Is AI the Next Big Bubble? Warning Signs Investors Shouldn't Ignore

1. Defining the AI Bubble Thesis: What Are We Actually Arguing About?

An economic bubble occurs when asset prices become detached from fundamental value, sustained by collective expectation rather than concrete evidence of returns. The question "is AI the next big bubble?" is being asked now because of three converging signals.

First, the sheer scale of valuation growth. Nvidia had a market value of $1.2 trillion at the end of 2023 and $3.28 trillion at the end of 2024, and as of July 2026, Nvidia's market capitalisation stands at approximately $5.02 trillion. Second, the speed of capital deployment: Microsoft (~$190B), Amazon (~$200B), Alphabet/Google ($175–185B), and Meta ($115–135B) plan to spend roughly $725 billion combined on capital expenditure in 2026. Third, AI startups with minimal revenue are being valued at hundreds of billions. OpenAI is currently valued at $852 billion following a $122 billion funding round, despite operating at a deep loss.

Not all layers of the AI investment landscape carry equal risk. Chip and hardware makers like Nvidia are reporting real revenue growth. Platform and model providers (OpenAI, Anthropic) are burning through capital faster than they earn it. Application-layer startups vary enormously. And public companies pivoting to AI often see stock price boosts simply for mentioning the technology in earnings calls.

The Gartner Hype Cycle offers a useful framing device here: technologies typically pass through a "peak of inflated expectations" before descending into a "trough of disillusionment," where only the genuinely useful applications survive. The question is not whether AI is real. It is whether the financial enthusiasm has outpaced reality.

2. Lessons From the Dot-Com Bubble: History Rhymes, But Does It Repeat?

The Dot-Com Parallels That Should Concern Investors

The similarities between the AI boom and the dot-com era of 1995–2000 are hard to ignore. Between 1995 and its peak in March 2000, investments in the Nasdaq Composite stock market index rose by 600%, only to fall 78% from its peak by October 2002, giving up all its gains during the bubble.

Then, as now, explosive stock price appreciation was driven more by narrative than by earnings. Retail and institutional investors were piling in out of fear of missing out. Companies with no path to profitability were achieving multi-billion-dollar valuations. The widespread belief that "this time is different" justified ignoring traditional metrics. And capital flooded infrastructure ahead of proven demand: fibre optic cables then, data centres and GPUs now.

AI stocks now account for roughly 44% of the S&P 500 market capitalisation, creating concentration risk reminiscent of what we saw in 2000 when tech stocks dominated indices at their peak. The AI stock market bubble risk, if it materialises, could be amplified by this concentration.

The Critical Differences That Separate AI From Dot-Com

The dot-com comparison, while instructive, is imprecise. AI is built on mature, proven internet and cloud infrastructure, whereas dot-com companies had to build everything from scratch. The leading investors this time are highly profitable incumbents, not speculative startups. Goldman Sachs Research notes that the appreciation of the technology sector has, so far, been driven by fundamental growth rather than irrational speculation, and the leading companies have unusually strong balance sheets.

AI is already embedded in real products used by hundreds of millions of people. ChatGPT crossed 100 million monthly active users in just two months after launch, making it the fastest-growing consumer application in history. By June 2026, the ChatGPT app had reached 1 billion monthly active users. Revenue growth at the infrastructure layer is real and substantial, not imaginary.

The Honest Verdict on the Comparison

The key takeaway from the dot-com comparison is sobering: even if AI is not a carbon copy of dot-com, bubbles can absolutely form within a genuinely transformative technology. The internet survived the dot-com crash, and it went on to reshape every industry on earth. But most dot-com investors still lost money. The Nasdaq didn't fully reclaim its March 2000 peak until May 2013, more than 12 years after the bubble burst. Technology succeeding and investors profiting are two very different things.

3. AI Valuation Concerns: When Stock Prices Lose Touch With Reality

Inflated Multiples and Hype-Driven Pricing

Goldman Sachs Research found that current valuations are high but still some way off the dot-com bubble peaks. The median 24-month forward P/E ratio across the "Magnificent 7" is 27x, roughly half the equivalent valuation of the biggest 7 companies in the late 1990s. That said, Nvidia trades at 40 times expected forward earnings, while most big tech stocks trade at around 30 times forward earnings.

The "picks and shovels" trade in GPU manufacturers and cloud providers is being priced for perfection. Meanwhile, concentration risk has reached extreme levels, with a handful of AI-adjacent stocks accounting for a disproportionate share of S&P 500 gains. In private markets, generative AI speculation has produced funding rounds implying astronomical valuations for pre-revenue companies.

