As AI Labs Burn Cash, the Real Question Is Which Suppliers Get Paid

As AI labs invest heavily in computing infrastructure, a crucial question arises: which suppliers receive the payments? Microsoft has recently abandoned a planned AI data center project in the UK due to the non-availability of electricity until the 2030s. This cancellation highlights the financial risks associated with AI infrastructure projects.

Companies like Nvidia, Broadcom, and AMD are among the key suppliers in this market. Nvidia sells GPUs and networking equipment, earning revenue upon delivery. Broadcom designs AI accelerators and provides networking equipment but earns payments for supplies rather than recurring fees for data center operations. AMD aims to compete with Nvidia by offering its Instinct AI accelerators and EPYC server processors, including a potential equity incentive tied to OpenAI's success.

Oracle is aggressively expanding into AI infrastructure through its cloud platform, generating recurring revenue. However, the company must secure the necessary equipment, power, and facilities. CoreWeave and Nebius Group operate similarly, developing AI computing infrastructure and selling access to customers. While these companies have long-term contracts, they still need to finance infrastructure, service debt, and deliver services on time.

Alphabet, through Google Cloud, benefits from selling computing capacity to customers, including Nvidia GPUs and its own TPUs. Despite the financial risks associated with AI infrastructure investments, Alphabet has established businesses that can help fund its expansion. The Microsoft-Nscale cancellation underscores the importance of understanding the financial exposure of suppliers versus the companies investing in AI infrastructure.