Reality check for Blackbird: Australia is not an AI leader

“Punching above our weight” is a meaningless boast if we’re still knocked out of contention by bigger players.

As featured on the Australian Financial Review

Jul 3, 2025 – 10.16am

Blackbird Ventures’ Tom Humphrey’s piece (“Australia is already an AI leader. So why aren’t we talking about it?“) reflects a comforting narrative, but one we should treat with real caution. These are the stories we tell to feel better, not to get better.

Tom was right to point out that AI is coming, and if we don’t act, it will sail past us. The urgency is real. But calling Australia a leader risks overstating the current picture. We’re not leading, we’re participating. And there’s a real danger in confusing the two.

That distinction matters. National confidence is good until it becomes an excuse not to compete. “She’ll be right” makes us comfortable. Tall poppy syndrome keeps us cautious. But AI rewards speed, scale, and ambition.

We fall back on “we punch above our weight”. But scale isn’t an excuse; it’s a challenge. Our size should push us to compete harder, not rest on relative success.

I say this with real affection for Australia and a global perspective: we are not an AI leader. We are a small – maybe high-quality – contributor. But talent is not power.

Tom is a smart investor, and Blackbird has helped shape Australia’s tech ecosystem with some of our most important companies. Its track record speaks for itself. That’s exactly why the narratives we choose matter: they shape how we frame progress and where we place our bets.

OpenAI’s Economic Blueprint for Australia, released this week, reinforces that reality.

VCs are naturally incentivised to promote ecosystems; it’s part of their value creation. But even the strongest ecosystems need more than private capital. They need depth, resilience, and long-term foundations that extend beyond any single firm’s reach.

The challenge isn’t the quality of our investors, who in many ways have outperformed. It’s that even exceptional VCs can’t build what only governments can: sovereign infrastructure, scaled compute, and long-term industrial capability. This isn’t a criticism of private markets; it’s recognition that some foundations require public investment. Without that broader platform, we risk mistaking private sector momentum for national durability – and that’s exactly what happens when we celebrate individual wins as systemic strength.

Tom’s piece highlights genuine successes: a wunderkind teenager, successful brothers, academic partnerships. We’ve also developed a habit of talking about Canva and Atlassian like clockwork. But they remain among the few Australian tech companies with sustained global impact. Do we really want to be referencing the same two examples in 20 years? They’re exciting, but they’re exceptions, not the rule.

Early-stage funding rounds and promising researchers do not change the fact that Australia lacks sovereign infrastructure, defensible IP, meaningful compute capacity, or globally scaled platform plays. These are real limitations.

Some foundations are appearing: early models, academic wins, emerging partnerships. But we need more. Without scaled platforms, research standards, commercial depth, and sovereign compute, we risk becoming a test market for someone else’s AI. That isn’t leadership. That’s product testing.

OpenAI’s Economic Blueprint for Australia, released this week, reinforces that reality. From tax incentives and national skills programs to investment in compute infrastructure and secure access to public data, their message is clear: OpenAI sees opportunity but not readiness.

Meanwhile, AI has become a capital markets race. Microsoft’s partnership with OpenAI is worth more than $13 billion. Leading AI startups like Anthropic, Mistral, and Cursor have each raised close to, or more than $1 billion in individual rounds. For comparison, Cut Through cites Australia’s entire startup sector raising just under $1 billion in Q1. Individual companies are out-raising our entire national ecosystem.

Much of our talent is employed by multinationals or backed by offshore capital. According to the National Artificial Intelligence Centre, just 100 companies account for 58% of all AI job postings, while our ecosystem “increasingly depends on access to AI foundation models developed and operated at the global level”.

That suggests intellectual property isn’t staying here. Despite producing 1.88% of global AI research, Australia accounts for only 0.18% of global AI patents – nearly 23 publications per patent. We’re capturing salary value, not enterprise value. We’re at risk of becoming a support node in someone else’s supply chain.

We’re heavily reliant on foreign-born talent, which makes up 93% of our AI workforce. That may reflect our ability to attract talent, but attraction isn’t retention. While 59% of international AI PhD graduates stay in Australia, the other 41% leave. Nearly half of those we train are walking out the door.

This creates dangerous dependency: we rely on others to supply talent, then rely on them staying. That’s not a national advantage. It’s a pipeline leak with a sovereignty risk.

As AI competition intensifies and other countries sharpen incentives, we can’t assume Australia will remain a destination of choice. If our edge in attraction erodes and our retention stays flat, we’re in real trouble.

This isn’t about pessimism. It’s about clarity. This is far too important a race to pretend we’re ahead.