For the past two years, artificial intelligence (AI) has mostly been about brute force. More data centres. More powerful chips. More electricity. Bigger budgets. That phase isn’t ending — but something important is changing.
Because at CES 2026, the world’s biggest tech companies revealed AI’s next chapter.
First: what is CES — and why does it matter?
CES stands for the Consumer Electronics Show.
It’s held every year in Las Vegas, and it’s where many of the world’s biggest technology companies unveil their next-generation products and strategies.
Think of it as a “preview” of what technology will look like in the next 12–24 months.
So when the biggest chipmakers arrive at CES and all push the same message, investors should pay attention.
And this year, the message was clear:
> AI is shifting from being expensive and centralised…
> to being cheaper, faster, and embedded everywhere.
AI is growing up: from raw power to real productivity
Until now, most AI systems have lived inside massive cloud data centres.
That’s where ChatGPT-style systems run:
* you ask something
* a server far away processes it
* you get an answer back
But that model has limitations:
* it’s expensive
* it uses huge amounts of energy
* it can be slow or unreliable in real-world environments
* it doesn’t work well for machines that need instant decisions
Now AI is moving into a new phase: real-time decision-making, in the real world.
That’s where three key concepts matter:
1) Inference (the part of AI that actually “does the job”)
AI has two phases:
* Training: teaching the AI by feeding it enormous amounts of data
* Inference: the AI using what it has learned to make decisions in real time
Most people experience inference every day without knowing it:
* facial recognition unlocking your phone
* a recommendation engine choosing what to show you
* a system spotting fraud on a credit card
* software identifying objects in a video feed
Inference is where AI becomes useful — and profitable.
And inference is about to get much cheaper.
2) Efficiency (doing more with less energy)
AI has been expensive not only because chips are costly…
…but because AI systems consume enormous amounts of power.
So the breakthrough isn’t just “more speed”.
It’s more AI output for less electricity, which drastically lowers the cost of using AI continuously.
3) Edge computing (AI leaving the cloud and moving into devices)
Instead of AI running only in data centres…
“Edge AI” means AI runs inside machines, such as:
* robots
* factory equipment
* cameras
* warehouse systems
* vehicles
* medical devices
* industrial sensors
In other words:
> AI is leaving the cloud… and moving into the physical world.
What the chipmakers revealed (in simple terms)
At CES, every major chipmaker delivered the same underlying message:
Nvidia: inference is about to become dramatically cheaper
Nvidia introduced its next-gen platform and claimed it can cut AI inference costs dramatically.
That matters because when AI becomes cheaper to run:
* more businesses can afford it
* more tasks can be automated
* adoption accelerates
And that’s where the next wave of growth comes from.
AMD and Intel: AI is moving into everything
AMD’s announcements targeted multiple markets — data centres, enterprise infrastructure, and AI-powered PCs — all pointing to the same theme:
AI will spread into everyday products and business systems.
Intel’s message was similar:
better efficiency, better on-device AI performance, and more capability built directly into chips.
So the combined story from CES was:
> AI is no longer just for giant tech companies.
> It’s becoming a standard feature of modern machines.
The biggest shift at CES 2026: Robots stopped being a “demo” and became a business
Because the most important part wasn’t a chip.
It was **Boston Dynamics**.
For years, Boston Dynamics robots have gone viral online — impressive movement, amazing balance, great demos.
But demos don’t make money.
What matters is whether robots can operate like real workers:
* reliable
* scalable
* low downtime
* affordable maintenance
* able to repeat tasks constantly
At CES 2026, Boston Dynamics basically signalled:
> We’re moving from entertainment… to deployment.
Atlas is now electric — and that changes everything
Earlier versions used hydraulics, which look powerful but have drawbacks:
* loud
* complex
* messy
* expensive to maintain
* harder to scale
Electric robots are easier to run like industrial equipment.
It can swap its own batteries — meaning less downtime
If a robot needs human help to recharge or reset, it becomes a headache.
Battery swapping in minutes means:
* higher uptime
* less labour support
* more predictable productivity
That’s the difference between a “cool robot” and a useful asset.
The real breakthrough: the robot’s “brain” got smarter
The game-changer isn’t the robot’s strength.
It’s its ability to understand tasks without being programmed for every tiny movement.
Boston Dynamics integrating advanced AI models means the robot can take higher-level instructions and work out the steps.
That’s where inference meets robotics:
> When AI becomes cheap and fast enough to run constantly,
> robots stop needing perfect pre-programming…
> and start adapting to real environments.
And that is what makes humanoid robots commercially viable.
Why Hyundai matters here
Hyundai isn’t just an observer — it’s deploying pilots inside real factories.
That’s important because it signals this isn’t “future tech”.
It’s early-stage adoption of a new industrial trend.
Even a small number of pilots today can become large-scale rollouts later — especially once ROI becomes obvious.
So what’s the actual investment takeaway for AI?
Here’s the gist in one sentence:
> AI is getting cheaper and more efficient — and that allows it to move from screens into machines, which triggers a productivity boom.
This is how big investment megatrends usually form:
1. Costs fall
2. Capability improves
3. Adoption spreads
4. Infrastructure gets built
5. Winners compound for years
The long-term winners won’t just be the most famous AI brands.
They’ll often be the companies supplying the “picks and shovels”:
* chips
* sensors
* automation systems
* robotics components
* industrial software
* power management
* factory infrastructure
Because when AI becomes physical, it doesn’t just improve convenience…
It improves output per worker, speed, efficiency, and margin.
And that’s where durable profits come from.
CES 2026 didn’t reveal “another AI gadget”.
It revealed something bigger:
AI is becoming a worker, not a feature.
And once that happens, the biggest money is usually made not in the hype phase…
…but in the long, steady rollout phase, when adoption spreads across the real economy.
Not a subscriber to Money Morning?
You can get free daily recommendations like these with Money Morning eletter. Just sign up here.
