AI 将远超此前所有技术革命：CES 2026 聚焦机器人、制造业与 AR 眼镜

We just had a very successful all-in spirited full contact debate in front of 1,700 people. A packed house. There were no open seats. Is the power of the all-in brand is really to have the conversations with a little bit of fun, a little bit of spiciness and no question, no topic can be banned. There was no censorship. No censorship. We're going to just go right after the hardest topics. But we had a really fun time. Discussed so many important topics. What a great panel. These guys are tip of the spear in terms of doing really exciting things in business.

Thanks for coming out, everybody. We're going to have a great, super-hardcore discussion about the future, specifically around AI, which I think is the most important theme not only of CES 2026, as we've seen with all the incredible gadgets and chips being launched and self-driving, but it's going to be the most important transformation of our lifetimes.

I think everything we've seen over the last 30 years of technology—from the PC revolution to cloud computing to the internet and mobile—all of that is going to be dwarfed in comparison to the impact that AI is going to have on society. If you're here at CES, you know that you're here for that reason.

We've got two amazing guests who are going to join me to have this debate. Additionally, I've brought my box, a box filled with all the ghosts and gadgets of Christmas past, and we're going to go through those at the end of our discussion. But here's a quick video of our guests who will be joining me today. From boardrooms to the White House and beyond, McKinsey's influence in business is virtually unparalleled. It's one of the largest and most influential consulting firms in the world. Enterprise can move faster than any of us expected, which is good news because these problems aren't going to be the problems of the next generation. They're going to be the problems of our leadership generation. Making this system of government better on both efficiency and effectiveness is key for economic growth and for national defense. We also think that some of the private sector insights that we have brought to the public sector can drive innovation. Our next guest leads venture capital firm General Catalyst with 40 billion in assets under management as of midyear. Our aspirations in venture capital is to be the best seed firm in the world. The decisions we're making, the companies we're building are going to impact the world for centuries to come. Ladies and gentlemen, please welcome Bob Sternfels and Hemant Taneja.

All right, gentlemen. Welcome. All right, team, let's do it.

How do you look at the pace of innovation and change in this past 2 years since ChatGPT was launched compared to the first 30 years of our careers? We're all of a certain Gen X age. Compare the last 2 or 3 years to the 30 before it.

Yeah. Well, first, Jason, thanks. It's great to be up here with you guys. I would just say this week is amazing. I think there's over 150,000 folks here this week. You talk about CES being back—I think CES is back. And that's great, right?

With all of these things happening, I think there's such a premium on folks from different perspectives getting together, because that's where new ideas are created. My big hope, and why we're here, is that when you mix and mingle with different folks, you come up with new things. The world needs new things.

What I love is that you mentioned a lot of the tech leaders. What's exciting about this is I think everybody sees tech as part of the equation. When I look at the folks here at CES, you see not only the technology leaders and the investors, but folks from almost every industry vertical. They're here now because they know that technology doesn't sit on the side; it's central to everything we do.

To get to your question, look, I think we're moving at literally warp speed now. It's just night-and-day different. It's almost a B.C./A.D. type of thing when you can see the change of pace. I haven't met a CEO yet who isn't talking about, "How do I get my organization moving faster?" It's quite frankly less about strategy. It's more about organizational speed.

Hemant, how does this feel compared to our first couple of decades, where companies would take 2 or 3 years to release a product, and now companies are releasing products in 2 or 3 weeks or 2 or 3 months?

Yeah. So, look, the world has completely changed, right? We've often said this is peak ambiguity. You have massive geopolitical change. You have an incredible amount of change around every country trying to drive strategic autonomy in different industries. All those dynamics keep changing: alliances, the new world order, everything.

Underneath that, our tool of implementation is technology, and that keeps changing, right? What the technologies can build today versus what an LLM could do, let's say, 2 years ago—or November 2022, when ChatGPT came about—is fundamentally different.

What are you building toward, as to what the world's going to look like, so you can have enduring value? And then what are you building with, where the technologies you're using aren't going to become obsolete and destroy your value proposition over time? It's just all kinds of change, and so it's a really dynamic time.

The other thing you will see is, you know, we invested in Stripe in 2010. It became a $100 billion company, let's say, 12 or 13 years later. You look at Anthropic, which we're also investors in, and that goes from a $60 billion valuation last year to, you know, a couple hundred billion.

By the way, with good economic progress, these are not pie-in-the-sky valuations. They're based on actual growth of the business. That goes back to your point, which is the compression of how fast value can be created when code self-writes and access to distribution changes.

Fundamentally, it's just really exciting, and I think it's going to accelerate from here. This was one of the statistics we would look at in venture capital: How long does it take this company to get to $100 million in revenue? How long does it take to get to $1 billion in revenue?

Unpack Anthropic and that journey, because this company's revenue—and you have OpenAI, obviously, contemporaneously trending toward $20 billion in revenue a year—where's Anthropic at, and what's the revenue mix? Where does the revenue come from?