The Market Narrative vs. Fundamental Analysis

Analyst upgrades and media coverage amplify price momentum in a self-reinforcing cycle. Passive index funds mechanically inflate AI stock prices as these companies grow in weight. Goldman Sachs' head of global equity research, James Covello, has been one of Wall Street's most prominent AI sceptics since he co-authored the original "Too Much Spend, Too Little Benefit?" report in June 2024. His central question remains whether the enormous capex being deployed can ever generate sufficient returns.

Goldman Sachs flagged several concerning developments: a significant rise in the valuations of many AI-exposed companies, continued massive investments in the AI buildout, and the increasing circularity of the AI ecosystem.

4. The AI Profitability Problem: Is Anyone Actually Making Money?

The Revenue vs. Cost Imbalance

The enormous cost of training and running frontier AI models is the elephant in the room. OpenAI generated $3.7 billion in revenue in 2024 before jumping to $13.07 billion in 2025. That sounds impressive. But operating losses widened from $8.78 billion in 2024 to $20.92 billion in 2025, despite the company's surging revenue. In 2024, OpenAI spent $2.37 to generate every $1 in revenue. In 2025, that ratio declined to $1.60 in expenses for every dollar it took in, but profits remain elusive.

The "AI ROI gap" extends beyond model providers. Morgan Stanley found that only 21% of S&P 500 companies could cite a measurable AI benefit at all. Forrester research reveals only 15% of AI decision-makers reported a positive impact on profitability in the past 12 months. PwC's 2026 CEO Survey found only 12% of CEOs have hit both revenue gain and cost reduction from AI.

Why Monetisation Is Harder Than It Looks

Commoditisation pressure is intensifying. As models improve and competition grows, pricing power erodes. A potential price war is emerging among generative AI companies, with OpenAI considering lowering the cost of its model access in response to expected moves by Anthropic. The race to the bottom in AI API pricing is compressing margins across the industry.

Consumer willingness to pay remains a challenge. ChatGPT now has more than 900 million weekly users, though only about 50 million subscribe to paid tiers. On the enterprise side, only 29% of organisations see significant ROI from generative AI. The profitability problem is compounded by a structural oddity in how AI demand is currently being created.

Watching the AI story unfold but want your money working in the meantime? Bleap's savings vaults offer 3.65% AER (Steady) or 3.83% AER (Dynamic) in USD, with just a $1 minimum deposit and 0% withdrawal fees. No speculation required. Open a Bleap account →

5. The Circular Investment Problem: AI Companies Buying From Each Other

One of the most underappreciated risks in the AI investment landscape is what analysts call the "circular economy" problem. Large AI companies are, to a remarkable degree, each other's biggest customers.

Consider the dynamics: Microsoft invests billions in OpenAI. OpenAI then paid Microsoft $10.59 billion for research and development expenses, $6.047 billion in cost-of-revenue charges, and additional amounts bringing total payments to Microsoft to $17.2 billion for the year. Microsoft, in turn, paid OpenAI $303 million in 2025. Amazon, Google, and Microsoft all invest in AI startups that spend heavily on their own cloud platforms, creating a loop of recycled investment capital that can look, from the outside, like organic market demand.

This is not unlike the synthetic demand structures seen in pre-2008 finance, though the comparison should not be overstated. The risk here is subtler: inflated apparent demand could mask a much smaller addressable market of genuine end-user spending. Goldman Sachs analysts found that while the US tech sector may not be in a bubble yet, Sequoia's David Cahn argues the only way to justify the large data centre buildout forecasted by 2030 is AGI.

If investment sentiment shifts, this circular demand could collapse simultaneously. When companies stop investing in each other, the revenue disappears from both sides of the equation.

6. AI Capital Expenditure: The Debt-Fuelled Infrastructure Gamble

The Staggering Scale of AI Infrastructure Spending

The numbers are hard to overstate. Combined hyperscaler capex roughly tripled from ~$226B in 2024 to ~$410B in 2025 and ~$725B guided for 2026. Goldman Sachs' baseline aggregate capex estimate stands at $7.6 trillion between 2026 and 2031, across compute, data centres, and power.

This spending is consuming an ever-larger share of these companies' financial resources. Consensus estimates of AI capex suggest it will climb to 94% of operating cash flows, minus dividends and share repurchases, in 2025 and 2026, up from 76% in 2024.

