In 1903, the Wright brothers changed the world forever, and today, some 123 years later, long-haul flying is taken for granted.

However, Aviation remains a high-risk investment, and airlines worldwide continue to experience high failure rates. Fortunately, there has been no shortage of cashed-up investors willing to invest in Aviation. Are there similar parallels with AI and where we are in the AI journey?

Constructing a 1-Gigawatt (GW) AI data centre is estimated to cost between $US30 billion and $ US80 billion. Alongside this, you have significant investments in infrastructure, power, and cooling. Electricity alone can cost over US$1.3 billion per annum.

The key players in AI are the megacompanies: Microsoft, OpenAI, Amazon, Alphabet, Oracle, and Meta in the US and Tencent, Deepseek and Alibaba in China. NVIDIA, now the world’s largest company by market capitalisation, provides the same GPU chips to all parties. So it looks like a pretty level playing field.

The problem is that much of the innovation by one can be mimicked by another. And in the world of sales, unless you can differentiate your offering, it all comes down to price. Given the enormous capital required to build these data centres, having the AI servers lying idle is not an option. Just like airlines, if the flight is taking off, the seats need to be sold at standby prices. Oracle and Meta will go cashflow negative in 2026.

All players are betting that demand will grow commensurately with construction – build it, and they will come. But what if consumer demand for AI doesn’t materialise?

We live in interesting times.

Sources: Bernstein Research, McKinsey, Warakiri Asset Management