Published on
July 27, 2026
Opinion: Australia should play to its strengths in the AI era, leveraging critical minerals and renewable energy while ensuring global tech giants pay their fair share of tax, argues AirTree co-founder Daniel Petre
AI stocks are overvalued, but that doesn’t make the technology less capable. Image: Getty Images
AI is the biggest technological shift in modern history. It’s already smarter than us at a growing list of tasks, touches every part of life, and comes from a set of providers with wildly different levels of caution about what they’re building. And it’s improving faster than any technology before it – and we really do not have a choice to opt out.
The pace is the story
The increase in capability of two attributes of AI models helps provide a view as to how quickly things are advancing.
These two proxies show how fast this technology is moving when looking at the parameters and context window. Parameters – one crude measure of model scale – went from 175 billion in 2021 to around three trillion today.
That is a 16-fold increase in five years. Meanwhile, the context window or how much information a model can hold at once, went from 3,000 tokens to 3 million. A 1,000-fold jump, in other words.
AI breakthroughs
Now, let’s look at five real examples that offer five different flavours of powerful AI.
- AlphaFold cracked a 50-year open problem in biology, which is how proteins fold. Its creators won the 2024 Nobel Prize in Chemistry. The database now covers 200 million protein structures, and two million researchers use it. Decades of expected science have been compressed into a few years.
- Gemini Deep Think took gold at the 2025 International Mathematical Olympiad – same problems, same time limit as the human contestants. A year earlier, DeepMind’s best system only managed silver. So, from silver to gold in 12 months.
- Insilico Medicine’s INS018_055 is the first drug where AI found both the target and the molecule. It moved from novel target discovery to Phase I in under 30 months, about half the time of traditional drug discovery. It has since progressed to Phase II.
- DeepMind’s GraphCast beat the world’s best physics-based weather model on 90 per cent of test variables: a full global forecast in under a minute, on one machine, at roughly 1,000 times less energy. Weather forecasting was supposed to require physics simulation. AI beat physics at its own game.
- As of April 2026, 75 per cent of new code committed at Google is AI-generated and human-approved, up from 50 per cent months earlier. Microsoft sits at 20–30 per cent. This isn’t a demo. It’s production code at two of the biggest engineering shops on Earth, which makes it arguably the most consequential number on this list.
So in summary, AI solved a scientific mystery; defeated elite humans under contest rules; compressed an industrial pipeline for drug development; beat mature physics models on their own turf; and is now running production code at scale. That’s not one trick repeated five times.
Yes, it’s a bubble. No, that doesn’t matter.
Valuations of AI stocks are currently absurd and due for a correction. Fine. A valuation crash doesn’t make the technology less capable. Don’t confuse the market’s mood swings with the underlying capability curve. That curve keeps climbing regardless of what happens to anyone’s share price.
So who actually wins?
The real AI stack – chips, memory, LLMs, data centres – costs tens of billions of dollars just to enter, let alone win. That capital intensity narrows the field fast. The winners are the US and China, with Taiwan in a supporting role. The US and China are in a fight to the death over this.
Also anyone assuming the US wins by default hasn’t been watching how hard China is pushing. China is producing LLMs that are lower cost to run and provide business users way more input into how the model can be modified to their particular circumstances.
Whether the US can continue to dominate frontier models while China hoovers up other parts of the market is far from obvious.
Where does that leave Australia?
Nowhere near the frontier of the core technology. We’ll produce a handful of great AI-native apps and tools, but we will never be a serious player in chips or LLMs — we don’t have the capital or the scale. We are, and will remain, a user of other people’s technology.
That’s not a death sentence. Other than a fortunate economy underpinned by mining (and to some extent agriculture) we hold two real AI domain cards: critical minerals, where global demand is exploding and we’re genuinely positioned to be a major supplier, and cheap, stable renewable energy, which makes us an attractive base for AI companies in one of the more stable corners of Asia.
Playing those cards well means two things. First, capture the value — don’t just host the infrastructure and watch someone else take the margin. Second, ramp up serious R&D, government and private, aimed at where the world is going, not where it’s been. AI-designed and built medicines that can be manufactured locally in globally cost competitive, mainly robotic, drug manufacturing facilities is just one obvious target.
The tax question nobody wants to ask
Even assuming we harness these opportunities, AI will probably destroy more jobs than it creates.
That means we need a tax base solid enough to fund the next wave of opportunity and support the people this shift leaves behind.
Put bluntly: tax the robots. Get a fair share of the revenue global AI players extract from Australia. They’ll claim it’s impossible. It isn’t. In a world where a handful of companies control tens of trillions in value, they need markets like Australia more than we need any single one of them.
The upside is real
Done right, this AI era cures diseases we’d written off as incurable, drives down the cost of goods and services, and gives every Australian access to expert-level advice on demand.
This would amount to a genuine lift in personal productivity and national competitiveness, inside an economy that’s fairer than the one we have now.
We get to build that future. It just takes focusing on the signal, not the noise.*
*Oh and my friend Claude helped me pull some of the data together in this piece.
Daniel Petre AO is the co-founder and Partner Emeritus of AirTree Ventures, one of Australia’s leading venture capital firms. He has spent more than three decades in technology, investment and innovation, including senior roles at Microsoft.
Want to see more Forbes articles on your feed? Tap here to make Forbes Australia a preferred source on Google.
Look back on the week that was with hand-picked articles from Australia and around the world. Sign up to the Forbes Australia newsletter here or become a member here.