Bubble won’t burst tomorrow, but prepare for the fallout if it does…
Warnings of an imminent AI bubble burst may be misplaced, but according to new research from Forrester, the AI industry faces a different challenge: Finding enough real-world value to justify the extraordinary levels of spending currently pouring into AI infrastructure.
The analyst firm’s modelling suggests AI must lift the level of US GDP by five percent by 2035 for today’s investment in AI software, cloud services and computing infrastructure to break even. Yet published estimates of AI-driven productivity gains average around two percent. That leaves a gap equivalent to roughly US$1.3 trillion in additional value that must come from benefits beyond productivity.
“Leaders shouldn’t pull back. Instead, they should keep investing where AI delivers value they can demonstrate… and be ready for corrections, large or small.”
The report, An AI Bubble Burst is Not Imminent, lands as tech giants continue an unprecedented spending spree on AI infrastructure. According to Forrester, Alphabet, Amazon, Meta, Microsoft and Oracle collectively devoted 31 percent of revenue to capital expenditure over the past four quarters, almost triple the 11 percent recorded in 2018 and 2019. Third quarter filings show Amazon spent $77.3 billion against $42.4 billion of cash, Alphabet spent $9.1 billion more than its operating cash, Meta outspent its operating cash by $1.2 billion and while Microsoft is still spending within its cash, it borrowed anyway. Estimates put spending by the five companies at US$806 billion in 2026, up 96 percent in a year, with spending expected to exceed US$1 trillion in 2027.
Forrester says that scale of investment resembles previous tech bubbles, but argues it is too early to conclude that an AI correction is imminent. Instead, the key question is whether enterprises can generate enough measurable value to absorb the capacity now being built.
To track that risk, the firm developed an AI Bubble Model that measures supply pressure, demand absorption and the financial health of companies funding the build-out. The latest reading shows supply pressure at an extremely high 4.95 out of five, while demand absorption sits at a much more modest 3.44. While infrastructure investment is accelerating rapidly, the largest suppliers remain financially strong and continue to generate significant cash flows.
But Forrester says the future of the AI market will depend less on how much infrastructure is built and more on whether enterprises can generate enough value from AI to justify it.
Brian Hopkins, Forrester vice president of emerging technology and principal analyst, says demand remains the biggest unknown because it depends on enterprises finding lasting value in AI beyond simple productivity gains.
Forrester points to the speed at which AI consumption is growing, versus the measurable economic returns. Google’s monthly token processing surged from around 9.7 trillion in May 2024 to more than 3.2 quadrillion in May 2026. Yet the firm’s modelling estimates AI has lifted US GDP productivity by less than half a percentage point so far, suggesting adoption and value creation are not necessarily moving at the same pace.
For Australian and New Zealand organisations, that gap may become increasingly important. Forrester warns if promised returns aren’t forthcoming and enthusiasm cools resulting in a correction, it will reshape AI portfolios and strategies.
“As capacity bought years in advance comes online, suppliers will have to cut prices, slow expansion, consolidate, or exit,” the report says. It notes that in July 2026, S&P cut Oracle’s credit rating to BBB-, ‘one notch above junk’, citing the cost of its AI build-out and a projected cash shortfall.
“Expect your vendors’ pricing, roadmaps, and partnerships to look very different after a correction.”
It won’t just be rapid shifts in pricing and licensing, but startup failures, vendor acquisitions and product discontinuations, the report says. “This will force you to rethink your foundational assumptions and investment priorities, and your architecture, platform and infrastructure decisions.”
The report suggests AI strategy can no longer be treated as purely a tech initiative, with supplier financial health becoming an increasingly important factor in platform and procurement decisions.
It recommends designing architectures that can withstand vendor failures, negotiating portability and exit provisions into contracts and avoiding long-term lock-ins at today’s pricing levels.
Forrester’s conclusion is not that the AI boom is about to end. Rather, it argues that the industry’s future will be determined by whether enterprises can convert growing AI usage into lasting business value.
For technology leaders, that means the biggest risk may not be an AI bubble at all. It may be investing heavily in AI without a clear path to proving its worth.
“Leaders shouldn’t pull back,” Hopkins says. “Instead, they should keep investing where AI delivers value they can demonstrate, watch the financial health of the companies building AI computing closely, and be ready for corrections, large or small.”