Are We in a Bubble

It’s worth noting that the historical pattern is remarkably consistent. Every major bubble — from tulip bulbs to dot-com stocks — follows the same arc: a genuine innovation or scarcity captures the imagination, speculative capital floods in, rational valuation gives way to the Greater Fool Theory, and eventually someone is left holding the flaming potato. The truly catastrophic ones (1929, Japan in 1989, the GFC (Great Financial Crisis)) aren't just stock corrections — they infect the broader economy through debt, leverage, or systemic financial contagion.

Dot-Com Bubble

Internet companies with no revenue or path to profit reached trillion-dollar combined valuations. The NASDAQ rose 572% in five years, then crashed 78%. Over half of all public dot-com companies failed by 2004. The NASDAQ took 15 years to reclaim its March 2000 high. The internet itself endured — and transformed civilization.

Severe

2003–08

U.S. Housing / GFC

Post-dot-com capital flooded into real estate, combined with subprime lending and mortgage-backed securities that spread risk globally. When subprime borrowers defaulted, the entire system unraveled. The S&P 500 fell 57%. Global GDP contracted for the first time since WWII. $11 trillion in household wealth evaporated.

Catastrophic

2020–21

COVID Everything Bubble

Zero-rate pandemic stimulus inflated nearly every asset class simultaneously — SPACs, meme stocks, crypto, NFTs. S&P 500 surged ~100% from its March 2020 low. The 2022 rate-hike correction cut 25% off the S&P, but recovery was swift. Many economists now regard 2021 as a bubble by classic definition.

Managed

2023–?

AI Boom / Potential AI Bubble

AI has driven a 125% NASDAQ gain since the ChatGPT launch in Nov 2022. The Magnificent 7 dominate index returns. $539B in AI capex is planned for 2026 alone. Markets are debating: genuine revolution with real earnings, or speculative overshoot? The answer may be both.

And now, we will focus on a comparison between Dot Com and Right Now.  While no consensus exists, a comparison of current (early 2026) market conditions to historical bubble periods, especially the Dot-Com era, reveals a mix of concerning parallels, but also significant, fundamental differences.

 Part II — Dot-com vs. today's AI market

Structural comparison: 1999 vs. 2026

The comparison is everywhere — but the data tells a more nuanced story. Key metrics reveal a market that is elevated but structurally more grounded than 1999.

Dot-Com Peak (1999–2000)

NASDAQ forward P/E~79×

Cisco trailing P/E (peak)~472×

Unprofitable tech stocks~12%

NASDAQ gain (5 yr)+572%

1999 IPOs 400+

Internet household penetration~50%

Revenue model Speculative

Crash depth (NASDAQ)−78%

 AI Market Today (2026)

NASDAQ forward P/E~25×

NVIDIA trailing P/E (peak)~56×

Unprofitable tech stocks~4%

NASDAQ gain (3 yr)+125%

2025–26 IPOs Sparse

AI enterprise penetration~88%

Revenue model Real earnings

Decline YTD (S&P 500)~−7%

 Key structural differences

1999: Valuations expanded without earnings. Companies lost money at scale.

2026: Valuations elevated but matched by explosive revenue growth and free cash flow.

1999: Lucent had $15B in vendor financing vs $300M operating cash flow.

2026: Hyperscalers funding AI capex from profits; $1T in stock buybacks signal insider confidence.

1999: IPO flood — "smart money" cashing out to retail.

2026: IPO drought — companies buying back shares; insiders holding.

Conclusion?  The dot-com comparison is the right frame, but it flatters 1999. The P/E ratio for tech stocks today is just 56% of what it was at the peak of the dot-com bubble, and for chip stocks even lower at 43%. The NASDAQ crashed 78% after its dot-com peak — and took 15 years to reclaim its March 2000 high. Over 50% of public dot-com companies failed by 2004, and venture funding collapsed 95% from its 2000 peak. Today's landscape is structurally different: while the weight of technology in the S&P 500 index today is similar to the late 1990s, the earnings contribution is dramatically different. During the tech bubble, valuations rested largely on speculative optimism. Today's technology leaders generate substantial profits, robust free cash flow, and strong balance sheets.

The most compelling case against "bubble" right now is the IPO silence. Three of the "Four Horsemen" of a bubble are present in early 2026 — valuations are high, retail investors are piling in, and sentiment is frothy — but the absence of issuance disqualifies the current cycle from bubble status. In 1999, over 400 IPOs hit the market. Today, companies are doing the opposite — running nearly $1 trillion in buybacks. Corporations are the smart money, and when they sell equity, that's a sign the equity is overpriced.

Part III — Bubble indicators: how does today score?

Reading the warning signs

According to Economist Owen Lamont's "Four Horsemen" framework: a true bubble requires overvaluation, bubble beliefs, issuance, and retail inflows. Three of four are present in early 2026 — the missing piece is the IPO flood that historically marks peak insider exit.

