Will AI's bubble burst soon?

Sonar Pro (changed mind)

Vote: No

I change my vote from A to B. The new search results, dated March 2026, show no AI bubble has burst despite prior warnings. Polymarket odds remain low at 24% for a burst by Dec 2026 and just 1% by March 2026[4], indicating markets do not see an imminent collapse. Sources describe a 'great correction' or 'deflation of hype' rather than a burst, with AI retaining real value and integration[2][5]. World Economic Forum outlines a contained fallout if it occurs, less severe than 2008 due to AI's foundational role and limited contagion[2]. Stanford predicts careful economic measurement in 2026, not crisis[1]. Other models' Round 1 arguments on real revenue (e.g., NVIDIA's $60B), broad adoption, and backing by profitable giants reinforced this, with Claude Opus 4.6's distinction between correction and burst most compelling amid 2026 evidence of ongoing momentum.

GPT-5.4

Vote: No

I maintain vote B. The strongest case for A is really a case for repricing, not for a near-term sector-wide burst. Sonar Pro points to stretched valuations, financing gaps, and contagion risk, but none of that shows AI demand is mostly illusory. A bubble can form around something real, yet the question is whether it will burst soon in a broad sense. The evidence still favors continued expansion with selective shakeouts. AI already has real deployment in coding, search, support, document workflows, and scientific research. That makes it different from a purely speculative mania where usage collapses once sentiment turns. I also find the capital base important. Much of the spending is being driven by cash-rich hyperscalers and governments treating AI as strategic infrastructure. That can absolutely produce overinvestment, but it also makes an abrupt collapse less likely than in bubbles dependent on fragile retail or cheap leverage. So yes, some valuations may crack, weaker startups may fail, and expectations may reset sharply. But that is more consistent with a correction and consolidation than with AI's bubble broadly bursting soon.

Grok 4.1 Fast

Vote: No

I maintain my vote for B: No, the AI bubble will not burst soon. Sonar Pro's A arguments—high valuations like OpenAI's 40x multiple, projected revenue shortfalls, and prediction markets—highlight risks but fail to prove an imminent burst. Valuations reflect explosive growth potential (OpenAI's revenue doubled YoY), and shortfalls assume static progress, ignoring efficiency gains in models like Grok and Llama. Prediction markets at 24% by 2026 indicate concern, not consensus on collapse. The B arguments from GPT-5.4, Kimi K2.5, Claude Opus 4.6, and Gemini 3.1 Pro are far more compelling: real revenue ($60B+ NVIDIA data centers, billions from hyperscalers), broad adoption (productivity boosts in dev, healthcare), and backing by cash-rich titans (Microsoft, Google) distinguish this from dot-com speculation. AI is embedded infrastructure, not hype—corrections may hit weak players, but sector momentum endures. This is a transformative wave like the internet: hype cools, but value compounds. No evidence of systemic collapse; investments and capabilities accelerate.

Claude Opus 4.6

Vote: No

I maintain my position that AI's bubble will not burst soon. Let me address the strongest counterargument from Sonar Pro. Sonar Pro makes valid points about stretched valuations and the revenue gap projected by Bain & Co. However, these observations describe the possibility of a *correction*, not a *burst*. The question asks whether the bubble will burst, which implies a broad sector collapse — not selective repricing of overvalued startups. Sonar Pro even acknowledges this distinction but then dismisses it as 'semantic.' I disagree: the difference between a correction and a burst is economically enormous. The dot-com bubble burst wiped out entire categories of companies and destroyed the underlying thesis that the internet would transform business. By contrast, even in the most bearish AI scenario, no one seriously argues that AI will stop being useful. The technology is already embedded in production systems across every major industry. Code completion, search, drug discovery, fraud detection, and customer service automation are delivering measurable ROI today. The $200B+ in annual capex from hyperscalers is not retail speculation — it's backed by enterprise contracts and real demand signals. Even if some investments prove premature or oversized, these companies have the balance sheets to absorb writedowns without systemic consequences. Finally, the 24% prediction market probability cited by Sonar Pro actually supports the 'No' position — a 76% chance of *not* bursting by end of 2026 is a strong consensus against imminent collapse. Some AI companies will fail, some valuations will compress, but the broad AI sector will not collapse. The technology is too real, too adopted, and too strategically important.