Who Bears the Risk If Demand Disappoints?

The hyperscalers have strong balance sheets, but they are making multi-year, largely irreversible infrastructure bets. Hyperscalers have burned through all their free cash flow from operations and are now issuing debt to fund the build-out. Data centre debt issuance doubled to $182 billion in 2025 alone.

Smaller AI companies and startups funded by venture debt face existential risk in a downturn. Data centres are physical assets. If AI demand underwhelms, these become stranded assets, much like the overbuilt fibre optic networks of the dot-com era. AI data centre facilities coming online face $40 billion in annual depreciation costs while generating only $15–20 billion in revenue at current usage rates. That math doesn't work long-term.

The link between AI capital expenditure and the broader tech bubble narrative is direct: the spending is being justified by revenue projections that assume AI adoption will accelerate far faster than current enterprise data suggests.

7. The Bull Case: Why AI May Not Be a Bubble After All

Real-World Adoption Is Accelerating, Not Slowing

The bull case for AI rests on one critical observation: the technology works, and adoption is accelerating. ChatGPT crossed 1 billion monthly active users in May 2026, the fastest any app has reached that scale.

Enterprise adoption is also gaining momentum. Improving productivity and efficiency top the list of benefits achieved from enterprise AI adoption, with two-thirds (66%) of organisations reporting gains. Code generation tools like GitHub Copilot show near-universal positive ROI evidence. AI-driven drug discovery, materials science, and medical diagnostics are producing genuine scientific breakthroughs.

Structural Differences That Suggest Durable Value

Unlike many dot-com promises, generative AI demonstrably does what it claims. Stock prices have been reflected by powerful and sustained profit growth rather than excessive speculation about the future. AI is being deployed across industries (finance, legal, healthcare, manufacturing, logistics), not just one sector. It is a horizontal enabling technology, like electricity or the internet itself.

Goldman Sachs' Peter Oppenheimer notes that radical new technologies tend to attract significant capital, and while they don't always end up with a spectacular bubble, there's usually a sharp, industry-wide decline in prices. "Even in cases where a bubble bursts and many companies eventually collapse, this does not mean that the technology itself fails."

Productivity Gains at Scale: The Long-Run Macro Case

Goldman Sachs estimates AI could deliver a 6.1% GDP uplift by 2034. Historical precedent suggests that electrification took decades to show up in GDP statistics but ultimately transformed economies. The "productivity paradox" may delay visible returns, but this does not mean returns won't materialise. The long-run macro case for AI remains compelling, even if the short-term market pricing is stretched.

8. Institutional and Regulatory Signals: When Central Banks Sound the Alarm

The Bank of England AI Warning

The Bank of England said that artificial intelligence poses a growing threat to financial stability, as investors bet heavily it will prove a success while the technology increases banks' vulnerability to cyberattacks.

The FPC's concerns are specific and worth paying attention to. In its October 2025 Record and December 2025 Financial Stability Report, the Bank flagged the potential financial stability risks posed by a sharp decline in AI-related asset prices, noting that while AI infrastructure investment has mostly been financed by cash flows of large, profitable technology companies, debt financing is increasing quickly. The potential for market concentration in AI-related services, including vendor-provided models, is a further challenge.

Other Institutional Voices

The SEC has intensified scrutiny of AI-related disclosures. The SEC's AI washing enforcement began in March 2024 with actions against two investment advisory firms for making false and misleading statements about their use of AI. SEC officials reiterated that "rooting out" fraud schemes related to AI washing is an immediate priority.

The EU AI Act introduces compliance costs that could significantly impact AI valuations, with penalties of up to €35 million or 7% of global turnover for prohibited AI practices. If regulation tightens significantly, the downside scenario for overvalued AI companies could be severe. These institutional and regulatory signals add credibility to the view that pockets of the AI market are stretched.

9. The AI Gold Rush: Winner-Takes-All Dynamics and Irrational Exuberance

The Race Nobody Can Afford to Lose

Every major tech company feels compelled to invest in AI regardless of near-term ROI. The logic is competitive survival: in platform markets, first-mover advantages can be decisive. The "AI winner takes all" thesis means that each individual company's decision to invest may be rational, even if the aggregate level of investment across the entire industry is not.

Goldman Sachs identified a striking dynamic: the engine driving the AI buildout does not appear to be a rational capital allocation process. It is insecurity, if not outright fear. Covello had predicted that if hyperscaler stocks underperformed, companies would cut AI capex. The opposite has happened. Microsoft, Amazon, Google, and Meta have dramatically increased spending even as their stocks have lagged the S&P 500.

