China's Building Something Whilst We're Still Talking About It
China's 15th Five-Year Plan mentions artificial intelligence 13 times more frequently than integrated circuits. That's not rhetoric. It’s a reallocation of national resources at scale, towards a systems play. The target is 90 per cent AI integration across their economy by 2030. Most people hear that and think it's ambitious. We think it's a deadline with consequences attached. The difference matters.
Geographic Co-location Solves What Kills Many Rollouts
China has spent years building geographically concentrated expertise hubs. Manufacturers innovate with and against each other in the same physical space. Market adjacencies and supply chain inputs get co-located where feasible. Knowledge transfer happens within kilometres, not continents.
This isn't new. China has over 2,000 industrial clusters, especially concentrated in eastern and central provinces. Proximity allows efficient supply chain integration, reduces costs, and improves responsiveness. Concentration promotes knowledge exchange and innovation.
The execution advantage is structural. When you need to integrate AI into 90 per cent of an economy, coordination becomes the bottleneck. Co-location is likely to remove a lot of that friction before it forms.
Humanoid Robots Aren't Science Fiction Anymore
UBTECH has shipped hundreds of Walker S2 humanoids to active factories in what it calls the world's first mass delivery of humanlike robots. The robots can remove and replace their own power packs within minutes without human assistance.
All of those UBTECH Walkers will help reduce manual roles that kept humans on their feet all day. That means the next time you purchase a Volkswagen, BYD, or Audi, there's every chance a humanoid robot will have put it together.
Many view 2025 as the year humanoid robots entered mass production. Chinese firms made up a majority of the estimated 16,000 humanoid robots sold in 2025. By the end of 2025, China was home to over 140 humanoid robot manufacturers, which collectively launched more than 330 different models.
China's working-age population has been shrinking since 2012. By 2035, over 400 million people—roughly 30 per cent of the population—will be aged 60 or above. Each worker will have to support more retirees as the workforce itself contracts. The alternative is demographic collapse.
Necessity drives adoption faster than innovation ever will. When the choice is between experimental robotics and economic contraction, you deploy the robots and iterate in production. That's what China appears to be doing.
What Australian Manufacturers Should Actually Monitor
China will still represent key inputs for Australian manufacturers. And we're talking about ideas as well as ever-more-cost-effectively made components and products.
Energy, momentum, and innovation rubs off. We all need to keep thinking wow, that's amazing, can we do something a bit like that?
Mark Twain suggested there is no such thing as a new idea around 1900. We humans, at least the ones who make the effort to notice what's going on in specific areas, amalgamate those ideas. It's why we bring people with different expertise and even personality types together if we want to build high-performing teams.
Together, we create a kaleidoscope of what we have seen before, and then execute to create something new.
Watching is going to be really important. Doing is going to be more important.
The gap between observing what China's doing and translating that into action in your own operations comes down to a simple question: Where do you spend the most money? Or time? Or have quality issues? Or bottlenecks to scale? AI can work on all of those things. But don't choose all of those things.
Pick one workflow, and add value to it. Then pick another key workflow, and use the money you've saved to invest in that one too. And so on.
Australian operations of all types can get started in this way, learn the new tools as they go, then accelerate.
It's Continuous Improvement With a Different Toolbox
The requirement is not ‘AI strategy’ in our opinion. It's continuous improvement, something we've been doing for decades, and it’s totally not scary at all. The toolbox has changed though, and so must we.
With AI, and especially at this moment, the capabilities and outcomes are being developed simultaneously. The ability to purchase software that solves your specific problem rather than something generic has been democratised to a large degree. Smaller teams can make bespoke software faster and more affordably than ever before.
But if you wait any longer to evolve, then you're going to be last to market. This is our new reality. You can build an AI agent to deliver each system integration. You can build an agent to help you work out how to attack each integration. You can build an agent to help organise all of the humans slowing down the integration.
The challenge is real, but the infrastructure to address it already exists.
It's time to go faster.
Most of the AI tools we were testing and investing in 18 months ago at Arrow are no longer the ones we leverage in our own workflows today. They helped us get to a certain point, but we noticed even better ways, so we integrated those ideas, pivoted a bit, deployed the new way, and focused on the next workflow so we don't get left behind.
We're selling new things. Doing the old things differently. Adding more value for clients. Evolving constantly.
And when we don't let the pace of it all scare us, we’re having more fun as well.
China's Track Record Suggests This Isn't Theatre
If you've ever made a financial investment, there tends to be a product disclosure statement, a warning document that suggests past performance doesn't guarantee future performance. That's fair enough.
We all still look closely at past performance, though. Because the past is full of facts, and the future is full of guesses.
