Is the AI Bubble About to Burst? (singularityhub.com)

The AI investment boom that pushed valuations into the trillions is showing signs of strain as markets wobble and investors question whether the massive costs of training and operating large models will translate into sustained, high-margin revenue. Recent market drops and nervousness around tech and crypto coincide with growing skepticism about business models: companies have been priced on assumptions of near-unlimited AI adoption, yet building the stack—chips, cloud infrastructure, data centers and models—is extremely capital- and energy-intensive. Triggers for a sharp revaluation are now likely to come from within the sector: missed earnings at marquee suppliers (e.g., Nvidia or Intel), a chip supply-demand mismatch that drives prices down, or a slowdown in returns from scaling ever-larger models. If a correction arrives, the most immediate impact would be on chipmakers and hyperscalers that have driven the boom; 2025 saw roughly $350 billion poured into infrastructure by Microsoft, Amazon, Meta and Alphabet, and Goldman Sachs projects up to $4 trillion in AI infrastructure spending by 2030—numbers that could be scaled back. That wouldn’t end AI but would shift the industry from speculative “build now, profit later” bets to pragmatic deployments that demonstrably cut costs or boost productivity, ushering a tougher, more mature phase after a painful revaluation for investors, suppliers and governments.