Amazon's $10 Billion Bet on OpenAI: A Win-Win Gamble?

TradingKey - According to a December 16 report by The Information, Amazon is in talks to invest at least $10 billion in OpenAI. As part of the proposed agreement, OpenAI plans to adopt Amazon's proprietary Trainium AI training chips.

This development comes less than six weeks after the companies signed a $38 billion cloud services agreement. The November agreement stipulated that OpenAI would spend $38 billion over the next seven years to lease Amazon Web Services (AWS) servers, gaining access to Amazon EC2 UltraServers equipped with tens of thousands of Nvidia GPUs.

However, the November agreement did not mention OpenAI's use of Amazon's proprietary chips. These agreements' rapid evolution highlights the swift escalation of collaboration between the two companies.

If this latest agreement goes through, it could propel OpenAI's valuation beyond $500 billion, while also reducing its reliance on Nvidia's chips and Microsoft's Azure cloud infrastructure. For Amazon, this partnership would introduce a new customer for its proprietary chips, demonstrating its progress in the AI sector.

OpenAI: Built by Nvidia, Bound by Nvidia

Previously, the agreement stipulated that OpenAI would purchase $38 billion in cloud infrastructure capacity from AWS. This collaboration, which introduces Amazon's self-developed chips, primarily aims to deeply integrate the computing power infrastructure of both companies.

Currently, OpenAI primarily utilizes Nvidia chips for model training, and its software framework is heavily dependent on Nvidia's CUDA platform. The introduction of Trainium chips could help OpenAI reduce its reliance on Nvidia's chips.

To mitigate its dependence on Nvidia chips, OpenAI signed a multi-year agreement with Broadcom in October this year, planning to collaborate on the development of custom chips and networking equipment.

The relationship between OpenAI and Nvidia is both close and complex. In September this year, the two parties signed a letter of intent for Nvidia to invest $100 billion in OpenAI, in exchange for OpenAI committing to deploy Nvidia systems with a capacity of at least 10 GW.

These collaborations undoubtedly provide OpenAI with state-of-the-art hardware, a robust supply chain, and financial security. However, from another perspective, OpenAI is almost becoming a subsidiary of Nvidia, not only losing its bargaining power but also being constrained by Nvidia's ecosystem. Should it choose to use non-Nvidia chips in the future, it would face significant architecture migration costs.

Furthermore, the capital market is not optimistic about the intertwined relationship between Nvidia and OpenAI. The biggest criticism is that Nvidia invests in OpenAI, and OpenAI, in turn, uses that capital to lease Nvidia's chips, which is almost like "lifting oneself by one's own bootstraps" to inflate valuations. This "circular funding" model, while potentially boosting valuations, does not generate genuine external growth, posing risks.

$500B Valuation vs. $100B Cash Burn: OpenAI’s Bottomless Financing Pit

OpenAI completed an employee stock sale in October this year, valuing the company at $500 billion in the transaction. This exceptionally high valuation reflects market expectations for the future potential of general AI. However, to meet market expectations and maintain its leading position, OpenAI must "burn through" significant capital to lease computing power and continuously update and iterate its models. This capital-intensive development model implies that OpenAI will face substantial capital requirements until it achieves profitability.

According to forecasts, OpenAI will consume over $100 billion for server expansion and talent acquisition. OpenAI previously stated that it plans to raise approximately $90 billion by 2027. Therefore, the $100 billion investment and chip "support" from Amazon will be a lifeline for OpenAI's current predicament.

Amazon’s $10B Investment: The Decisive Strike into the AI Inner Circle

Although Amazon's AWS has become the largest provider in the cloud business sector, it relies on Nvidia's chips. AWS essentially serves as a 'showcase' for Nvidia's chips, , which places Amazon in a somewhat awkward position within the AI sector. For Amazon, a successful agreement with OpenAI would be tantamount to its self-developed Trainium chip receiving a powerful endorsement from a leader in the large language model industry. Given that OpenAI possesses the most sophisticated models, if these models can run successfully on Trainium chips, it would signify that Amazon's AI chips have officially entered the forefront of the AI arena.

Among current large language models, Microsoft's Azure and OpenAI's ChatGPT, or Google Cloud and Gemini, have all become deeply integrated pairings. This agreement, requiring OpenAI to run its models on AWS, could undoubtedly disrupt the duopoly of Google and Microsoft, attracting more customers to access large language models from the AWS platform.

Beyond market positioning, profitability concerns constitute another key consideration. Should Trainium's performance successfully pass OpenAI's validation, Amazon would gain the leverage to 'de-Nvidia-fy,' no longer needing to rely on Nvidia. Instead, it could choose to deploy the AWS platform using its lower-cost, self-developed chips, thereby significantly reducing computing costs and boosting profit margins.

Furthermore, according to prior reports, OpenAI and Amazon have also explored cooperation in the e-commerce sector, including transforming ChatGPT into a shopping hub and integrating OpenAI's inference capabilities into Amazon's e-commerce platform. If this collaboration materializes, Amazon would gain an ideal traffic gateway, given ChatGPT's high daily active users and strong user stickiness, which would benefit Amazon in competing against rapidly emerging rivals such as TikTok Shop and Temu.

As of September 30 this year, Amazon's cash and cash equivalents stood at a substantial $66.9 billion, ranking among the top tech giants. While $10 billion represents a significant investment, it would not impose a financial burden on Amazon.

Of course, whether this investment can be successfully realized remains uncertain, primarily depending on the actual performance of Amazon's Trainium chips. If successful, the current competitive landscape in the AI sector could be reshaped.