Google is spending money faster than it can print it, and investors are finally starting to ask uncomfortable questions. During its latest quarterly report, the company handed the market an unpleasant surprise: its capital expenditure forecast has skyrocketed to $205 billion. This is a sharp jump from previous expectations of $190 billion. The problem for the market isn't just the astronomical figures, but the fact that the new forecast's floor—$195 billion—now exceeds what was considered the absolute ceiling just a quarter ago. This numerical leapfrogging reveals Google's total inability to forecast its own spending in the current infrastructure race.
A failure in financial planning
From an investor's perspective, a lack of predictability at these stakes is a massive red flag. As Billy Leung, a technology strategist at Global X Management, noted, such financial maneuvers highlight the colossal strain that AI construction is placing on budgets. We are witnessing a classic economic deadlock: Google and its peers are forced to pour billions into data centers and chips while simultaneously engaging in a race to the bottom in the API market. To compete with aggressive Chinese players, these giants must keep model prices low, which looks like a surefire way to incinerate margins against the backdrop of rising infrastructure costs.
Essentially, Google has admitted it cannot adequately estimate the cost of its own banquet, and that is terrifying the market.
Consequently, a situation has emerged where tech giants pay increasingly more for the same amount of revenue—or sometimes even less. The success of Chinese startups, which are releasing competitive models despite GPU shortages, proves that the golden age of hardware manufacturers and unchecked budget bloating may end sooner than expected.
Signs of an overheated market
Nervousness has permeated the entire supply chain. Investors are spooked by Oracle’s debt load for data center construction, which many view as a health indicator for OpenAI. Even Nvidia has been forced into negotiations for deals worth three-quarters of a trillion dollars to sustain the pace of expansion. Specifically, Nvidia's decision to act as a guarantor for $250 billion of OpenAI’s debt suggests that real market demand may be far weaker than aggressive chip procurement implies. Even SpaceX shares have lost nearly half their value from peak levels, completing a general picture of industry-wide anxiety.
For the corporate sector, this is a warning signal: the era of subsidized AI is coming to an end. If the world’s wealthiest companies cannot forecast the cost of operating their infrastructure while facing deflationary pressure on their AI products, the "buy all the hardware in the world and figure it out later" business model is officially bankrupt. At some point, the cost of entering this race will simply outweigh the value of the prize itself.