Well, look, Anthropic builds language models. It's got some of the best models out there. There are a couple of companies that are doing a good job at that. Then they've got Claude on top, which is, to me, the essence of transforming the engineering department of an enterprise, right? That's a killer application where everybody is now using these tools.

That business, when we invested, was doing about $880 million, which was 10x growth from the year before.

10x in a year?

10x growth from the year before. And then this last year they've announced that they're growing another 10x or more.

When you look at that, we invested at the $60 billion valuation, assuming it was going to be, like, 3x growth from there, because those are staggering numbers. And if it does 10x, you can't predict it, but to see adoption is so fast.

We ended up investing at an $8 billion, $9 billion, $10 billion kind of run-rate business at $60 billion. That's the cheapest deal that got done last year in venture capital on a financial basis.

We just have to get our heads around what scale really means. Are we in the business of creating what we used to think was, like, "Can we create decacorns?" Now we're talking about, "Can we create trillion-dollar companies?" That's not a pie-in-the-sky idea with Anthropic and OpenAI and a couple of others. The game-changing scale of technology is fundamentally different in what it can do.

Bob, what's behind this massive revenue ramp? You get to see all the incumbent businesses. You get to see the elite businesses that are growing 2x or 3x each year. You also get to see the ones that are struggling, and then you see these large numbers and 10x growth. What's driving this in your mind, and is it sustainable?

I hate to give you the classic consultant answer, but I do think it depends. I think we're at a tipping point this year, and I'll tell you why.

We work with most of the large enterprises in the world across all industry verticals, and what we have seen is a huge uptake in leveraging these technologies, like Anthropic. We're leveraging Anthropic, and large enterprises are using technology at a scale and rate that they haven't before.

If you look at IT spend as a percentage of revenue, et cetera, all this stuff has gone up, and I think that is propelling the 10x to 10x. The conundrum is—and it's been widely written about—realizing enterprise-at-scale value in nontechnology companies is proving harder than people think.

Got it. So, in plain English, that means, hey, you've got a travel company. There's somebody deploying AI, and you're watching what's happening at Tesla or Google, and they're getting these phenomenal results, but maybe that legacy business is having a harder time achieving those results.

I'll make it even simpler. A typical nontech CEO might say, "Hey, Bob, do I listen to my CFO or my CIO right now?" The CFO is saying, "We've spent all this money. Why do we need to be the fast adopter? I'm not seeing the ROI yet. Can we pause?" The CIO is saying, "Are you freaking crazy? This is the moment that, if we don't, we'll be disrupted."

We think—now, I will say the shining part is, I think there's a path where you bring those two together as allies.

And you say, “Yeah, but let’s rethink this, get out of pilot purgatory, really rethink the reorganization, all this stuff.” There is a path, but I think right now most CEOs are getting torn a bit between, “Do I listen to my CFO or do I listen to my CIO?” I think this is a really good jumping-off point, Hemant, for your strategy at General Catalyst.

You and I have known each other for a long time—really a long time, decades—and you always prided yourself on being the great seed fund. We’re going to get to these companies when they’re $10 million and 10 people and put that first check in. But then I saw this news item go by a month ago that you raised N billion, and then I see you’re buying companies. So are you out of the seed business and now doing random acts of private equity? What’s going on here? Explain to me the strategy at General Catalyst.

How much time do we have? It’s going to take some time. So, look, we very much view ourselves as venture capital for America. For the 25 years we’ve been around, it has been about meeting founders where they are. What that means is essentially helping them navigate ambiguity in the past when they’re beginning and the business isn’t clear, all the way to figuring out how to scale in the complex markets that they go into.

Everything we’ve done has been in that context of creating these catalysts: the flexible capital they need, the policy capabilities they need, the market access they need, and those relationships globally to actually build an enduring company. So that hasn’t changed.

So why did we acquire a health system in Ohio? It was a nonprofit. We worked with the attorney general and converted it. I say this, by the way, with a great sense of responsibility, because that’s a community in Akron, Ohio, that we take care of. So that hospital has to continue operations in all the dimensions. It takes care of people.

We bought it to actually have a place where we can work with our founders and transform with AI, create abundance and resilience for this health system, so we can take care of the people a lot better. And if we did that, then we can go do that for the other hundreds of systems across the country.

Some of it is market access. It’s very hard for healthcare startups to deploy successfully at scale in these systems. We’re going to go show how. We’re going to actually go on the ground with them and show the world how, so it can transform the health system.

The other point about buying companies: We look at that as a lot of workforce transformation happening. Bob and I talk a lot about this. I think work fundamentally in these companies is going to change. So if you’re a call center in an emerging country today, that’s a declining asset value because you know it’s going to be displaced with AI.

