From dot-coms to AI start-ups: Same boom, different bust

The rise of corporate investment in artificial intelligence (AI) recalls the dot-com boom of the late 1990s: Both were fueled by transformative technology, abundant capital and a fear of missing out on the next dominant platform. In each era, investors funded companies whose valuations depended less on current earnings than on anticipated future adoption. Yet while both booms produced speculative excess and eventual financial distress, the bankruptcy issues arising from dot-com failures and AI-company failures are materially different.

Industry observers have described current AI spending as a modern gold rush. Goldman Sachs Research estimates that global AI-related investment will exceed $1 trillion in 2026, including approximately $581 billion in the United States. The reasons for this investment are straightforward: businesses expect AI to shift costs from human labor to scalable digital infrastructure, while startups market AI as the core of their products, services or platforms. But the pace of capital deployment has also increased the risk that many companies will fail before proving sustainable demand, defensible technology or a path to profitability.

The dot-com boom followed a similar pattern. The rise of the World Wide Web produced extraordinary investor enthusiasm, and internet-based companies raised billions of dollars despite limited operating histories. Some existing companies added ".com" to their names to capture market attention. When the bubble burst, many companies discovered that brand recognition and market share were not substitutes for durable revenue, operating discipline or manageable debt.

The resulting bankruptcies illustrate both the common logic of technology bubbles and the distinct risks facing creditors. Dot-com debtors often failed after overbuilding physical infrastructure, overspending on marketing or misjudging consumer adoption. AI debtors may fail for those reasons too, but their cases can also involve more difficult questions about intellectual property, data rights, licensing restrictions, model ownership and whether the company's claimed AI assets exist or can be monetized.

Webvan Group Inc. is a useful dot-com example. On July 13, 2001, Webvan and three affiliates filed chapter 11 petitions in the United States Bankruptcy Court for the District of Delaware. Webvan had expanded into multiple metropolitan markets and spent heavily on automated warehouses, logistics systems and delivery trucks before confirming sufficient customer demand. Its bankruptcy therefore reflected a familiar dot-com problem: the company built an expensive operating platform faster than revenue could support it.

EToys Inc. presents a related example. On March 7, 2001, EToys and three affiliates filed chapter 11 petitions in the Delaware Bankruptcy Court. The company had anticipated a major online holiday shopping season, entered into an $18 million marketing agreement with America Online, and accumulated substantial inventory. When sales came in far less than expected, EToys was left to liquidate tangible goods whose existence and ownership were comparatively easy to identify.

Recent AI bankruptcies show a different profile. On Aug. 27, 2026, BioXcel Therapeutics, Inc. and two affiliates filed chapter 11 petitions in the Delaware Bankruptcy Court. BioXcel, an AI-driven biopharmaceutical company focused on neurological medicine, disclosed estimated liabilities of $100 million to $500 million and assets of $10 million to $50 million in its petition. Its filing followed cash constraints, debt pressure and efforts to pursue strategic alternatives, including a sale, merger or licensing transaction. For creditors, the value of such a debtor may depend not on inventory or trucks, but on regulatory assets, patents, trade secrets, data and licenses.

That difference matters because a bankruptcy estate includes the debtor's legal and equitable interests in property, including intellectual property such as patents, patent applications and trade secrets. See 11 U.S.C. §§ 541(a), 101(35A). The estate also includes executory contracts, generally understood as agreements under which material obligations remain on both sides. Although 11 U.S.C. § 365(f)(1) generally permits assignment of executory contracts notwithstanding anti-assignment clauses, 11 U.S.C. § 365(c)(1) limits assignment where applicable non-bankruptcy law permits the nondebtor party to refuse performance from a third party and that party does not consent.

Those limits may be especially important in AI cases. AI companies frequently rely on licenses covering software, models, training data, cloud infrastructure, research collaborations or regulated technology. If those rights are nonexclusive, personal to the licensee or subject to consent rights, they may be difficult to assume and assign to a buyer. By contrast, Webvan sold software and domain-name assets in bankruptcy, and the transferability of those assets does not appear to have been a central obstacle. Creditors in AI cases therefore should scrutinize whether the debtor owns its core technology outright or merely uses it under contracts that may not be transferable.

AI cases may also present heightened diligence issues because some startups engage in "AI washing," exaggerating or fabricating automated capabilities to attract capital or inflate valuation. Engineer.ai Corp., doing business as Builder.ai, filed a voluntary chapter 7 petition in Delaware on June 2, 2025. Public reports described allegations that its popular "Natasha" AI assistant was substantially supported by human engineers rather than any autonomous AI. Unlike EToys, where the liquidation involved identifiable tangible inventory, an AI debtor's asserted value may depend on disputed or opaque technology claims. Creditors and investors should therefore verify not only the value of intellectual property but also its existence, ownership, transferability and practical utility.

In sum, while there are similarities between the dot-com and AI "gold rush" periods and their subsequent failures, the bankruptcy filings by companies in those industries demonstrate significant differences of which creditors should be aware. Dot-com cases often centered on overbuilt infrastructure and tangible assets, while AI cases may turn on harder questions involving intellectual property rights, data, licenses and the reliability of claimed technological value.