Investing for the Long-Term: A Conversation with Ruth Porat

First, I have to say how amazing it is to walk on stage and see a room full of women. So it is wonderful to be here. To all of you, thank you.

Well, actually it goes back to 2010, and I would say it even goes back earlier than that. When Google went public, a core part of the founders' letter for the IPO was, we will invest in long-term growth. And we were always very deliberate about what that meant. So even at a company as innovative as Google, we said, let's create organizational structures to nurture entrepreneurial innovation above and beyond everything we're doing in our business. So we created something called the 20% project, where if you had a great idea, you could go off and pursue it. Why am I going back to that? One of the first things that came out of the 20% project was Google Brain, which was our first AI business. In 2014, we acquired DeepMind. And then only in 2016 did we say, you know what, now is the time to really pivot from being mobile first to AI first. And I think one really important lesson from that is you must invest for long-term growth. By the time you need it, it's too late to say, 'Okay, now we're going to start. It's 2016.' So when our CEO Sundar Pichai said in 2016 we are pivoting to AI first and it will be a full stack approach — at the time it wasn't really clear what that meant — but the foundation was already there. It was an extraordinary research team. It was the development of models that were needed. We had already been developing our own proprietary chips, TPUs, and then already using machine learning, a precursor into what we have here today, to improve the customer experience both for consumers and for enterprise and for government. So it was really at this point saying, full stack approach, everybody, eye on what that means for model development, for chip development, infrastructure development and deployment.

So there are a couple parts to that. One, we've said publicly that we're looking at about $200 billion of capex this year, and that we expect that will increase significantly next year. And we've also said, as have a number of our peers, that we are capacity constrained. The demand for compute, because of the value that our partners are seeing in the enterprise, in governments, what we're seeing across our platform, suggests that we're the buildout here now. If you could snap a finger, we could deliver value. That's an important point, but I bet if I did a show of hands, everyone here would say, but we're so early on the journey. And I think that is true. Many of us are still grappling with, if you do triage on the full set of opportunities within our enterprise, where can we get the highest return? One of the things that struck me is when Demis Hassabis won the Nobel Prize, and you and I have talked about this, he was embarking on something that had been a 50-year grand challenge, and that was to predict the protein structure for all proteins known to humanity. This comes back to your question. And at the time, it took four years for just one protein. There are 200 million proteins known to humanity. And members of his team said, 'How is this even possible?' And his answer was, 'Why not?' And his answer when he was awarded the Nobel Prize shortly thereafter, four years later, was the benefit of this technology is that we can solve the most intractable problems that we have in each one of our businesses, government, wherever it happens to be. That suggests that really the magic in this, and the answer to your question, is how deeply do we dig into sort of creativity about how do you expand the scale and scope of your business. So we do see this as an opportunity that extends for quite some time. Point number one. Point number two: we, each of us, to generate an attractive return on investment, need to be focused on two things. One, what is the value that we add to partners so that you can actually deliver attractive revenue growth. And the other is how efficiently do we deliver that compute capacity. And so on everything from model efficiency to infrastructure efficiency, we have historically and continue to be very focused on driving greater performance per dollar of investment.

So it's interesting you use the word liquidity. When I arrived at Google back in 2015, I was asked a lot about the financial crisis. Probably one of the most meaningful roles I played in my career was advising then Secretary of the Treasury Hank Paulson. And one of the key lessons from the financial crisis was that you should identify your greatest source of vulnerability early and fortify against it, because you can't address it in the moment. Now for banks, it was liquidity. Without liquidity, you couldn't run a trading operation. Without liquidity, you shut down. And I've been asked that question frequently since. What is the greatest source of vulnerability today? I think for all of us across industry, one of the most important elements is innovation in your business, because if anyone is connecting with your customers in ways that you are not, because they are using AI and you are not, or if anyone in your industry is looking about how to manage supply chain or other elements, you're at a competitive disadvantage. So that's one part of it. I appreciate that's not what you were asking. You were really going to, but what about this question about funding and financing? And I would say one of the important lessons there from other platform shifts is it is imperative that each one of us is delivering differentiated solutions. I ran technology banking at Morgan Stanley in the late '90s, and what you saw going through the late '90s is those that really kind of sustained growth over the long term were those who were using technology in a differentiated way. If you, at the time I said, if you're just, if you're using the internet just as a way to have an online catalog, that's not a sustainable differentiator. Similarly today, if you're using AI simply to try and extract efficiency from a business, that's nice, but that's not long-term sustainable. The real question is what should be different? What is available today that wasn't available previously? One of my favorite examples on that, probably because I spent so much of my time in my career in financial services, is in financial services. If you think about the service that was delivered to ultra-high-net-worth individuals for decades, it wasn't available to everyone else because the cost of delivery was too great. But with AI, and agentic AI, the ability to actually aggregate more information, deliver great insights in a more cost-effective way, means unit economics have gone down. You can deliver for an entire customer base you previously couldn't deliver for. That is better for everyone in society, that everybody is getting access to quality information. But most important, when unit economics can change, you can identify new markets that you maybe wanted to identify in the past but couldn't. Which is why I think that Demis' comment that we should each look at the most intractable problem, a market you wanted to address, a way you wanted to deliver that you couldn't, and this will make a difference for it.

So I think the word 'win' in the context of this technology platform is an important one to drill down on. Oftentimes it's viewed as where are you at any particular point. There's third-party leaderboards that gauge sort of where are you on reasoning or logic, math, etc. Our view is the most important element is to, if you want to call it win, lead in responsible deployment of technology, because the thing that is true throughout history is anything that is not built for the long run won't endure for the long run. And so in our view, the single most important thing we're focused on — well, the single most is three. So the three most important: if we don't collectively fortify the downside risk collectively, we will not have the opportunity to capture the upside. And you've hit on one part of it, but I think very importantly where we've been focused literally since we said we're shifting to AI first is, one, what about governance around models. So concurrent with announcing we're moving to AI first, mobile first moving to AI first, we said we're going to develop and operate with a code of principles, and since then have been very focused on what's the governance framework that is needed to provide the responsibility, the scientific rigor, the resilience that's needed with a technology that continues to evolve. And our colleague Demis Hassabis, again, the Nobel Prize-winning scientist, amazing leader, has codified that, really bringing in what across Alphabet we think is the most important way to approach this, which goes to levels of transparency, testing models before release, bringing in a whole of community — government, developers, those of us who are model builders out at the frontier, etc., academia — to actually have a rigorous foundation. That's one part of it. The other part of it is how to ensure when you're building the infrastructure to deliver this, it's done responsibly. And the third part that's super important, you and I have talked about it, is ensuring that this upside benefit actually addresses the needs across economies, across society, so that we're narrowing inequities in society, addressing inequities. So again, it builds something that is sustainable.

There always are. And platform shifts, fortunately, I think hopefully the market shakes that out.