So we look at that and say, “Well, those are customers on the other side. If we bought that as a piece of the puzzle, we could work with an early-stage founder to learn how to quickly accelerate adoption of AI into the call-center space, serve these customers, and scale a lot faster.” The compressed value creation we’re talking about—that is a new playbook.

So this is not about trying to be PE. This is about acquiring businesses in PE that actually have declining value but have important customers that need to be served, and helping them get to that AI transformation Bob’s talking about faster by getting our founders in there.

This is extraordinary, Bob. When you think about it, just so the audience can get their head around this, venture capitalists used to back founders to then be the barbarians at the gate trying to take on these big industries. Now these big industries, in some cases, are in significant decline, struggling, and the venture capitalists are coming in and saying, “We’ll just buy the castle, open the drawbridge. We’re going to buy it so that we can take our startups and accelerate—whether it’s healthcare, financial services, or customer support and outsourcing, business-process outsourcing.”

Essentially, we don’t care about that business economically necessarily as much as we care about it for access to that customer base. Running McKinsey, this is a playbook that is like coming out of the future in a time capsule and saying, “We’re going to just upend the entire ecosystem.” Yeah.

Yeah. I mean, I just gave a talk at a university and was talking to some potential folks to join us, and I said, “Look, I’m jealous. I’m jealous of all of you because you have a lot more time to do what we do than I do. And you’re doing it at a time where it’s going to be a lot more exciting.”

What I love is that, effectively, you’re creating a new asset class, right? This is not private equity. This is about how you transform incumbent entities into something different, right? Private equity typically optimizes an existing asset class at a certain scale. This is about transformation.

So you think of a large existing enterprise, and I think you have a choice. You have a choice: transform or die.

And so there’s this wonderful moment, but because of some of the incumbent advantages, I wouldn’t say that it’s predetermined which way you’re going to go.

Right. You can actually do this quite quickly, and I think you’re showing the power of private capital can actually do this.

So we’ve chatted a lot about this, right? One of the things, when you think about transforming a large enterprise, is what do you really need? You need a few pieces. One is data infrastructure that can ready you for the enterprise. You need the models adapted to you, and then you actually need a new model for how the workforce is going to function, because you have agents and humans and there’s a massive change-management exercise.

A lot of our partnership has been about figuring out what that new model is going to be to transform these businesses. What does that mean when you get on the ground and look at department by department? Take HR: How do you drive transformation of healthcare and how you take care of your people? That process is horrible today, and we have a business called Transcarent that essentially creates abundance in that regard, which uses AI.

You have direct access, telephonically, to all kinds of healthcare services and can be routed whether you need surgery, cancer therapy, or mental health. It does it in a way that is seamless, cost-effective, and helps enterprises take control of their cost structure.

You need coding to fundamentally transform. That’s what Anthropic does. There are companies working on transforming the call centers. There are companies working on transforming your sales and marketing. But when you have these technologies in there, how are the actual people going to do their work in concert with these agentic capabilities?

That is a whole new model. You guys are inventing a lot around that, because there’s a lot of innovation that needs to happen in that whole workforce transformation. I think that’s ultimately where the rubber is going to meet the road on how quickly teams embrace it, customers embrace it, and we can actually diffuse AI into these businesses.

And Bob, you’ve had to deal with this internally at your organization. What’s the right size? And what happens when a piece of technology takes a career and takes out the first 5 years?

What happens to an organization when you just basically gut the first 5 years of development? This is why some people in the economy are looking at AI and they’re scared, and they’re looking at AI and saying, “Is this going to benefit me, my family, my kids who are graduating from school?”

This technology—I think management consulting is the perfect place to look at it. Tell me honestly: The first couple of years, you’re training up one of these really smart kids to write up reports and do analysis that can be done with AI today perfectly, close to perfectly.

Yeah. So I’ll give you a couple of stats first on us, then more generally: 25 squared, 40,000, and 25,000. What do I mean by that?

So let’s look at McKinsey as a bit of an incubator. The 25 squared is that we’re simultaneously doing 2 things at the same time. We have client-facing folks, which most of you in the audience would know and think about when you think about a McKinsey consultant. We’re growing that body at 25% next year—25%. An unprecedented number of new hires, because the work is changing.

They’re not doing the stuff that you talked about. We saved—we looked at it—we saved 1.5 million hours in search and synthesis last year, but we’re dividending that to solve more complicated problems and do different things. You’re probably sick of McKinsey charts out there. We have agents that do this. They just gave you 2.5 million of them in the last 6 months. I want to get rid of charts.

But the consultants are doing different things. We’re adding 25% to that body.

So they’re moving up the stack.

They’re moving up the stack and doing these more complicated problems. At the same time, though, about half of our firm are non-client-facing folks. We’re down 25% in that group, with a 10% increase in output.

And so simultaneously—and I know this is hard also, particularly for folks to get—we’re going to be adding and shrinking simultaneously with the 2 halves.

