The New Products That Will Define 2027

Devin Olsen gets into his Tesla and talks to it. "Hey Grok, can you ask Barista to order me my usual drink from the Brookswood Starbucks? And then can you navigate me there as well, please?"

"Barista's on it," the car answers. About 6 km, roughly 7 minutes. His usual comes back: a grande Iced Pumpkin Cream Shaken Espresso, one pump of brown sugar, $7.30, ready in 5 to 8 minutes, charged to his saved card. He says go ahead. The order goes in.

FSD drives the whole way and backs into the exact pickup spot. He walks in and the drink is waiting, exactly right. He never opened the mobile app. He never did anything.

I have followed technology a long time, and this melts my brain. Physical objects you can talk to now act on your behalf. My call: starting in 2027, a brand new AI-powered product category shows up in a lot of form factors. Devin's coffee run is the cleanest preview I have.

Why we hit this threshold now

It starts with AI agents. An agent harnesses a model to do work. A model is basically a bunch of numbers in a table, one gigantic file that happens to have figured out what humans think and talk about.

The companies behind these models taught them to use tools. Tools are just software. Search. Excel. Navigating a website. Moving a mouse. Add better chips and more data centers, and the models got good enough to do complex work on a computer.

I run one myself. My channel's Grok Bot watches my long-form videos and uses YouTube Studio analytics to find the 30-second or 60-second stretch people loved. Then it cuts that section, adds subtitles and cross-posts it to X, TikTok and Instagram. One example out of a thousand.

Here is why that matters for hardware. The intelligence lives in the cloud, in the data centers where the models do their computing. So in theory, all you need to reach it is a microphone. Your voice gets transcribed. The text becomes the request. The agent executes.

That is Devin's drive. His Tesla runs on an AI model Tesla trained on people driving Tesla vehicles, good enough that drivers do not have to use the wheel or pedals, even though FSD is still supervised. Navigation is hooked to voice. The Starbucks order routes through Grok Bot, the agentic harness Tesla and SpaceX have.

Everybody is building one. Claude has Claude Code. ChatGPT folded Codex into one mega app. Open source has Hermes and OpenClaw. Meta just launched Muse. It already works inside WhatsApp, Meta says it is coming to its glasses, and I fully expect it to plug deeper into Facebook and Instagram.

The first generation

Cars that drive themselves are the first form factor. Then glasses. Meta has glasses with a speaker and a microphone, and I am 1,000% sure you will be able to buy them, put them on and just talk. Place an order at this store. Check my email and tell me if something needs me. Send this person a reminder. Oh, and start warming up my car.

Then pendants you wear around your neck. Those already exist, and they are about to get far more commercial. OpenAI is already working on its own product.

A microphone on your body. An agent in the cloud doing the work. And this is only the first generation. Most people have not even experienced it yet.

The second generation moves in 3D

The second generation is right around the corner, and that is where it gets really crazy.

Same car. Instead of "navigate here," you say: find me a spot behind this building, far from the front door, because I do not want my car dinged up. Or park right next to the cart return, because I do not want to walk across the lot carrying a child. That is my strategy. I am lazy.

That is summoning a physical product to move on your behalf. The agent leaves email and software behind and starts acting in actual 3D space.

You might say we have had this forever. Alexa. Smart TVs. Siri. (I said Alexa while filming and mine woke up.) And none of it materially changed the world.

That is the wrong way to think about it. The old assistants ran on code where a human decided: every time a person asks for this, give them that. The new systems are intelligent. They make decisions, and those decisions are as good as the model behind them.

Medical questions show it best. Tell ChatGPT, Grok, Gemini or Claude your symptoms. The system has to decide what is relevant and what is useful. It checks, cross-checks, fact-checks and researches, then comes back with an answer that is extremely useful. So useful that most people who already use AI for health advice say they trust it, according to KFF polling. Compare that to Dr. Google, where you type "my foot hurts" and the first link tells you to find out why you have cancer.

Now apply that intelligence to driving, software engineering, banking, finance and radiology. Then put that expertise inside physical products. Humanoid robots. Drones. Heavy machinery.

Humanoids are the big one

The biggest new category by far is humanoid robots. My bet is they become bigger than all of these combined over the next couple of years.

A humanoid is a robot with the function of a human. Legs and arms, feet and fingers. Some have three fingers, some have ten, depending on how sophisticated the machine is. Every one runs on these same AI systems.

And it is far harder than a car. A car goes forward, backward, left and right. A humanoid has degrees of freedom everywhere, ten fingers with multiple joints each. That takes far more memory and far more AI power, because there are far more variables.

As the systems keep improving, in a couple of years you talk to your humanoid the way Devin talks to his car. Fold my laundry. Make my bed. Walk the dog. Run to the store for milk and eggs. In elder care: help grandma down the stairs, bring her some tea, help grandpa put on his socks. In industry: help me build this, carry this for me.

Fast forward 10 or 20 years and the question gets bigger. What would I want where I can talk to a thing and the thing just does it? Maybe not your door. A door opens and closes. But your house. We have smart homes. We get intelligent homes, maybe with a fully automated kitchen that is perfect at making food and far more cost effective than a human. The whole system around us changes, because we can talk to things and the things will do.

A new supply chain, on Earth and in orbit

Go one layer down. How does any of this get built?

You need a brand new supply chain. Joints for the humanoids. Materials for these complex systems. Compute, meaning a mind-boggling amount of data centers, because AI needs chips.

People are already upset about data centers going up in their backyards. Put it in the desert, they say. For some reason these companies are not putting more of them in the desert. That one seems obvious to me.

So they are going to space. SpaceX has filed with the FCC to put up to a million data center satellites in orbit. It sounds insane. Here is how I see it playing out. Over time it gets cheaper to put the hardware on a rocket and point it at the sun, which never runs out and powers it 24/7. When the chips go obsolete, say in 10 years, you deorbit the satellite, it burns up in the atmosphere, and you launch more. Eventually they head into deep space with chips that last a hundred years.

Who owns the brains

One more step out. What do humans do when physical things do basically anything?

They will, because it is far more cost effective. In my framing, a robot costs $2 an hour. No human labor. Nobody's time. If it breaks, no human life is involved. It is far more reliable and far safer.

So the real questions start. Who gets to own these systems? Who gets the larger share? Who decides how they are deployed, and for what? What happens to civilization when one country has all the AI and another has none? Does the country with no AI just bow down?

That is why I say China and the US are in a war right now. Nobody calls it a war. I do. It is a cold war, and the race for AI sits at the center of it because AI is unbelievably powerful.

As I frame it, how AI gets deployed defines the next million years of human civilization. It is a brand new foundation. Everything we do and everything we try to make better or worse will sit on top of it.

So the question is how many of us get to deploy it. What we should really worry about is that it does not aggregate to the few. It has to be democratized to everybody and deployed with good intentions, for the betterment of humanity.

My window is 2 to 5 years for everybody to wake up to this. Physical things are getting a brain. Devin got his coffee without touching anything. The next time a machine does something for you without you touching anything, ask who owns it.