Table of Contents

• You might want to know


• Main Topic


• Key Insights Table


• Afterwards...

You might want to know

Could the AI industry’s rapid growth be vulnerable because it relies heavily on just two startups?

How might cheaper Chinese open-weight models reshape the competitive and pricing landscape for major cloud providers?

Main Topic

Steve Eisman, the investor famous for profiting from a bet against the housing market before the global financial crisis, has voiced concern that the current artificial intelligence boom is increasingly concentrated around two companies: OpenAI and Anthropic. Eisman noted that these two startups represent a substantial portion of AI-related revenue for large cloud and technology firms. According to his remarks, OpenAI and Anthropic together account for approximately 70% of AI-related income at companies such as Microsoft, Amazon, Google (Alphabet), and Oracle. That level of dependence suggests that a meaningful share of those firms’ AI-driven futures is effectively a bet on the success of these startups.

Eisman expanded on the potential financial impact by estimating that revenue tied to these startups may make up as much as 25% to 35% of those companies’ cloud revenue. In other words, a heavy portion of cloud monetization could be linked directly to the performance and adoption of a very small number of third-party AI providers. For large cloud vendors that have invested heavily in AI infrastructure and marketing, this concentration amplifies exposure to the fortunes of a few partners.

One of Eisman’s central warnings is that the market faces an external competitive threat: Chinese open-source and open-weight AI models. These alternatives tend to be less expensive and, by some accounts, are gaining traction. Eisman suggested that if OpenAI or Anthropic were to encounter setbacks, the lower-cost Chinese models could rapidly win market share. That shift would likely trigger fierce price competition among cloud providers, compressing margins and creating significant revenue pressure. This potential for a price war — driven by cheaper open-weight models — represents what Eisman calls the industry’s Achilles’ heel.

Eisman’s perspective contributes to a broader debate about whether the large-scale investments fueling the AI boom will deliver adequate returns. Some prominent investors worry that the rush to build and deploy AI capabilities may not translate into sustainable profit growth for the major technology firms that are the primary spenders and beneficiaries of the trend.

The conversation echoes the views of other skeptics, notably Michael Burry, another investor known for betting against the housing bubble and featured in The Big Short. Burry has expressed an even more skeptical outlook, questioning whether much of the present and projected AI demand is genuine end-customer demand. He has argued that a portion of spending may be driven by circular or self-reinforcing arrangements rather than organic, customer-led adoption. Acting on that skepticism, Burry has taken bearish positions against several companies seen as prime beneficiaries of the AI surge, including semiconductor leader Nvidia and other firms tied to the chip industry.

Taken together, these critiques highlight two linked concerns for the AI investment thesis: concentration risk and the durability of demand. Concentration risk arises when a large share of an industry’s value depends on a small number of players, increasing systemic vulnerability if those players falter. Durability concerns question whether current spending patterns reflect long-term customer value or are temporarily inflated by hype, vendor financing, or nonrecurring arrangements.

For corporate and investment decision-makers, acknowledging these risks can inform strategy. Firms heavily exposed to a few AI partners might diversify their provider base, invest in proprietary model capabilities, or seek locked-in commercial agreements to reduce dependence. Investors may weigh valuations more conservatively, stress-test cash-flow assumptions for cloud and AI revenue, and monitor competitive developments among lower-cost global providers.

Key Insights Table

Afterwards...

Looking ahead, several areas of technology and policy merit further attention to mitigate the risks Eisman and others have identified. First, advancing model interoperability and open standards could reduce vendor concentration and allow more competitive sourcing of AI services. Second, investment in efficient, cost-effective inference techniques and model compression would make providers less vulnerable to low-cost entrants. Third, clearer commercial models and transparency around customer demand would help determine whether spending represents sustainable adoption or transient effects. Policymakers and industry leaders should also monitor cross-border competition and consider how trade, licensing, and security considerations intersect with market dynamics.

For investors and executives, the prudent path includes diversification of AI suppliers, rigorous scenario analysis of revenue dependences, and an emphasis on building proprietary capabilities where possible. The conversation around concentration and pricing pressure is still evolving, and staying informed about global model development, total-cost-of-ownership trends, and customer adoption patterns will be critical in assessing the long-term prospects of AI-driven businesses.

Ultimately, while AI promises transformative value, Eisman’s warning underscores that a narrow dependency on a few startups — coupled with the rise of cheaper alternatives — could create significant vulnerabilities. Stakeholders should incorporate these considerations into strategic planning and risk management as the industry matures.