And this has never happened in the history of the firm. Our model has always been synonymous with growth only occurring with total headcount growth. Now it’s actually splitting. We can grow in this part, the client-facing side, and we can shrink in this part and have aggregate growth in total.

And that's a new paradigm and a new dynamic.

We're seeing this in venture. You and I were around for the days where you'd give a team $3 million, and they would come back in 18 months having spent it on data centers and building a team of 20 or 30 people. Then we'd see the first version of the product 18 months later. And then the $3 billion now.

Yeah, it's insane how much more is getting done with less. And so, do we worry about society's and our industry's ability to communicate this change to society?

Exactly.

What? Why? How come? It doesn't make any sense. And then young people are graduating, and they're sending out 100 or 200 resumes and getting no job offers.

We were sitting here 10 years ago. Every graduate from a decent school was like, “I have an Uber, a Coinbase, and a Google offer. Which one should I take for $150K?” And those offers just aren't there. So how do we communicate better as an industry? And what's the advice to young people coming into the workforce?

Yeah, look, every company in Silicon Valley—I was just going to say Silicon Valley, but broadly in the tech industry, in the startup industry—essentially looks like a C corp with a bunch of engineers. But in a world where code writes itself, what is that next-level innovation? What are these companies actually going to do? I think that's ultimately the transition we're going through: What does innovation actually mean?

It's going to be less about being able to write code fast. It's much more going to be about, systemically, how do we adopt this into the world? And, to your point about ambiguity in the opening, because we don't know about the capabilities of these technologies or how the world's shaping up, there's a lot of ambiguity.

So to me, the companies that do really well—and the way we guide the founders—a lot of it is: become iterative, constantly change, constantly move forward, as opposed to what used to be before: become precise, find this narrow edge, create your growth loop, and go build a company.

Now it's like: constantly iterate. And in order to have customers give you the license to iterate, it comes down to trust and relationships. So founders that are very good at engaging with customers, building trusting relationships, and saying, “Hey, we're going to go figure this out together. We know how to leverage this technology, but we don't really know how it leads or what the possibilities are. Will we co-create?”

The advice I always give is, it's all about radical collaboration. In this next phase, we've got to figure this out together, where different stakeholders that all touch a system are figuring out what this means to them, and then what it means in terms of an overall optimization and the transformation that we can do with it.

You know, one maybe exciting part to this, because I think you framed it as: you're a graduate, and how do you get into the workforce, and is it getting tougher? We did a little bit of work that asked what kind of skills folks are going to need in an AI-infused world.

From an employer's point of view—less so the startup, more an at-scale enterprise—what can the models not do? And therefore, what role will humans play?

So, leadership and goal-setting.

Human.

Judgment, right? And we've seen a lot around evals in this room. But there's no right and wrong in these models. So how do you set the right parameters—the architecture—based on firm values, based on societal norms, whatever? How do you build the skills to set what the right parameters are?

And then finally, true creativity, right? The models are inference models—the next most likely step. How do you think about orthogonal stuff?

Some of the work we've been doing with large enterprises, if you believe in some of that, can take you back to challenging some of your assumptions about where you look for talent. It actually means that where you went to school matters a lot less.

So do you start looking for raw intrinsics? Can you widen the base? Can you actually look at—let's take a tech background, not which university you graduated from, but what does your GitHub profile look like? Let's actually get to the content. And could that actually start meaning that a wider set of people can enter the workforce through different pathways?

One of the things you said that really resonates is around creativity, because when we were going to college, it was all about learning how to solve problems really well, right? And now, in a world where we have this technology that can solve problems for us, it really is about asking the right questions. It's like going back to that Socratic dialogue.

It is about creativity and who can imagine best what the world's going to look like and then leverage these technologies to go shape the world toward that. To your point about vision and teaching our kids, I get this question a lot: What do you want? How do you want your kids growing up?

It's like learning how to ask the right questions versus solving hard problems. It's a very different mindset, and it is about curiosity—and kind of back to being kids when you're growing up. It is about challenging your curiosity. Can we actually rethink our pedagogy in a way that we can develop this next generation to be more that than, “It's 8:00 on Wednesday morning and I'm going to factor polynomials because I'm in 7th grade,” which is what our system looks like today?

Yeah. The advice I've been giving to young people is: There's nobody coming for you. There's no training program. You have to make that for yourself.

And do not go in through the front door with a resume. Just email the CEO of the company and redesign their landing page. Say, “These are the 3 things that I think could be better, and I saw you speak on this podcast. I think your company is incredible. I would love to come work there, and I did this spec work.”

Now, people are like, “Why should I do free work to get a job to prove you actually have a skill that is meaningful?” You're not going to be able to get into a training program. So many folks now in corporate America, especially the people who are onboarding people, are just like, “Hiring somebody and training them is going to take longer than building an agent. I can build an agent.”

