For the past two years, tech companies like **Meta**, **Google**, and **Microsoft** have been on an AI binge, splurging on the latest AI models and computing power to stay ahead of the curve.
The Era of Subsidized Intelligence Ends
But a growing number of companies are now balking at the soaring bills that come with it. As the initial hype fades, businesses are starting to question the real value of their AI investments. The era of subsidized intelligence, where companies could easily justify the cost of AI without much scrutiny, is ending.
Just a few years ago, AI was the holy grail of tech. Companies like **Meta** and **Google** were willing to invest heavily in AI research and development, pouring billions of dollars into projects that promised to revolutionize everything from customer service to data analysis.
AI Agents and Computing Power Get More Expensive
But as AI models became more sophisticated and computing power grew more expensive, the costs began to add up. The average cost of training a large AI model has skyrocketed from around $100,000 a year ago to as much as $1 million today.
Meanwhile, the prices of computing power – the foundation upon which AI relies – have increased by as much as 50% in the past six months. This means that companies are facing a perfect storm of rising costs and decreasing returns on investment.
Companies Seek Smarter Spending Strategies
So what’s the solution? Companies are starting to look for smarter ways to spend their money on AI. They’re focusing on more practical applications of AI, like automating routine tasks or using AI to improve customer service.
They’re also exploring more cost-effective options for training AI models, like using cloud-based services or collaborating with other companies to share the costs. And they’re starting to question the value of AI in the first place – wondering if the benefits really justify the expense.
What this means: The AI binge is over, and companies are starting to get realistic about the costs and benefits of AI. As a result, we’re likely to see more practical and cost-effective applications of AI in the future – and fewer grand, unproven projects.
