The International Energy Agency surveyed organisations trying to deploy AI at scale. They found something interesting: The biggest barrier to AI adoption isn't cost. It's not technology. It's people.
But that's only half the story.
While organisations are struggling with skills, something else is happening behind the scenes. Tech companies are investing $400 billion in data centre infrastructure in 2026—a 75 percent jump from last year. Australia is now a Tier 1 global data centre market. NSW approved A$51.9 billion in projects. This infrastructure is being built at venture-capital speed.
The problem: electricity grids, supply chains, and regulatory systems move at infrastructure speed. Not venture speed. That mismatch is reshaping your technology costs right now.
What This Means for Your Business
Your cloud bill is rising. Not because you're using more. Because the underlying system has to upgrade to handle the load.
Think of it like a suburb suddenly getting popular. Everyone moves there at once. The council needs new roads, new power lines, and new water mains. That infrastructure bill gets factored into what residents pay.
Australia's in that moment. Data centres submitted over 10 gigawatts in grid connection requests—more than half of NSW's peak demand. Energy providers have warned that without clear policy, those grid upgrade costs could be passed to existing customers.
But the real cost driver is something most people miss: memory chip allocation. Data centres went from taking 32 percent of global DRAM supply five years ago to 50 percent last year. By 2030, that number hits 60 percent. Memory manufacturers are choosing data centre builders over everyone else. Prices are up 40-100 percent. Some products are up 1,000 percent.
Your cloud provider is paying significantly more for infrastructure. That cost eventually flows to you.
Why This Matters
Here's the paradox nobody talks about: AI is becoming wildly more efficient per task. A simple text query uses less power than leaving a TV on. Impressive progress. But organisations are still building massive new data centres.
Why? Because the work being asked of AI is exponentially more demanding. Video generation. Reasoning tasks. Multi-step workflows. These consume thousands of times more energy than simple queries. So even as efficiency improves at the task level, complexity and volume are growing faster. Infrastructure needs keep exploding.
This means your cloud costs aren't just exposed to normal inflation. They're exposed to structural supply chain constraint and policy uncertainty. For the next 18 months, that constraint is real.
How Leaders Should Actually Respond
First: understand what you're dependent on. Which systems run on cloud? Which vendors? When do contracts renew? If you're on month-to-month billing, you'll feel price increases immediately. If you're on fixed contracts, you're protected until renewal. Know which situation you're in.
This isn't technical. It's financial planning.
Second: acknowledge that skills and infrastructure are dual challenges. Yes, organisations are struggling to find people who know how to implement AI properly. That's the adoption barrier. But the infrastructure side is where costs surprise you. Both are true. Both need attention. The organisations winning aren't the ones who moved fastest. They're the ones who moved thoughtfully—building capability internally or finding smart partners who could bridge the gap.
Third: model cost scenarios. Build three forecasts for cloud spend over 18 months: normal inflation (5-10%), constraint scenario (15-25%), shock scenario (30%+). Run your business through each. The constraint case isn't theoretical. Memory prices are already up 40-100 percent.
Fourth: understand your vendor spread. If all computing runs on one provider, you have supply chain risk. Australia being Tier 1 now means local options exist. That optionality has value.
What Actually Matters
Policy is still being written on how grid upgrade costs get allocated. Supply chains are under real stress. Skills shortage is real. Infrastructure costs are rising.
The businesses that understand this landscape aren't making faster decisions. They're making more informed ones. And right now, that difference is the only thing that matters.