Young people coming into the workforce whom I have to train are annoying. Setting up an agent that just does the work is easy. That's the game on the field right now that people don't want to talk about.

Which means, to stand out, you're going to have to show chutzpah. You're going to have to show drive. You're going to have to show passion. And what college is doing that? What college is teaching that? What course is that?

Look, I think there's a massive gap in resilience. Yes—resilience. Because what you've got under that is: you're going to get knocked down.

Yeah, right. The question is, do you get back up?

And how do you get back up? I think the educational system today doesn't necessarily build institutional or individual capability in resilience.

If we could wave a magic wand—just to go off on a complete tangent here—what should the education system look like in 2026? Because you're buying businesses, and you have one in health care. That's one of the 3 hardest businesses to make change in, historically.

The other 2 here in America that have the most regulation, are the most expensive, are the hardest, and that Americans are suffering under the most, are housing and education. Those are the 3 big ones. When I run for president, that's going to be my platform: those 3. I'm going to solve those 3.

But go ahead and solve education for us right now. And are you going to buy a college next? [laughter]

Transform? Basically buying all the businesses that make no money. Is that where we're going?

I would say—

Oh, the businesses that are the most—

Yeah, but the ones that need to endure for the longest, actually. That's the way I look at it.

So here's the thing about education. This idea that we spend 22 years learning and then we spend 40 years working is a broken idea. If the learning and development of technology is going to be so dynamic, what about going from a 4-year college to a lifelong college, where your relationship with learning is a lifelong skilling and reskilling outcome experience?

We've talked about this before as well. There are some innovative college presidents who are thinking about that, which is, first of all, better business and better lifetime value if you're a college and you actually have a client or a student for perpetuity, versus paying you for 4 years.

It's much more useful for us to be able to have that capability and constantly learn what these technologies are doing, how the workforce is evolving, and how to stay ahead in terms of where the opportunity is. Learning has to become much more fluid, and we need to become a community of lifelong learners as we adapt to a world where AI is diffusing through us over the years.

And I would just add: I’m with you on this. The system built close to 700 years ago was designed around a high fixed cost—libraries and professors—to then take you out for a finite period of time to learn, and then effectively you’re set off into the workforce.

If you start to think about the half-life of skills getting shorter and shorter, we’ve done some work at the McKinsey Global Institute that said, for an employer, the return on investment that you give an employee in terms of skills has shrunk by about half over the last 30 years. It used to be about 7 years’ return; it’s less than 4 years—about 3.6 years now—and that’s only getting shorter and shorter as things change.

Absolutely.

Where is it working really well, and where is it not working well? It works when you have a specific domain area where value can ultimately be created. For us, that’s in structured problem-solving, search and synthesis, and more effective communication—these types of domain areas.

But where I was going with this is, the skill is: Are you skilling people to actually become superhuman by leveraging agents? Do you have that ability? One of the things that we’ve now indexed on—and I mentioned this—is 40,000 and 25,000. That is the number of humans we have and the number of personalized agents we have as of last week at McKinsey, and I think we’ll be at parity by the end of this year. So you’re literally deploying agents that can do a full 360-degree trusted job function.

Right. Right? That becomes a skill. And I don’t think we’re actually equipping people for that right now. It’s a bit more random, or sometimes actually excluded in the classroom, as opposed to embracing it and figuring out how you actually take advantage of it.

It’s almost like we need to train people to go from being part of the orchestra to everybody being the conductor.

And everybody having their own orchestra of agents working for them. I always look to startups because they’re resource-constrained. I was at a dinner in Singapore, and I had a dozen founders there. I said, “Has anybody hired anybody in the last 60 days?” They all raised their hands. Then I said, “Okay, how many of you have an HR person who wrote the job description?” Nobody raised a hand. I said, “How many of you typed into an LLM, ‘Write a job description for this’?” All 12 hands went up.

So now you think, in HR, the entire blocking and tackling has been writing the job description and sorting through the resumes. Then I asked the next question: How did you sort through the resumes coming in? They said half of them had built agents—

Mhm.

—to sort through the resumes and stack-rank them using AI. I said, “Whoa, holy cow. This is like the typing pool, the mailroom, the photocopy room—for those of you who are under 40 years old. We had a room called the typing pool. Then we had one called the mailroom, where packages came in and messengers worked. All those went away. That floor of the building got redeployed.” I think that’s what we’re going to see: the HR department, the legal department, all getting compressed. Really interesting.

It’s already happening. As we think about our own transformation for our own business, we basically say every department needs to have AI teammates now. Are those AI teammates like a copilot or pilot? Can you fully empower them to do stuff, or are they giving you efficiency? That depends on how well the technology works, how complex the problem is, and how severe the problem is.

In healthcare, for example, if it’s life-and-death decisions, you want humans making those today because that technology isn’t as reliable. So I think having a framework but saying every one of your departments is going to have these AI agents—if you’re not doing that, then you’re not preparing yourself for this next phase. That’s a lot of what you’re seeing. You’re already going to be one-to-one; that’s an enormous ratio.

