AI gets physical as manufacturing looks good again
Reports last month that Amazon founder Jeff Bezos is attempting to raise US$100 billion to buy up manufacturing businesses are a sign that perhaps a bigger game is afoot.
As the value of software on its own diminishes, thanks to AI coding assistants, businesses that actually make things have been made more interesting.
I share a belief with certain forward-lookers that progress is pushing not just the value of software creation, but labour, logistics and energy towards zero (in the long run.)
Whether or not this makes me fun at parties is for somebody else to say, but I’m in both the AI optimist and the sustainable abundance camps.
Until just this year, enterprise-quality “vibe coding” would’ve been impossible. The resulting SaaSpocalypse has been terrible for vendors, but great for their (in some cases former) customers.
When an e-commerce business founder wrote that Claude Code allowed his team to build replacements for certain digital tools – saving maybe a quarter of a million dollars through nine days of AI-assisted coding – I cheered.
I remember 15 years or so ago when a manufacturer might’ve paid a million dollars a year for an ERP that was, quite frankly, not fit for purpose.
(Incidentally, we at ACM CRC, are developing our own ERP creation tools, where a company can input their needs and a pretty good ERP system is spat out.)
The abovementioned apocalypse has led certain folks to muse about whether automated coding means their money is safer in the real (read invested in the physical) world. The bits (software) vs atoms (tangible products) paradigm is shifting.
This shift was already underway pre-Claude Code, in a few ways.
The AI revolution was never purely bits versus atoms, if you consider the massive capital investments in data centres and everything in and around them.
Serious investments are being made in a growing number of physical AI companies, such as Bezos’s Project Prometheus (raising a reported $US 6.2 billion late-last year), Yann LeCun’s AMI Labs (which raised $US 1 billion last month.)
In these cases and others, it’s bits AND atoms.
Uber co-founder Travis Kalanick’s group of companies – chasing “physical automation for food, mining and transport” – was literally rebranded as Atoms last month. And readers will have followed the development of all-purpose (including factory work) Optimus robots by Tesla’s Elon Musk over the last five years.
Which brings us back to the big bucks and Bezos story of last month: his reported courting of sovereign wealth funds, JPMorgan and others to raise $US 100 billion for a vehicle that will buy up manufacturers and “seek to use AI technology to accelerate their path to automation."
The fact that Bezos hasn’t commented publicly on the “manufacturing transformation vehicle”, or on Prometheus makes both even more interesting.
Bezos is co-CEO of Prometheus, alongside Vik Bajaj (formerly of Google X.) It is Bezos’s first formal operational role since he left as Amazon CEO in 2021
The relationship between the startup and the buyout fund is not yet clear. The Financial Times reported in February that the fund “would invest in groups that Bezos and Bajaj anticipate will be hit by its technology.”
Prometheus’s early moves include acquiring agentic AI startup General Agents, it reportedly seeks to apply AI to aerospace, automotive and computer manufacturing, and it appears to have several employees specialising in computational fluid dynamics.
According to FT this week, the current $US 10 billion fundraising round is nearing completion, and will value the startup at $US 38 billion.
Other than that, the company has no products and no website, so exact plans are anybody’s guess.
One conclusion of the shift towards bits and atoms is that as some of the value evaporates from previously safe-looking software businesses – due to the ease of starting a new one or of customers creating bespoke enterprise tools – there’s also an opportunity.
As Joe Fath from venture capital firm Eclipse put it, “AI is a powerful complement” in physical industries, making things “faster, cheaper, less capital-intensive, and ultimately driv[ing] better returns. But it's not a full replacement in the way it will be across large parts of the service economy.”
My own guess is that, in the bigger picture, the masters of the universe see manufacturing as critical for creating and capturing value (as it always has been, even if appreciation for it waxes and wanes.)
I’m optimistic that as AI gets better and automates service and other roles, it will lift productivity, increasing everybody’s living standards. Along the way, progress will see the cost of inputs like energy, labour and transport head towards zero.
The world’s richest are in a race to dominate the important intersection of the digital and physical worlds, identifying that whoever can develop the best, lowest-cost tools will own a pretty darned important market.
So where does that leave Australia? I would argue that we double down on smart manufacturing, and that those who run manufacturing businesses do their absolute best to make use of the current and upcoming AI offerings.
Australia is uniquely positioned to dominate in this new Industrial era, with our abundance of materials, land for generating solar and other energy sources, and no shortage of smart people: all critical inputs for manufacturing! All while our traditional blockers to manufacturing such as cost of labour and logistics costs (our tyranny of distance) drastically reduce in cost.
Luke Preston is CEO of the ACM CRC. This article was first published at InnovationAus. The original version can be viewed here.