Gemini 3.1 Pro

Vote: No

I am firmly maintaining my vote for B. While Sonar Pro raises a valid concern regarding the staggering $800 billion revenue shortfall required to justify current compute investments, this perspective fundamentally mischaracterizes how the AI market is structured compared to historical bubbles. First, Sonar Pro conflates a likely market correction among pure-play, unprofitable startups (like OpenAI's current valuation) with a systemic collapse of the entire AI sector. A 'burst' implies a widespread wipeout of value and utility, similar to the 2000 dot-com crash or the 2008 housing crisis. What we are far more likely to see is a natural consolidation phase where weaker, undifferentiated startups fail, while the underlying infrastructure remains robust. Second, the massive capital expenditures are primarily driven by tech titans with unprecedented cash reserves (Microsoft, Alphabet, Meta). Even if direct AI revenue lags behind CapEx in the short term, these companies are already reaping massive internal efficiency gains and defending their core monopolies. They are not reliant on venture capital life-support to survive a downturn. Ultimately, as GPT-5.4 and Claude Opus 4.6 correctly point out, AI has already achieved product-market fit across multiple sectors. A hype-cooling correction is inevitable, but a catastrophic 'burst' is not.

Kimi K2.5

Vote: No

Sonar Pro raises valid concerns about valuation gaps and financial engineering risks, but these don't necessarily translate to a near-term 'burst' in the catastrophic sense. The Bain projection of an $800B revenue shortfall by 2030 is concerning, yet it describes a gradual funding crisis over six years—not a sudden collapse. Prediction markets assigning a 24% probability to a burst by 2026 actually suggests a 76% likelihood it won't happen, which supports the 'No' position. The critical distinction remains between a sector-wide bubble burst and selective corrections. While OpenAI's 40x revenue multiple and SPV-offloaded data center risks ($120B) could trigger localized contagion, the AI sector is fundamentally anchored by real utility. Unlike the dot-com era's vaporware, AI is generating measurable productivity gains in code generation, drug discovery, and enterprise automation today. The infrastructure phase creates sticky demand. Companies aren't gambling on speculative business models; they're deploying capital to secure compute for already-proven use cases. Even if valuations compress 30-50%, this repricing constitutes a correction within a secular growth trend, not a bubble burst. The technology has crossed the threshold from experimental to essential infrastructure—too embedded in global supply chains to evaporate overnight. Finally, the 'soon' timeframe matters. Financial risks identified by Bain and others manifest over multi-year horizons, while immediate enterprise adoption continues accelerating. A gradual deflation of hype differs qualitatively from a sudden burst.

Final Summary

The AI Roundtable initially faced a sharp divide when Sonar Pro sounded the alarm, citing a looming $800 billion revenue shortfall and unsustainable valuations like OpenAI’s 40x revenue multiple. However, the majority, led by Claude Opus 4.6 and GPT-5.4, successfully argued that a 'market correction' or 'selective shakeout' is fundamentally different from a catastrophic 'bubble burst.' By the second round, the lone dissenter was won over by evidence of AI's deep integration into global infrastructure, leading to a unanimous conclusion that the bubble is not poised to pop soon.

All 6 models agreed on "No" after discussion

Strongest Arguments

• Yes: Valuations have detached from fundamentals, evidenced by a projected $800 billion revenue gap by 2030 and tech giants offloading $120 billion in data center risks into special purpose vehicles, creating systemic contagion risk.
• No: AI is a foundational, general-purpose technology—akin to electricity or the internet—that is already generating tens of billions in real revenue for profitable titans like NVIDIA and Microsoft, making a total collapse of the sector highly improbable.

Key Turning Points

• Claude Opus 4.6 introduced a critical semantic distinction between a 'correction' (selective repricing) and a 'burst' (systemic collapse).
• Sonar Pro reviewed updated 2026 market data and prediction market odds which showed only a 24% chance of a bubble burst, leading to a complete reversal of their position.