In the past, China has accelerated in manufacturing, urbanisation, automation, green energy, and numerous other areas after its 5-year plan declared them priorities.
For China, KPIs are Kept Promise Indicators, and they've got a track record of meeting or exceeding them.
Beijing is now measuring success by how deeply computing infrastructure penetrates the economy rather than chip production volumes. The 70 per cent semiconductor self-sufficiency target has been quietly deleted and replaced with a deployment metric of digital economy value-added at 12.5 per cent of GDP by 2030.
This represents a fundamental strategic pivot. China is building a name for itself as a provider of computing infrastructure to the developing world. DeepSeek demonstrated in early 2025 that frontier AI capability is achievable with constrained hardware. The focus has shifted from making chips to deploying capability.
The Guardrail Problem Changes Everything
The technology is going to be smarter than we are this time round, and that's different from any other period of technological advance.
Our best AI Engineers won't even be able to understand exactly how it works for much longer. And that means we can't just integrate something because it works.
We'll need to develop some key software for ourselves. We'll need to put the data where we can control it. We'll need to protect access better than ever before. And the best way to do it is likely to be different each time you look at a new project brief.
Basically, we need to be leaner and more innovative with every workflow we touch from here on in.
You don't want your special sauce to be packaged up and re-sold to other organisations, especially competitors. You don't want your customer data put at risk. And you don't want too many of your core workflows to be accessible by people or systems, including robots, outside of your sphere of control.
Competitors will try to copy or infer what you are doing if they think it will benefit them. Many AI tools have little regard for the original source of ideas they find on the internet.
Yes, bad things will happen where we don’t have guardrails, and this is a conversation you should be having with your technology partners.
Thankfully, the very technologies that put your business at risk also allow you to build your own AI, with guardrails, for a fraction of what it would have cost even a year ago.
China Has a Speed Advantage, But Australia Has Options
China has an ecosystem of suppliers, manufacturers, and technical vendors just the same as we do here in Australia. They are going faster because their government is giving it more oomph. More investment, more proactively pushing for progress.
China has the speed advantage. That's a fact.
But we have similar solutions available, whether we have the bandwidth to attack one workflow, ten workflows, or many processes. We’ll certainly fall further behind if we don't start choosing and doing now though. Many Australian manufacturers don’t have a software development specialisation. But any organisation can find a trusted local advisor and have them spin up a project team with just the right skills to address the risks, technical hurdles, and assurance of outcomes when creating something new.
For Australian operations, inertia is our biggest risk. So do something.
Here at Arrow, we made artificial intelligence and evolution a priority a couple of years back, set some KPIs around creating new capability, and held team members accountable. We do it as a small consultancy. We have customers that do it as medium-sized operations. And this article is noting that a whole country, China, has made it their thing.
It's not a size thing. It's a prioritisation thing.
When You'll Know You've Waited Too Long
The Aussie businesses that wait too long to evolve will see it reflected in the balance sheet, unfortunately.
In the space of just a couple of fiscal quarters, they'll realise their market has changed so much, so quickly, that they can no longer afford to react.
China's industrial robot output grew by 14 per cent year-over-year in 2024, reaching an estimated $33.4 billion in revenue. This growth rate doubled in 2025 to 28 per cent year-over-year.
China has recorded 7,705 humanoid patents over the past five years—five times as many as the US—and accounted for 54 per cent of global industrial robot installations.
Xiaomi's automobile plant can complete inspection of large die-cast components in two seconds using AI vision models—ten times faster and more than five times more precise than manual checks. That’s just one enhancement contributing to its ability to ship a Xiaomi SU7 sports car every 76 seconds!
China already enjoys a per-unit cost advantage of roughly a third compared to the West, but robot workers will slash labour costs and could double productivity simply because robots can work all night.
The competitive pressure is not coming. It's here.
What's Actually Being Misunderstood
China has proven over the course of decades that they do "we're all in this together" a lot better than many other countries. It sets them up well to leverage this great new technology revolution.
They are no less deserving of success than we are.
It's time for us in Australia to identify what's unique about what we deliver. Protect it. Do it better, faster, cheaper, and at greater scale than we have before.
And to keep our eyes open to absorb and integrate great new technologies and ideas wherever they come from.
The past is full of facts. The future is full of guesses. But the pattern is clear enough: organisations that treat AI integration as continuous improvement with a different toolbox will move faster than organisations that wait for a perfect capability they can install.
The window to react is closing in operational terms, not geopolitical ones. By the time it shows up in your balance sheet, the decisions that would have mattered have already been made.
Pick one workflow. Add value to it. Then pick another.
That's the move.
By Jeff Anderson, Founder & Lead Consultant, Arrow Strategic Communications
