Goldman Sachs’ 1-Delta Desk recently shared critical insights on a crucial chart that illustrates the competitive landscape between open-source and closed-source AI models. This chart serves as a key indicator for where the AI infrastructure space is headed.
Their analysis, dated June 9, 2026, presents a straightforward assertion: The AI sector is shifting from scarcity to abundance. This transition poses significant challenges for firms that price proprietary models at a premium.
#Why Are Open-Source Models Gaining Ground?
The Goldman team highlights that certain open-source models, including Gemini 3 Flash, Kimi K2.6, and DeepSeek V4 Pro, are either matching or outperforming their closed counterparts like GPT-5.5. The notable aspect is that they are achieving this at approximately half the operational cost.
The focus is on real-world compute pricing and model performance metrics instead of getting lost in the current hype. For instance, DeepSeek has dramatically cut its token prices by 75%, intensifying what the desk describes as a “token price war.” In this scenario, AI providers are fiercely competing by reducing inference costs, further disrupting the market.
#What Are the Supply-Side Risks?
Investors should be aware of two main supply-side pressures that could negatively impact valuations despite growing AI adoption. The first is equity issuance, wherein AI companies increasingly seek funding from capital markets. This often leads to dilution, putting downward pressure on per-share valuations, even amid rising revenues. The second pressure comes from local inference, where smaller and more efficient AI models led to a shift away from centralized cloud services, favoring on-device or on-premise deployments.
#What Does This Mean for Investors?
In an environment defined by scarcity, the power lies with entities that control critical resources like GPU manufacturers and cloud service providers with exclusive model partnerships. However, in a setting characterized by abundance, the advantage shifts to companies that successfully deploy AI at scale, lower costs for enterprises, and establish strong positions through strategic distribution and data advantages.
For traders and portfolio managers, it is crucial to keep an eye on specific metrics highlighted by the Goldman team. Real-world compute pricing trends, benchmark performance between open and closed models, and the rapid changes in token pricing across major providers are all important indicators to watch.
Moreover, the potential for re-rating of infrastructure premiums is significant. Should open-source models continue their progress in closing the performance gap, it could lead to meaningful reductions in these premiums. Conversely, if closed models can solidify their performance advantage, they may maintain their current market multiples. Thus, the pivotal question is not if open-source AI can compete, but rather if closed-model providers can justify their high prices when faced with comparable alternatives.