The arms race in foundation model development (GPT-5, Gemini Ultra, Claude, Llama) ensures that no company can afford to slow down without risking permanent disadvantage.

Retail Investor and Media Amplification

Media coverage of AI milestones triggers retail investor inflows, creating a feedback loop. Social media influencers amplify AI hype narratives, and smaller AI-adjacent companies sometimes experience meme-stock-style price movements. Three-quarters of executives (75%) admit their company's AI strategy is "more for show" than actual internal guidance. Nearly half (48%) call AI adoption a massive disappointment.

The classic warning sign of a bubble is when everyone from professional investors to casual observers is recommending the same trade. When the consensus is overwhelming, the risk is highest.

If you find yourself caught up in investment hype cycles, having a stable financial foundation matters. Tools like Bleap's savings vaults, offering 3.65% or 3.83% AER in USD with no lock-ins, provide a predictable, low-risk place to park funds while you evaluate higher-risk opportunities carefully.

10. What Would an AI Market Correction Actually Look Like?

Trigger Scenarios for an AI Market Correction

Several plausible catalysts could trigger an AI market correction. A high-profile AI product failure or safety incident could dent public confidence. An earnings disappointment from a major hyperscaler revealing AI monetisation shortfall would be particularly damaging. A significant AI startup collapse would serve as the "Pets.com moment" for generative AI. OpenAI has an incredible $1.4 trillion in spending commitments over the next few years, making it particularly vulnerable if revenue growth stalls.

Macro triggers also matter: rising interest rates make speculative growth stocks unattractive, and regulatory shocks (a major ban or liability ruling) could reshape the landscape overnight.

The Mechanics and Magnitude of a Potential Correction

A correction could take different forms. A sector rotation, where AI stocks fall while others rise, would be disruptive but manageable. A systemic market event, given AI stocks' weight in major indices, could drag broad markets down significantly. A fall in AI-related asset prices could adversely impact US economic growth through a fall in business investment and a consumption response through wealth effects.

Contagion to private markets would mean VC portfolio markdowns, unicorn "down rounds," and talent market contraction. Forrester predicts a market correction, with enterprises deferring 25% of planned 2026 AI spend into 2027.

Historical comparison provides sobering context. The dot-com bust resulted in a 74% total-return drawdown, 31 months from peak to trough, then nearly 12 years to climb back. Nobody should assume an AI correction would be mild or short-lived.

11. After the Hype Cycle: Which AI Applications Survive and Who Wins Long-Term?

Gartner's Hype Cycle and the "Trough of Disillusionment"

Generative AI appears to be transitioning from the peak of inflated expectations toward the trough of disillusionment, the phase where failed pilots pile up, subscriptions get cancelled, and vendor consolidation begins. S&P Global found that 42% of companies abandoned most of their AI projects in 2025. IBM put the number of initiatives delivering expected ROI at 25%. Gartner projects over 40% of agentic AI projects will be cancelled by 2027 due to unclear ROI and weak governance.

This is not a death sentence for the technology. Cloud computing passed through a similar trough post-2001. SaaS companies endured a reckoning post-2008. Both emerged stronger. The trough separates the genuine from the speculative.

Post-Bubble AI Survivors: The Use Cases With Real Staying Power

Not all AI applications are created equal. Code generation and software development productivity show near-universal positive ROI evidence. Healthcare diagnostics and drug discovery offer measurable, defensible value. Customer service automation, legal and financial document processing, and scientific research are demonstrating durable utility.

Use cases likely to underperform or disappear in a correction include generic chatbots without differentiation, AI-generated content at scale (already facing quality and legal backlash), and novelty consumer apps built on hype rather than genuine utility.

Who Are the Long-Term Winners?

The dot-com aftermath offers the clearest lesson. Amazon's stock fell over 90% during the crash but went on to become one of the most valuable companies in history. Google launched during the bust and thrived. The bubble's lesson wasn't that internet companies were worthless. It was that speculative pricing detached from business reality doesn't hold up, and that the distance between a promising idea and a sustainable company is enormous.

The long-term winners in AI will likely include a small number of hyperscalers that own the compute, and application-layer companies with proprietary data, deep domain expertise, and strong distribution. Business model durability matters more than technological novelty.

