The Declining Monetizable Spread in AI Models: A Threat to Closed-Source Valuations

The debate over the value of closed-source versus open-source artificial intelligence (AI) models has intensified as the capability gap between them narrows. While the raw performance difference, or capability spread, is a critical metric, it's the monetizable spread that truly matters for valuations. This subset of the capability delta-what enterprises are willing to pay a premium for-is declining more rapidly than anticipated, posing significant risks to the equity of frontier AI labs.

The Shrinking Capability Spread

At the end of 2023, the best closed-source model scored approximately 88% on the MMLU benchmark, a standard measure of knowledge performance, while the leading open-source model managed about 70.5%. By early 2026, this gap has effectively vanished on knowledge benchmarks and is single digits on most reasoning tasks. The time lag between state-of-the-art closed models and their open counterparts has also decreased dramatically. In late 2024, Epoch AI measured this lag at about a year; by the present, it has compressed to roughly three months.

DeepSeek's V3 base model exemplifies this trend. It achieved comparable performance using just 2.6 million GPU hours, compared to Llama 3 405B’s 30.8 million-a tenfold improvement in training efficiency. The R1 reasoning model built on top of DeepSeek's base model matched OpenAI’s o1 at approximately 3% of the cost. This efficiency and performance parity are eroding the traditional advantages of closed-source models.

Investment Implications

The declining monetizable spread is a critical factor that markets have yet to fully price in. For investors, this trend suggests that the equity valuations of leading AI companies like OpenAI and Anthropic may be overpriced if they are primarily valued as utilities. The standard pushback includes enterprise agreements, safety certifications, distribution networks, research talent, and regulatory positioning. While these factors provide some moat, they are not sufficient to justify current valuations in the long term.

The markets are already showing signs of caution. Amid the excitement of the AI revolution, tech stocks have become creatures of hype and price momentum. However, recent wobbles suggest that a more grounded assessment is taking hold. NVIDIA's potential $250 billion financing guarantee for OpenAI underscores the high stakes involved but also highlights the risk of overreliance on financial support to maintain valuations.

For investors, the key takeaway is to reassess the value proposition of closed-source AI companies. The rapid advancement of open-source models means that the premium paid for proprietary technology may no longer be justified. This shift in dynamics could lead to a revaluation of these stocks, potentially resulting in lower multiples as the market adjusts to the new reality.

The declining monetizable spread is not just a technical issue but a financial one with significant implications for investors and stakeholders. As open-source models continue to improve, the competitive landscape will become more crowded, and the differentiation between closed and open models will diminish. This trend demands a reevaluation of investment strategies and a closer look at the underlying fundamentals driving AI valuations.