In the last several months, new large language models (LLMs) leveraging open-source AI architectures have emerged, producing robust capabilities with significantly lower cost structures. These advancements present risks for investors, particularly concerning the cost-competitiveness of traditional AI data center investment models. There are likely many such innovations underway that are not yet transparent to the public markets, but some recent public examples can help to illustrate the trend.
FoxBrain
On March 10, 2025, Taiwan’s Foxconn introduced its first large language model, “FoxBrain.” Based on Meta’s Llama 3.1 architecture, FoxBrain was trained using 120 Nvidia H100 GPUs over about four weeks.1 Optimized for traditional Chinese and Taiwanese language styles, FoxBrain demonstrates reasoning capabilities close to world-class standards. Its applications include data analysis, decision support, document collaboration, mathematics, reasoning and problem-solving, and code generation. Foxconn plans to collaborate with technology partners to expand the model’s applications and promote AI integration in manufacturing and supply chain management.
DeepSeek
Similarly, DeepSeek, a Chinese AI startup, has garnered attention for its efficient and cost-effective LLMs.* In December 2024, DeepSeek unveiled its R1 model, which matches the capabilities of leading models like OpenAI’s GPT-4 but at a fraction of the cost. The R1 model was trained using approximately 2,000 Nvidia H800 GPUs over 55 days, with a total expenditure of around $5.58 million.2 DeepSeek’s training costs for its V3 model could be higher as the cost cited by the startup only includes the amount spent on computing power. Little is known about the costs of building the more publicized R1 model.
Still, it is a far cry from the $250 billion analysts estimate big U.S. cloud companies will spend this year on AI infrastructure. And for some, these approaches challenge the notion that tremendous market opportunities in high-performance AI development require substantial financial investments.3
Investor Risks and Implications
The advent of LLMs like DeepSeek and FoxBrain poses several risks to investors:
Valuation Adjustments: The emergence of lower-cost AI models has led to significant market reactions. For instance, DeepSeek’s R1 model prompted a substantial selloff in tech stocks, with companies like Nvidia experiencing notable declines in market capitalization. This volatility underscores the potential for rapid valuation adjustments in response to technological advancements. 4
Pressure on Traditional AI Investments: Companies heavily invested in traditional AI infrastructure may face challenges as more efficient models reduce the need for extensive data center investments. This shift could impact the profitability of firms specializing in AI hardware and large-scale data centers 5.
Increased Competition: The accessibility of open-source architectures enables new entrants to develop competitive AI models without substantial capital, intensifying competition and potentially eroding market share for established players.
Regulatory and Ethical Considerations: The proliferation of advanced AI models raises concerns about misuse, necessitating robust regulatory frameworks. Investors must closely monitor regulatory risk and potential liabilities associated with deploying AI technologies.
Strategic Responses
The Information Technology sector has the highest level of self-reported risk to regulators across the global economy, with an accelerating trend over the last several years. Going forward, monthly monitoring will reveal the extent to which AI cost structure disruption is building into a potential regulatory and market value recalibration.6
To mitigate these risks, investors could consider the following strategies:
Monitoring Technological-Risk Trends: Staying informed about emerging technology risks and potential disruptions is crucial for proactive investment decision-making.
Understanding Regulatory Developments: Tracking, understanding, and anticipating regulatory changes can help investors navigate the evolving landscape of AI ethics and compliance.
Diversification: Investing across a spectrum of AI technologies and related sectors can help cushion the impact of rapid advancements in specific areas.
Conclusion
The rapid evolution of LLMs, as demonstrated by DeepSeek and FoxBrain, presents both challenges and opportunities for investors. While these advancements can disrupt traditional investment models, they also open avenues for innovation, growth, and democratizing access to advanced AI technologies. Ongoing management of this evolving risk and a nuanced understanding of these developments is essential for building informed investment strategies in the AI Economy.
[1] Foxconn Unveils First Large Language Model | Reuters
[2] AI companies are copying each other’s homework to make cheap models | Business Insider
[3] DeepSeek’s low-cost AI spotlights billions spent by US tech | Reuters
[4] Blackstone holds firm on data center investments despite DeepSeek turmoil | Reuters
[5] DeepSeek Threatens an AI Revolution outside of the US experts warn | The Australian
[6] The new All Material Risk Investment Strategy (AMRIS) deployed at Boston Common Asset Management has produced a novel array of insights, updated monthly, on risks disclosed to global regulators that are most material to publicly listed equities.
This piece is intended as an editorial review of a survey of risk considerations and does not constitute investment advice, a solicitation, or an offer to buy or sell any security. Past performance does not guarantee future results. All investments involve risk, including the risk of loss of principal.
References to third-party research, publications, and data sources are provided for informational purposes only. Boston Common Asset Management does not warrant the accuracy or completeness of third-party information cited herein. Readers should consult original sources and qualified professional advisors before making any investment decision.