Want your spending to work harder while markets sort themselves out? Bleap's self-custodial Mastercard gives you 0% FX fees and up to 20% cashback on everyday spending. No monthly subscription, no hidden charges. Get the Bleap card →

12. Verdict: AI Bubble, Necessary Correction, or Justified Growth?

Having examined the evidence from every angle, the honest answer is this: AI is probably not a single, uniform bubble, but pockets of speculative excess clearly exist.

Three scenarios frame the range of outcomes:

Scenario A (Soft landing): AI monetisation gradually catches up with investment. Stock prices consolidate without crashing. Productivity gains become visible over 3–5 years. This is the most optimistic scenario, and it requires enterprise AI adoption to accelerate meaningfully beyond current levels.

Scenario B (Sector correction): Overvalued AI-pure-plays correct sharply, perhaps 30–50%. Infrastructure spending moderates. Winners and losers become clear. The broader market is relatively unaffected. This scenario has the strongest historical precedent and the most supporting evidence.

Scenario C (Hard bust): A major negative catalyst triggers a cascading correction across AI stocks and into broader markets, amplified by concentration risk and debt-funded infrastructure. Parallels to the dot-com crash intensify. This scenario is less likely but not implausible, particularly if the circular investment dynamics unwind simultaneously.

Based on the evidence presented, Scenario B appears most probable. The technology is real, but the financial enthusiasm has outpaced reality in measurable ways: Goldman Sachs notes that "56% of Americans say they use AI, yet 85% of the workforce does not have a value-driving AI use case." The gap between usage and value creation is the central tension.

Whatever happens with AI stocks, protecting your own financial position matters. While the macro picture sorts itself out, focus on what you can control. Bleap's savings vaults offer 3.65% AER (Steady) or 3.83% AER (Dynamic) in USD, starting from $1, with 0% withdrawal fees and no lock-ins. Pair that with a self-custodial Mastercard that charges 0% FX fees and gives up to 20% cashback on everyday spending. You don't need to predict the AI market's future to make your money work harder today.

Try Bleap →

FAQ

Is AI actually in a bubble right now?

AI is not in a uniform bubble, but specific segments of the market show clear signs of speculative excess. Pre-revenue AI startups with multi-billion valuations, circular revenue relationships between hyperscalers and their portfolio companies, and forward earnings assumptions that require perfection all point to overvaluation in pockets. The infrastructure layer (Nvidia, cloud providers) is backed by real revenue growth, which distinguishes it from the worst dot-com excesses.

How does the AI bubble compare to the dot-com bubble?

The parallels are real but imperfect. Both feature explosive stock price growth driven by narrative, FOMO-driven investing, and infrastructure buildout ahead of proven demand. At the peak of the dot-com bubble in late 1999, the PEG ratio was 3.7x versus 1.7x today, pointing to a more conservative pricing picture than in the late 1990s. The key difference is that AI's leading investors are profitable incumbents with strong balance sheets, not speculative startups.

What could trigger an AI market correction?

The most likely triggers include an earnings miss from a major hyperscaler revealing AI monetisation shortfalls, a high-profile AI startup collapse, macro shifts like rising interest rates, regulatory shocks from the EU AI Act or SEC enforcement, or a significant AI safety incident. Any of these could shift investor sentiment rapidly.

Is it safe to invest in AI stocks in 2026?

That depends entirely on which layer of the AI stack you are investing in and at what valuation. Infrastructure companies generating real revenue are lower risk than pre-revenue application startups. Diversification matters more than ever, given that a few AI-adjacent stocks account for a disproportionate share of index gains. If you are looking for lower-risk ways to grow your money, consider stable options like Bleap's savings vaults (3.65% or 3.83% AER in USD) alongside any speculative positions.

Which AI companies are most likely to survive a correction?

Historically, post-bubble survivors share three traits: real revenue from real customers, efficient cost structures, and defensible competitive positions. In the current AI landscape, this favours hyperscalers with proprietary data and distribution (Microsoft, Google, Amazon), infrastructure providers with genuine demand, and application companies solving specific, measurable problems in healthcare, legal, finance, or software development.

What did the Bank of England say about AI risks?

In April 2025, the Financial Policy Committee set out the ways by which AI adoption in the financial sector could pose financial stability risks, focusing on four areas: greater use of AI in core financial decision-making, greater use of AI in financial markets, operational risks from AI service providers, and the changing external cyber threat environment. The Bank has consistently flagged concentration risk and the potential for a sharp decline in AI-related asset prices to have wider financial stability consequences.