Well, the problem, I think, Jason, that you alluded to earlier in this is there’s all this potential, but folks aren’t thinking through the dynamic implications in their enterprise model versus the static.

The static might be, hey, there are all these departments; I can apply this, I’ll radically shrink it, I’ll reduce the number of layers in an organization, and I may slow hiring on the inbound, to your point. The dynamic is, okay, but what does your company look like in 5 years’ time?

What I also often ask a CEO is, okay, you’re doing all this stuff—what’s the pathway to your job?

How does somebody get to your job in the organization of the future? You had a pathway. It’s not going to be that same pathway, but if you don’t hire inbound folks, you can’t continually laterally bring in a CEO. It’s literally like taking the bottom 4 rungs off the ladder to save money today.

And then everybody’s jumping up trying to get into the organization. It’s like, well, we don’t have a path there. You’re going to have to be really thoughtful about making that investment. It feels like the first 2 years of AI were about cutting jobs, and we really need to think about, hey, it’s not just about efficiency; it’s about opportunity.

Exactly. What’s that other 25%? Right. That’s what I think we’ve got to lean into. Let’s take a little diversion here before I open my box.

The black box.

My box here of all the great CES innovations over the last 20 years. Physical AI. We’ve been talking here very cerebrally about what’s happening in enterprises and what’s happening in software, but self-driving is probably the theme. I would dub 2026 CES “self-driving CES.” I will dub 2027 “robotics”—humanoid robotics specifically.

We’re starting—obviously, people are showing off all these incredible robots here—but I think consumers will be experiencing them in 2027. Consumers are experiencing self-driving this year. Nuro and Lucid have an incredible product. Zoox has been here. Obviously, Elon is doing great things with robotaxis. It feels like he’s closing in on a solution and getting very close. Waymo obviously is leading the pack.

But then you also have Baidu, Alibaba, WeRide, and Pony.ai. This is a global race. What will the world look like in 2026 in terms of self-driving, and what are the second- and third-order impacts of that? Then do the same for robotics.

If you go around the world today, you go to Europe, you go to the Middle East, where there’s a focus on interesting luxury products, there’s a market for it. BYD and a lot of these Chinese companies are actually penetrating deeply everywhere because these companies have all the features and functionality, and they’re really low-cost.

So one thing is that the dynamic of the auto industry—and European automakers are all very dejected because they don’t know how they’re going to compete with Chinese industry—the U.S. has innovation in self-driving, which allows you to say the next generation of winning automotive companies will take advantage of this platform shift.

The U.S. has the technology, but it doesn’t have the manufacturing capabilities to actually make it as cost-effectively as a Chinese maker is going to be. So it’s not as easy to figure out how the world order around automotive is going to reshift around the world.

Part of physical AI and the use of AI in manufacturing is to figure out how you design and manufacture next-generation products right here in the U.S. in a way that mimics the cost advantages of China, so that our innovation can then carry the day for us to be the global leaders yet again in this next phase.

We have a company, Re:Build Manufacturing, that’s focusing on this, for example. There’s a lot of focus that needs to get on that, because if it’s self-driving and it’s not cost-effective, some of us will buy it, but it’s never going to be a mainstream product. Cost has—I mean, there’s a reservation price that really shifts the demand patterns around automotive. You probably have good data on this as well. We should talk about that. But we’ve got to get the AI right, and we’ve got to get the manufacturing cost right as well.

No, I think we’re going to see a massive move down the cost curve on this. I’m with you, Jason. I think we’re going to see, literally over the next 12 to 24 months, a massive transformation. I think the race is afoot, right? The race is afoot between, let’s say, a Western stack and a Chinese stack on this, and in the rest of the world, it’ll be interesting as a battleground to see where that plays out.

You and I were talking a little bit about this. I think that is a massive trend. I think a larger trend will be the trend to robotics, and not just for human interaction but in manufacturing. When you think about the challenges that the Western world faces, take the U.S. I was talking to the CEO of one of the large contract manufacturers, and she has 50,000 job openings right now for U.S. manufacturing jobs in America that she can’t fill.

Right? And our demographics aren’t getting better on this front. Germany is in an even worse situation. Korea—

Germany—

Like, another level.

Yeah.

And I think the only way that you build resilient supply chains at the cost point that you’re talking about is that robotics is going to be at the heart. This race, I think, is wide open. Korea leads the way in robots per worker. They’re about 1 to 10 right now.

Germany and China are tied for second, and the U.S. is a distant third. So there’s a real race. You talked about autonomy; I would actually jump to robotics and wonder how I—

One of the issues in robotics is that, when you build the LLMs, you can dump them in the cloud, experiment with something called ChatGPT, and it becomes pervasive. If you have good robotics models, what’s next? You don’t have a hardware capability that’s like an API infrastructure that diffuses those models quickly. There’s a lot that needs to get built.