Overvaluation (CAPE ratio near 1999 levels). High

Bubble beliefs / frothy sentiment Elevated

Retail inflows / speculative participation. Elevated

Equity issuance (IPO flood — the key signal). Low

AI capex concentration risk.  High

Circular financing (hyperscaler feedback loops). Emerging

Fed rate environment (tightening = danger) Supportive

Underlying earnings quality. Strong

Part IV — Risk factors for a burst

High risk

AI capex ROI disappointment

$539B in planned 2026 AI capex. If AI revenue growth fails to justify this, a sudden repricing of hyperscaler earnings expectations could cascade across the market. OpenAI is still running ~$7.8B in annual operating losses on $4.3B revenue.

High risk

Circular financing unraveling

Some AI hyperscalers are financing compute access for AI companies who then pay back the hyperscalers — artificially inflating revenues on both sides. Similar to Lucent's vendor financing in 2000. Vulnerable to any shift in conditions.

High risk

Geopolitical / supply shocks

The Iran conflict has disrupted helium supply from Qatar (1/3 of world supply, used in chip manufacturing). Energy costs for data centers rising. These external shocks can rapidly compress the earnings base that current valuations depend on.

Medium risk

AI scaling law limits

If performance gains from scaling model size plateau, investor expectations — built on infinite improvement curves — would face sharp repricing. The DeepSeek shock of Jan 2025 (wiping $589B from NVIDIA in one day) previewed this risk.

Medium risk

Fed hawkish pivot

Most major bear markets begin when the Fed tightens aggressively. The current dovish path is supportive, but persistent inflation or an energy-driven price spike could force a reversal — the most historically reliable bubble catalyst.

Lower risk

SaaS / software displacement

The "SaaSpocalypse" has already begun — Salesforce and ServiceNow both lost ~30% since early 2025 as investors anticipate AI agents replacing traditional software. This sector bubble may have already partially deflated, limiting broader contagion.

Part V — Will it burst? When? How bad?

Three plausible scenarios

Capital Economics argues the AI stock bubble has already partially burst — tech P/E ratios have compressed since their October 2025 peak. What remains debated is whether a broader market crash follows, and when.

Soft landing — broadening bull

~35% probability

AI productivity gains spread to the broader economy. Earnings growth "catches up" to valuations across more sectors. Market leadership broadens beyond Mag 7. A managed rotation, not a crash. Parallels the 1995–96 "mid-cycle" scenario where Internet productivity arrived 3 years after the initial euphoria.

Significant correction — no recession

~45% probability

AI capex disappoints or circular financing cracks. Tech stocks correct 20–35% (Seeking Alpha estimates 20% S&P downside; NVDA, MSFT, AMZN facing 20–50% retracement). Economy avoids recession due to solid consumer base, fiscal stimulus, and accommodative Fed. Recovery within 2–3 years.

Crash — recession triggered

~20% probability

A perfect storm: geopolitical energy shock raises inflation, forces hawkish Fed pivot, AI revenue disappoints, circular financing unravels simultaneously. S&P falls 40%+. Compares to 2000–02 scenario. Most analysts consider this unlikely without a recession first — and current fundamentals don't indicate one imminently.

When might a major correction arrive?

The most reliable timing signal remains the IPO cycle. Watch for a surge of AI unicorn IPOs — OpenAI, Databricks, Anthropic going public would signal that "smart money" is seeking the exit. Lamont's framework says when insiders rush to sell equity at peak prices, the top is near. That flood has not yet materialized. Until it does, the consensus is that we are in an elevated, risky bull market — not a classic bubble peak.

But one bubble has already partially burst. For information technology and big tech, the price-to-earnings ratio fell from its highs and is now the smallest since the pandemic. The SaaS sector has been hit hardest — both Salesforce and ServiceNow have lost about 30% of their respective values since the beginning of the year, as investors fear agentic AI replacing traditional software business models.

The critical unknown is AI capex ROI. Some large tech companies are increasingly financing their enormous investments through complex structures, with circular financing between providers and users of AI computing capacity artificially inflating revenues. These models become vulnerable as soon as conditions change. The S&P 500 is already down about 7% from its all-time high as of March 31, 2026, as investors reassess risk.

The bottom line

Today's market is elevated and concentrated, but not a dot-com replay. The key differences: real earnings backing today's valuations, strong corporate balance sheets, no IPO flood signaling insider exits, and a supportive Fed. The S&P 500 is already ~7% below its 2026 peak. The more honest answer is that a partial AI bubble has already begun to deflate — software stocks have cratered, and tech P/E ratios have compressed from their October 2025 highs. What comes next depends almost entirely on whether AI capex generates the productivity gains investors have priced in. If it does, this looks like 1997 — early innings. If it doesn't, the unwinding could be sharp. History's lesson is humbling: even when the technology is real (railways, the internet), the financial excess can be catastrophic. The internet destroyed the NASDAQ by 78% — and then changed everything anyway. AI is almost certainly transformative. Whether current valuations survive that transformation intact is a different question entirely.

The honest verdict: this is a market at elevated risk, with a partial correction already underway. A dot-com-scale catastrophe requires conditions — a complete lack of underlying earnings, an IPO feeding frenzy, and a Fed tightening cycle — that don't yet exist. But history never promised those conditions would stay absent forever. Watch the IPO calendar.