DeepSeek Announcement & Hidden Risks in the Information Technology Sector
Weeks after the market responded to DeepSeek, there are still lessons to be learned. We’ve discovered that AI exposure is now behaving like a systematic risk factor—one that traditional models fail to capture. Our quantitative analysis using Omega Point’s Thematic Package and Noonum’s AI-driven indexing tools reveals that the market reaction to DeepSeek’s announcement wasn’t just about a new competitor—it signaled a fundamental shift in how AI-driven disruption is priced.
We know that OpenAI triggered an AI arms race, sending valuations of AI-driven companies soaring until DeepSeek’s announcement on January 27, 2025, changed the narrative. The company claims to have built a large language model (LLM) that delivers performance competitive with ChatGPT—at a fraction of the cost. While the AI community has debated the validity of these claims, the market’s response speaks for itself: investors are no longer treating AI companies as a single cohort, and the risks associated with AI exposure are becoming clearer.
Until DeepSeek’s announcement, markets had treated the beneficiaries and facilitators of AI driven disruption as a cohort and assigned them astronomical valuations based on future growth prospects. DeepSeek’s announcement shattered that monolithic image and highlighted fundamental differences between companies that are likely to benefit from AI transformation (AI Beneficiaries) and those that are will facilitate the said transformation (AI Suppliers).
Risk Analysis: Goals
DeepSeek’s announcement created top line (revenue growth) uncertainty for AI Suppliers and margin uncertainty for AI beneficiaries with risks skewed to the downside for the former and upside for the latter. We analyze this phenomenon by building AI Beneficiaries and Supplier indices using Noonum, and examining them through Omega Point’s Thematic Package. Our analysis has a dual focus.
First, we are interested in identifying hidden systematic risks associated with the broader AI theme as characterized by the AI Beneficiaries and Supplier indices. Commercial risk models can capture systematic risks associated with long-term risk factors such as Value, Growth, Momentum, and various industries. We believe that short-term themes such as AI also carry systematic risk that largely remains unaccounted for when examined through the lens of a commercial risk model. Indeed, for a fund manager, the distinction between a long-term and a short-term risk factor is a distinction without a difference, and hence our emphasis on quantifying the latent (hidden) systematic risk in AI indices.
Second, we want to understand the evolution of the risk characteristics of the AI theme within the past one year. For a field that is breaking new ground every few weeks, AI disruption represents unique challenges in the context of portfolio management; DeepSeek announcement and subsequent volatility in the TMT sector being a good case in point. To this analysis, we focus exclusively on stocks within the Information Technology sector.
Risk Analysis: Insights
Noonum platform is designed to seamlessly translate investment objectives stated in plain English into investable indices. Exhibits 1 and 2 show the investment objectives that were used to create the AI Beneficiaries and AI Supplier indices, respectively. These indices are weighted by Noonum’s “exposure” score, a proprietary metric that uses LLM based tools to ascertain the relevance of the investment objective to various stocks within the index. To assess the risk characteristics of these indices, we ran the following experiment.
Using Omega Point’s Thematic Package, we generated historical betas for various US securities to the AI Supplier and Beneficiaries indices after controlling for the usual style and industry factors. Thereafter we ran daily cross-sectional regressions to determine the relationship between residual returns and historical betas. Exhibits 3 and 4 show the daily R2 of these regressions for stocks in the Information Technology sector using the AI beneficiaries and Supplier indices, respectively. It is interesting to note that these indices can explain roughly 1-2% of the residual returns on average, and can have considerably higher R2 on certain days. While 2% explanatory power might seem insignificant on its own, it needs to be examined in the context of risk adjustments in our experimental setup that already control for a variety of traditional long-term style & industry factors.
Exhibits 3 and 4 suggest that AI theme is being priced in the market like a systematic risk factor. We obtained independent validation of this hypothesis using Omega Point’s Thematic Strength Indicator (TSI) metric (cf. Exhibits 5 and 6). TSI is a model agnostic measure of the systematic risk of a theme, and represents the average explanatory power of a theme with respect to residual stocks returns. TSI-Z is a Z-scored version of TSI that ranks more than thirteen hundred themes; a TSI-Z score more than two suggests that the theme is systematically important to the cohort of stocks under consideration.
Exhibits 5 and 6 show the TSI of AI Suppliers and Beneficiaries indices. As demonstrated in the charts, a high TSI-Z score corroborates our hypothesis regarding the latent systematic risk associated within the AI theme. While the actual DeepSeek announcement happened in January 2025, the upward momentum in the TSI score starting in 2024-Q4 suggests that market had already started to price in the uncertainty associated with this development much earlier.
To better appreciate the evolving systematic risk characteristics of AI themes, we generated “term” curve of TSI (cf. Exhibits 7 and 8) by varying the lookback period that is used to compute the metric; the default version of TSI reported in Exhibits 5 and 6 uses a lookback period of 252 trading days. The steepening of the TSI term curve in the last one year is noteworthy (cf. Exhibit 7). Among other things, this suggests that the latent systematic risk associated with AI themes has increased in the recent months. The upward shift at the short end of the TSI-Z term curve captures the impact of the DeepSeek announcement on the AI theme vis-à-vis other themes in our universe.
Conclusion
Rapid technological advancements have a unique way of reminding us the importance of the cliched risk – return tradeoff. While the rest of the world was focused on upside opportunities of the AI driven disruption, DeepSeek announcement on January 27th 2025 revealed downside risks especially for AI Suppliers. For Portfolio Managers, navigating the AI landscape requires effective modeling of the underlying theme and a process-oriented holistic approach to thematic risk management. Noonum’s LLM driven platform for custom index creation and Omega Point’s Thematic Package are designed exclusively for those goals.
Anureet Saxena, Ph.D, CFA
CEO & Founder, Alignment Trio Management