I actually think robotics will be slower than people think in terms of really taking hold. But it’s essential to lead in that if you’re going to lead in manufacturing and therefore have that core advantage to play up the stack in industries like automotive. There’s no other way to do it.

Yeah. I don’t want to name-drop, but 2 Sundays ago I went to Tesla with Elon and visited the Optimus lab. There were a large number of people working on a Sunday at 10 a.m. I saw Optimus 3.

I can tell you now, nobody will remember that Tesla ever made a car. They will only remember the Optimus and that he is going to make a billion of those. It is going to be the most transformative technology product ever made in the history of humanity, because what LLMs are going to enable those products to do is understand the world and then do things in the world that we don’t want to do.

Yeah.

I believe it’ll be a 1-to-1 ratio of humans to Optimuses. I think he’s already won, but I don’t want to speak out of school. But I do have a box. We go to the box.

I have a box, and these are all really interesting technologies that we all got to see. How many people owned one of these? I mean, Michael Douglas made this famous. Remember Wall Street, on the beach making trades?

There was an amazing AT&T “You Will” commercial. Remember the AT&T “You Will” commercial? This was one of them: “You’ll be able to work remotely from the beach.” What is the equivalent of this today? What do you think we’re going to look back on this year and laugh at in 30 years? This is something from the ’80s, so I guess this is 30 years ago. What are we going to look at that we’re all enamored with today that we’ll get a little guff out of?

Well, I’ll tell you. By the way, I love it. It says “California mobile phone” on this. That was the brand associated with it. Two memories come to mind for me on this. One was envy, because when I started, only the most senior people could get one of these and I couldn’t. I was just like, “I want one of those.” It was $4 a minute. What was it—$3 or $4 a minute?

And its battery lasted about 30 minutes.

Exactly. Some new associate doesn’t get one of these.

When did you have your first mobile phone?

I’m too young for this.

Too young for that? You’re such a liar. You had the StarTAC like me.

But the second—and this was made infamous—was one of the great things, unfortunately, great failures that we had. We did a project that was published a while ago for AT&T in the mid-’80s that said these things were never going to take off.

I don’t know why you burned me with this one.

By the way, let me remind you of something today to answer your question. Think about a lot of the eyeglass innovation that’s happening. This was with your ears. The innovation we’re trying with the eyes is how to intelligently navigate. I think there have been so many attempts that have not worked in the last year or something.

There we go. All right. It’s a really good segue, because here’s Google Glass.

Now, as ridiculous as I look right now—and I can hear the cameras taking my picture—you will not be spared, because you’ll be wearing them as well. I remember when Larry and Sergey started walking around with these.

In fact, Larry—I was at a party, and he came onto the dance floor with these. I said, “Larry, take those off. All the girls are going to stop dancing if you keep walking around with them.” He goes, “Really?” I was like, “Yeah, that’s not how dancing works.”

If you think about this product, why did they stop making this? They should have kept iterating.

And this was AR before AR. You see right through it.

Yeah. Ahead of its time, right? Go ahead and try it out, Bob. There you go. And now forever you will also be in infamy. There you go. Your turn. Smart to do it.

All right, I’ll do it. But by the way, the new ones aren’t much better. The form factor is better, but the utility isn’t there.

Now here’s one. This is a miniature version. I tried to get this, and if anybody can get me this, I’ll pay $10,000 for it—maybe $25,000. The Theranos one-drop blood machine. This was one of their tchotchkes.

But in truth, you’re now in healthcare. This may have been a fraud, allegedly. In reality, she’s in jail, I guess. I don’t want to—I mean, maybe there’s a chance she’s innocent. Who knows? I’ll leave that possibility out there. Allegedly.

But the promise of this captured people’s imagination: a small amount of blood to get a lot of data back. In fairness to Elizabeth, she was able to do a couple of interesting tests with a small amount. This was a great product idea, correct?

Yes. Yes.

Will somebody create that with AI in the next 10 years?

I think it’s very likely, because the challenge with this is: How can you actually manufacture those nanodevices where you can take really low volumes, be accurate, and measure these things? The technology wasn’t there.

Going back to our hardware manufacturing innovations, I think they will catch on to enable this. You want to be able to have real-time diagnostics. Think about a modern physical and being much more preemptive about healthcare. Pervasive, effective capabilities like this—these endpoints will be useful for that.

You have Function Health. You have Superpower now doing this. I don’t know if you guys use either of those products, but getting your blood work done every year and having a concierge talk to you about it for about $800, $1,000 a year, or $600 a year—obviously, consumer-led healthcare and the Theranos vision.

I think there’s a growing movement around longevity. It’s become a cultural phenomenon. First of all, consumers have a propensity to pay. We have a company called Ro, for example, that focuses on GLP-1s. Because there’s that propensity, it drives innovation to create more products like this that are focused on keeping you healthy.

