By SavantCat
In my previous piece I wrote about Wonderful, the company with a 24-year-old CTO that reached RMB 33.5 billion in 20 months by refusing to show off models and instead sending people on site to help enterprises actually use AI.
After that piece went out, a friend who works in smart home left me a comment: “You say the software side is converging. What about the hardware side? Should you take a look at the IFA that just closed?”
So I went through the first-hand coverage of this year's Berlin IFA 2026 (the Zhidongxi piece from walking 50,000 steps, Zhongguancun Online's observations from the show, and the official preview), and the more I read the warmer the back of my neck felt — because hardware and software are converging on the same direction.
In this piece I will not talk about the flashy humanoid robots. I want to talk about the most honest sentence from this IFA, and about a signal most people probably missed.
1. At this IFA, AI itself converged first
Start with something counter-intuitive: this IFA had the largest humanoid robot presence in its history. Among more than 1,900 exhibitors from 49 countries and regions, Chinese manufacturers numbered over 1,000, nearly half; Xiaomi took a 3,300 square metre booth and showed more than 380 products.
But the official tone from every company was strikingly restrained.
The organiser's own theme was “The future is now”. Translated, that means: do not give me science fiction, give me things I can buy and actually use.
A few of the most typical convergences:
First, AI is no longer a badge stuck on the outside; it has sunk into the foundations of the product.
Samsung's fridge expanded ingredient recognition to more than 3,000 items, judges shelf life and suggests recipes — and the key point is that the whole thing runs on the device locally, not in the cloud. The official logic is blunt: lightweight models have matured, so local devices can run inference themselves, faster and more privacy-preserving. My smart home friend says this local approach has become increasingly common on European home appliances over the past two years, almost standard equipment.
Second, the shift from AI features to AI scenarios.
Vendors no longer boast about having an AI chip or a large model; they dig into concrete life scenarios: ingredient management, recognising grease on dishes, sensing sleep state. I noted down the exact line from Zhongguancun Online's observation — “AI no longer pursues showy features; it converts compute into a practical experience users can perceive.”
Third, even ecosystems are converging, in order to break fragmentation.
UGREEN built something called HomeAgent, a neutral local hub, not tied to its own brand, designed to connect smart devices from different vendors. They see it clearly: the biggest pain point in smart home is not too few features, it is a pile of devices each speaking its own language and refusing to connect.
2. Robots are not being hyped any more, which makes them more valuable
The robot zone at this IFA had 300 companies. But if you go in hoping for science fiction you may be disappointed — because nobody is claiming humanoid robots are about to enter your home.
Officially it was candid: humanoid robots are still at the technology validation stage, with a long cycle before large-scale household purchasing. What is genuinely being pushed are task-specific robots — cleaning, tidying and the like, going into one concrete scenario to do the work first.
To an outsider that sounds deflating; to me it is reassuring. Because it means even the most capital-hungry hardware sector has started talking about solving one real problem first.
One detail I found especially interesting: the show featured a class of humanoid developer platforms in the ten-thousand-yuan range, from RMB 8,888, with open interfaces and support for peripherals and secondary development. They are not positioned as a nanny you bring home, but as an experimental platform for developers.
To someone doing delivery work like me, the signal is one sentence: even the hardware threshold is dropping towards letting more people take part.
3. Software and hardware are converging on the same word
The Wonderful piece I wrote earlier is the software representative — the most valuable thing in AI is no longer the model, but the people who can install the model in an enterprise and make it work reliably.
This IFA is the hardware answer — the most valuable thing in AI is no longer showing off, but sinking into devices, digging into scenarios and solving real pain points.
Put the two lines together and you get the sentence I most want to say:
The first half of AI was a contest of who was cooler; the second half is a contest of who is more useful. And the money in useful is exactly the part ordinary people can reach.
Big companies can burn tens of billions making a fridge do local inference, but they cannot reach and do not want the work — should the small restaurant downstairs take AI orders? Who answers the enquiry that hardware shop missed at midnight? Who helps the small company that cannot even organise its own documents start using AI?
These unglamorous jobs: Wonderful proved with USD 5 billion that the software side is worth money, and this hardware fair proved through convergence that AI money is flowing from telling stories to solving real problems.
What it comes down to for you and me — three sentences
First, do not be led away by showmanship; watch for the real pain points inside scenarios. Whether software or hardware, money flows towards whatever can be delivered. You do not need to build a large model or a robot; finding one concrete scenario with rigid rules, high repetition and a boss willing to pay is enough to make a living.
Second, local, no big cloud is actually good news for small merchants. Lightweight models running locally, balancing privacy and cost, mean small businesses finally do not have to beg their way onto the cloud to use decent AI. That door is opening for exactly the kind of small shop you want to serve.
Third, do not wait for the opportunity to mature. Fridges recognise ingredients now, which shows AI delivery has reached the capillaries. The AI available today is already enough for you to solve one concrete problem for a small shop near you — use it first, get one working, then talk about scaling.
When the AI story is told to the end, only one word is left: use.
Geniuses proved in the first half how strong models can be; Wonderful and this IFA proved the same thing in the second half — whoever can put AI to work on real people stands where the money flows. Models and robots will keep getting cheaper, but helping real people actually use AI will always need people willing to sit on site with a client and grind through the details.
Is there a small business near you with very rigid rules and heavy manual effort where you think AI could help? Or are you yourself working on bringing AI into a concrete scenario? Talk about it in the comments — these unremarkable corners may be the next opening.
#AIDelivery #IFA2026 #OnDeviceAI #EnterpriseAI #AIMoney