The artificial intelligence boom shares important similarities with the dot-com bubble of the late 1990s, but fears of an imminent market collapse and widespread unemployment remain premature, according to veteran technology investor Paul Wick.
Wick, chief investor at Seligman Investments, part of Columbia Threadneedle Investments, told Business Insider that the current surge in AI startups and data-center investment resembles the period around the turn of the century, when internet companies proliferated, and telecommunications operators aggressively expanded wired and wireless networks.
Both periods have featured heavy infrastructure spending, rising energy prices, and higher interest rates. Yet Wick argued that the present market is healthier than the one that preceded the dot-com crash, pointing to the absence of an equivalent initial public offering frenzy and continued strength in enterprise AI spending.
Register for the next Tekedia Mini-MBA.
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
Seligman Investments manages more than $40 billion in assets across a hedge fund, a mutual fund and a venture capital arm, giving Wick exposure to different parts of the investment landscape.
His assessment offers a counterpoint to warnings that the AI industry is repeating the excesses of the internet boom. While the scale of current investment raises questions about future returns, Wick’s argument is that the market has not yet reached the same point of speculative excess or deterioration in corporate spending that characterized the earlier cycle.
Dot-Com Parallels, but a Different Market Environment
During the dot-com era, the rush to commercialize the internet encouraged companies to spend heavily on infrastructure and pushed valuations to extreme levels. Telecommunications operators expanded network capacity in anticipation of demand, while investors poured money into internet businesses whose commercial models had yet to demonstrate sustainable profitability.
The eventual collapse exposed the gap between infrastructure investment, market expectations, and the revenue many companies could generate. Wick sees similarities in today’s AI ecosystem, where technology companies are committing enormous sums to computing infrastructure and startups are competing to build products and services around sophisticated models.
However, he identified important differences. The current market has not experienced the same initial public offering frenzy seen during the dot-com period, nor the “frenetic environment” in which stocks could rise fivefold or tenfold in a single day.
He also noted that, by late 2000, forecasts of a slowdown in information technology spending for 2001 were already emerging. Enterprise spending on AI, by contrast, remains strong, while recent technology earnings have been robust.
“Healthier” was Wick’s description of the current environment compared with 2000.
Infrastructure booms do not necessarily end when investment reaches a high level. Thus, the decisive question is whether demand and revenue can grow sufficiently to justify the capital committed. Continued enterprise spending and strong earnings provide some support for the AI investment case, although they do not guarantee that every company, project or valuation will prove sustainable.
AI infrastructure investment also involves a different commercial landscape from the early internet era. Data-center operators, chipmakers, cloud providers and model developers are increasingly connected through financing arrangements, supply agreements and investment relationships. Those connections can accelerate deployment, but they also make it important to assess whether reported demand ultimately translates into independent customer spending and durable cash flows.
Burry’s Warnings and Nvidia’s Investment Strategy
Wick also addressed concerns raised by Michael Burry, the investor who gained prominence for anticipating and profiting from the collapse of the US housing market in the mid-2000s.
Burry has questioned the sustainability of the AI boom, highlighting aggressive accounting, circular financing, hidden debt, and heavy investment in depreciating assets such as microchips. His concerns center on whether the industry is overstating the durability of demand and understating the financial risks associated with its infrastructure expansion.
Wick pushed back on treating Burry’s reputation as sufficient evidence that his latest warnings will prove correct, saying the investor has been “early or downright wrong on many other things.” Burry has defended his record by pointing to previous correct predictions and publishing charts illustrating the timing of his market calls.
The disagreement comes off a broader challenge for investors attempting to evaluate AI valuations. Evidence of excessive investment may emerge before a downturn becomes visible in company earnings, while warnings made too early can also miss years of further growth. The relevant test is more about the capacity of capital spending, revenue growth and returns on invested capital remaining aligned as the industry matures.
Wick was particularly skeptical of the argument that Nvidia should direct its excess cash primarily towards share buybacks rather than investing in data-center developers and software companies that use its chips.
He described Nvidia as “extraordinarily profitable” and suggested that supporting its broader ecosystem could be more valuable than simply returning additional cash to shareholders.
That strategy has a commercial rationale. Nvidia benefits when customers build more AI infrastructure and develop applications that increase demand for its processors. Investments in customers and adjacent businesses can help expand that market and reinforce the chipmaker’s position within it.
However, the approach also raises questions about the independence and quality of demand across the ecosystem. When a supplier invests in companies that purchase its products, investors need to distinguish between demand generated by end customers and activity supported by financing from companies that stand to benefit from continued spending.
Nvidia also faces a longer-term competitive challenge as some of the industry’s largest technology companies develop their own AI chips to reduce reliance on its hardware. Supporting customers may help preserve demand, but it does not eliminate the risk that customers eventually shift some workloads to alternative processors.
Beyond infrastructure spending, Wick identified pressure on AI pricing as a trend investors should monitor.
Enterprise customers may choose older AI models or develop their own systems rather than pay for the newest commercial offerings. At the same time, wider adoption of open-source models could put downward pressure on token prices, affecting the economics of companies such as OpenAI and Anthropic.
Token pricing is central to the commercial model of many AI providers because it determines how much customers pay for the volume of text or other information processed by their models. If prices fall faster than providers can reduce computing costs, revenue growth may fail to translate into stronger margins.
Lower prices, however, are not necessarily evidence that the market is weakening. They can also expand adoption by making AI more affordable and encouraging companies to integrate it into more products and workflows.
The risk for investors is that a rapidly expanding AI market may still generate disappointing returns for some of its most prominent companies. Strong demand for AI tools does not automatically mean that every model developer can sustain premium pricing or that the capital invested in data centers will earn an adequate return.
AI Job Loss Fears May Be Premature
Wick also challenged predictions that artificial intelligence will rapidly eliminate large numbers of jobs, arguing that the consequences for entry-level employment remain uncertain.
He said the traditional route to success through a computer science degree may no longer be the only attractive option for young people. With demand growing for electricians and heating, ventilation and air-conditioning technicians, he argued that skilled trades can offer strong earning potential without requiring a four-year university degree.
“It used to be that a degree in computer science was a ticket to success,” Wick said.
His argument reflects the physical requirements of the AI expansion itself. Data centers require electrical systems, cooling infrastructure, construction and ongoing maintenance. As investment in computing facilities grows, demand for workers who can build and operate those facilities can rise alongside demand for software engineers and AI specialists.
Wick also described the “doom-and-gloom notion that AI is going to eviscerate demand” for human labor as “a little premature.” He noted that unemployment has yet to soar and that businesses and consumers typically take time to change their behavior in response to new technologies.
That does not establish that AI will have a limited employment impact. Automation could still alter entry-level roles, reduce demand for particular skills, and change how companies organize their workforces. The pace and scale of those changes remain uncertain, particularly as businesses move from experimenting with AI to deploying it across routine operations.
For investors, the labour question is also an economic one. If AI improves productivity and reduces costs, companies may benefit from higher margins. But if adoption is slower than expected, or if cost savings fail to materialize, the financial case for some investments could weaken.
Therefore, Wick’s assessment is neither a dismissal of AI risks nor an endorsement of every investment being made in the sector. It is a warning against treating the current boom as an inevitable replay of the dot-com collapse, while recognizing that pricing pressure, customer self-sufficiency and capital allocation will determine which businesses ultimately capture lasting value.