How many people owned one of these? Raise your hand. All right. How many people have 3 of these in their closet that they can’t throw away?

The keyboard.

This was the greatest product ever.

I did a startup out of college. One of the very first non-email apps on this was a merchandising app for Red Bull. We wrote that so they could do inventory tracking in a store. This was an amazing product.

I’m still faster on this keyboard, right? I mean, for McKinsey, this was your cocaine.

We had some very senior people who wouldn’t give it up.

I have to tell you a story. It just gives me anxiety. I grew up writing apps on this, and then in 2011 I moved to the Valley and had my BlackBerry. I put it on a table like this. I met with somebody who was a well-known person in the Valley. We had a good conversation. At the end of it, he said, “You still use a BlackBerry?” I was like, “Yeah.” He’s like, “Stop doing that. You were judged in this meeting.” I kid you not. I was like, “Okay.”

I don’t want to touch it. You were holding it away from me.

Just think about how many carpal tunnel surgeries this created.

Oh, absolutely.

I mean, this was great for the economy.

This is an interesting one. How many people owned a PalmPilot? How many people owned one of these? Incredible, right? This one has a stylus and an antenna. We got this off eBay, thanks to my friends at CES.

You had to learn Graffiti, and you would be very good at spending 3 or 4 minutes at a party typing in something. You’d have to have your phone separately, right? These were 2 different devices.

If you really wanted to have a lot of swagger and a lot of rizz, you would have this on 1 side of your belt. I know you had this problem. You did have, didn’t you?

You have to be equal.

And the BlackBerry on the other side. That was like you were a gunslinger.

Yeah. And then in the early days, when the BlackBerry didn’t have the phone, you had the phone, too. So then you looked like a utility guy.

Hey, Hemant, I know that in college you lost a lot of brain cells to this one. The first ad on the internet was a banner ad for Zima.

Oh, boy.

How many people have had a Zima?

Oh, too many.

These are headaches. This was the most repulsive drink in the world.

We got an empty can of it. It's still available, I think, in Sweden. I think there's 1 place that still has the license and produces this horrific beverage.

But you look at all the carbonated stuff now. I think a version of this is White Claw.

Yeah, I think that's that generation's version.

You want one? No, thanks.

I actually ran a marathon with one of these on my waist in New York City. The Sony Discman didn't skip when you were running.

You see, this is a very good point. I had the advanced one that had a 10-second buffer.

Oh, this was elite at the time. It's an extra $50, but it would buffer 10 seconds.

And then, obviously, the iPod came out.

What do we think, in terms of the limited capabilities but the inspiration of this, we'll look back on at this moment in time? In other words, a device that could go 1,000x in its capability but provide similar functionality—in this case, being able to have portable music?

That's interesting because you think of the Walkman before this, which was the cassette. The one that skipped was durable.

An advance in technology, moving from analog to digital, but less durable.

Yeah, but better fidelity.

The transition to the iPod solved both equations. And it gets you thinking: What are the transition technologies we're in right now? One of the places I come back to is health wearables. So many different health wearables are out there, and they're all attacking the problem from slightly different angles.

Yes, some advances, but I think we're on the cusp. I go back to marrying this plus wearables to having more continuous monitoring and data. This might be the transition step between your Eight Sleep, your Oura, your WHOOP, all of that, your blood work coming together and giving you customized medicine.

I think that's a better answer than I was going to give, but my answer is LLM hallucinations. When you think about the intelligence, it's actually unreliable in a lot of ways, just like the music was unreliable with this. Is that going to change fundamentally over the next one?

This was a very interesting device because, for people who don't know, this one might have text messaging on it, but it used to just tell you the phone number of the person who called. So now, if you were dating and you were in the dating pool and you got that text from that special number, you're like, “Oh, how many minutes before I call back? I have to go find a pay phone and call back.”

You used to be able to send a number. So after you paged somebody, you could put in a couple-digit code. We started to have our own vernacular: 411 or 911. You could append some numbers to your beep, like your location or the street number you were on.

It was really an interesting product in how we never got to turn off work, which led to always-on doomscrolling—the never-ending nature of our commitment to work. In some ways, now we're starting to see a reverse of that. People are buying phones. I understand a lot of millennials now are buying digital cameras so they can leave their phone at home, and they're getting flip phones, so they've unbundled it. Really interesting. Any memories of the pager for you?

Yeah. First of all, they always say all the money is made in bundling and unbundling, and that is happening. And I think it is about how we go back to human connection and engaging in person, as opposed to trying to be fulfilled online and being lonely offline. That's probably the behavioral change that's going to happen. What enables that? I think there is probably some social engineering that's going to drive that.

This has been an amazing hour. Well done, gentlemen. Big round of applause for our guest.

Thank you so much for hosting us. This is incredible. Thank you. It's been a great audience.