“Is the AI ‘bubble’ about
to burst or just starting? ”

The Jaw Dropping Speed of AI Growth  

Let’s talk about speed. The sheer growth in capability of AI training models is jaw-dropping, increasing at a rate of four times per year. To put that into perspective, that’s faster than the growth rate of mobile phone adoption during its peak years, even faster than the rate at which we sequenced the human genome. AI is scaling so quickly that it’s leaving other tech revolutions in the dust. Moore’s Law, which famously said the number of transistors on chips would double every two years, suddenly feels like a relic from a slower time.  

This isn’t just hype. Consider the latest wave of investments: OpenAI is raising billions of dollars to refine their models, while Nvidia and other hardware vendors are ramping up chip production at extraordinary rates. Cloud providers like Microsoft and Amazon are expanding their data centre capacity and even exploring novel power sources such as nuclear energy to meet AI’s insatiable demand.  

The investment is real, the infrastructure is real, and the commitment is real. It’s hard to argue with facts: investment in AI is staggering and shows no signs of slowing down.  

But Can AI Deliver ROI?  

Here’s where things get tricky. Massive investment doesn’t automatically mean massive returns. And this is where the sceptics start sharpening their knives. Can AI justify the insane amount of capital flowing into it? The answer depends on one key factor: ROI. At what point will investors stop throwing money at AI startups and organisations?  

Google’s latest earnings call highlighted this issue, its Q2 expenditure soared to $13 billion, largely directed towards data centres and services underpinning AI capabilities. Google framed this as a strategic, long-term investment, stating that it’s better to overinvest in AI now than risk falling behind. Similarly, OpenAI reported $540 million in losses during the development of ChatGPT, attributing the shortfall to the immense costs of training and developing its models.   

These companies aren’t chasing hype, they’re chasing ROI. And here’s the catch: you can’t treat AI as a shiny object. Instead of asking, ‘What cool thing can we do with AI?’, the question should be, ‘How can AI solve real problems, drive revenue, and scale operations?’ And the answers must be specific – specific problems, outcomes, and goals.  

It’s Not All Smooth Sailing  

Despite optimism, scaling AI isn’t easy, and some of the technical hurdles it needs to overcome are immense. Data centres supporting massive models consume astonishing amounts of energy. The projected power needed for AI by 2030 is staggering, with some estimates suggesting data centres will need access to 1 to 5 gigawatts of power per campus—enough to power a small city. Chip manufacturing is another bottleneck, as AI models require specialised chips that are in short supply. And while we have more data than ever, training next-gen AI models demands astronomical volumes.   

So, is AI in a bubble? If it is, it’s the most productive bubble we’ve seen. But the focus now needs to shift from building AI tools, to deploying them in ways that deliver real business value. For businesses and individuals alike, it’s time to take a pragmatic approach to AI investment.  

Just as the world’s leading tech companies are refining their offerings to focus on ROI, so too should organisations adopting AI solutions. The key is to prioritise providers who understand the true costs, complexities and change management involved and the means to deploy solutions quickly and effectively, with end-to-end outcomes in mind.  

AI isn’t just a technical deployment. To maximise ROI, focus on the business case: What problem are you solving? How will you measure success? And how will ongoing investment drive value? The companies and investors willing to answer these questions will reap the biggest rewards.  

Original article source: The Australian

Read More