Ben Horowitz: From a macro perspective, the history of technology is a history of things always getting better. But the transition period is always frightening. For example, in 1750, about 93% or 94% of people in the US were farmers. If you told a farmer back then what a "product marketing manager" was, he would think it’s the stupidest, most不可思议 job in the world because you neither produce food nor build houses. Most people find it hard to see the other side of the transition. The great economist Keynes during the Great Depression predicted that when material abundance is great and basic needs (housing, food) are met, people would only need to work 15 hours a week. But he didn’t realize that human ability to generate new demands is incredible; we need a car for every person, televisions, computers, even exquisite meals prepared by chefs for 10 hours – these demands didn’t exist back then. In the next 15 years, I think everyone, both in the US and the world, will live a better life in terms of quality of life, luxury享受, and access to information than the person with the best life in 1980. So, although the transition is unsettling, we really shouldn’t be angry about it.
Related reading: a16z Wealth Manager: Embrace 40% Market Drawdowns, Don’t Invest 80% of Your "First Pot of Gold" in Friends' Startups
