In 10 years, will humanoid robots be folding your laundry? This question—which might seem like science fiction—is becoming a real possibility to the founders and investors driving today’s robotics race. Recent reports that OpenAI is hiring robotics engineers has fueled new speculation about the future of high-tech tools in the workplace and our homes.
Andy Wu, Melwani family associate professor of business administration at Harvard Business School, studies the tech industry with a focus on strategic timing: when is the right time to adopt, or compete in, a new technology? In this Five Questions interview, Wu shares his thoughts on the market potential—and financial reality—of emerging consumer technologies including humanoid robots, virtual reality, and more.
This interview has been edited for length and clarity.
Humanoid robotics is one of the hottest investment areas right now. Is that hype warranted?
My honest view is that humanoid robots are suboptimal for nearly every conceivable use case, except possibly interacting with humans. Specialized non-humanoid robots are both less expensive and more effective. Consider locomotion: wheels and four legs are almost always better and cheaper than two legs.
The real question is whether humanoid robots become a dominant robotic format—so that the supply-side economics drive its cost below what specialized robots would cost.
A smartphone today is drastically cheaper than a barcode scanner, even though the smartphone is far more complex, because we manufacture over a billion smartphones a year and only a fraction of that in barcode scanners. Scale and learning wins.
The only way to hit that kind of scale for humanoid robots is for the technology to become general-purpose, with millions or tens of millions of units manufactured and deployed each year. Achieving this would require significant consumer adoption. Consumers may have different tastes than a factory manager about what the robot looks like and how they interact with it.
Even if both a humanoid and a non-humanoid robot could fold laundry equally well, the open question is whether consumers prefer the one that looks like them. If they do, the cost curve bends and humanoids could become viable in industrial settings too, almost as a byproduct. That’s the trillion-dollar bet.
A few years ago, many companies were betting heavily on the metaverse. What happened?
I still believe in the metaverse. The subtlety is in how you define it. Many industries are already operating in the metaverse—defined as 3D virtual worlds—today. The entire business of autonomous vehicles is built around 3D simulation of virtual worlds, which allows engineers to test driving algorithms.
A significant share of individuals already spend many hours a week in 3D virtual worlds like social media and in gaming platforms like Minecraft, Roblox, or Fortnite.
Do you think Meta Platforms still has a compelling long-term vision for the metaverse?
Meta bet on a very specific vision of the metaverse, centered on virtual reality (VR) headsets and augmented reality (AR) wearables. Most observers would say they were too early on VR.
That said, after Nvidia, Meta is one of the biggest winners from generative AI. I often use a “gold rush” analogy: OpenAI, Anthropic, and xAI are out there digging for gold. Nvidia is the consummate shovel seller, designing the chips needed by the diggers. And Meta is the consummate jewelry maker. Meta’s social media, advertising, wearables, and metaverse businesses all stand to benefit from advancements in generative AI, wherever it comes from and whenever it arrives.
So even though Meta’s original metaverse bet hasn’t paid off, generative AI strengthens the long-term thesis.
What current technology trend feels overhyped to you?
I'll give a counterintuitive answer: generative AI is both, depending on the time horizon.
In the short term, there are serious risks. AI may be a bubble—which I define as a mismatch between the vision for value creation and the current reality of value capture. Everyone can imagine how useful AI will be, but no one has figured out yet how to make money.
Do you think generative AI is on a sustainable course?
Generative AI today has high variable costs and insufficient revenue to cover it. Every time we ask a question to a chatbot, that costs real money; running an AI agent costs exponentially more so.
Most users get it for free, and power users and businesses mostly pay a flat subscription fee. Eventually, the most viable model will be something equivalent to pay-per-usage. But when we start paying for usage, generative AI becomes much less appealing.
Meanwhile, the companies pouring hundreds of billions into the infrastructure buildout face real financial obligations today that they may not be able to meet. OpenAI has promised $100 billion contracts to several vendors despite not generating anywhere near the revenue to pay them.
What separates a hype cycle that fades quietly from a truly destructive bubble is the amount of leverage and risk being taken by investors and vendors. By that measure, this one has the ingredients to do real damage if growth disappoints.
In the long term, what’s underestimated is all the opportunities AI will create, independent of whether and when we achieve superintelligence. The key opportunities are in complementary technologies like robotics, the metaverse, and wearables, and in redesigning legacy organizations around AI workflows.
