“股权换采购”--AMD与OpenAI的协议是“半导体历史上罕见”的

Core Viewpoint - AMD and OpenAI have announced a groundbreaking GPU supply agreement worth up to $90 billion, utilizing an unprecedented "equity-for-purchase" model that could redefine financing for AI infrastructure [3][4]. Group 1: Agreement Structure - The agreement involves OpenAI purchasing up to 6 gigawatts of AMD Instinct GPUs, with potential sales reaching $90 billion. In exchange, AMD will issue warrants for up to 160 million shares at an exercise price of $0.01 per share [3][4]. - If AMD's stock reaches $600, the value of the 160 million shares could total $96 billion, equivalent to the hardware's value in the agreement [3]. - The structure of the warrants is designed to function like performance-based equity incentives, linking AMD's long-term valuation directly to OpenAI's infrastructure growth [6][9]. Group 2: Strategic Benefits - For AMD, this agreement represents an innovative customer acquisition strategy, converting traditional upfront discounts into equity costs tied to future performance, thus securing a minimum revenue of 1 gigawatt while mitigating dilution risks [9][10]. - OpenAI benefits by ensuring a stable source of non-NVIDIA hardware in a tight supply market and creating a potential self-funding pathway through the appreciation of its AMD shares, which can be used to finance future GPU purchases [11]. Group 3: Industry Implications - This transaction signifies a shift in AI computing power from a mere capital expenditure to a financialized and securitized asset class, showcasing a different ecosystem-building paradigm compared to NVIDIA and Intel [12][14]. - AMD's model does not require partners to co-invest directly but maintains the core of commercial transactions, allowing OpenAI to purchase hardware fully while incentivizing them to expand procurement through equity rewards [15][16]. Group 4: Risks and Challenges - The agreement faces execution risks, including the lack of transparency regarding key timelines and technical triggers for equity vesting, making it difficult to predict revenue recognition [18]. - The target of 6 gigawatts is a ceiling rather than a firm commitment, with future purchases dependent on OpenAI's business needs and performance [19]. - AMD must maintain a stable supply chain for the delivery of such a large GPU cluster, with potential disruptions in any part of the supply chain posing significant challenges [20].