Cloud Giants’ AI Revenue: Wall Street’s Roundabout Worries

Wall Street is getting increasingly concerned about the glittering numbers reported by big tech companies in their cloud revenue sheets. The worry stems from the suspicion that these growth figures are being artificially inflated by massive investments in AI startups and related projects. This skepticism has led both analysts and investors to focus on a practice known as “revenue round-tripping.”

Revenue round-tripping, in simple terms, is when a big tech company invests in an AI startup, and then that startup uses a portion of the investment to purchase cloud and AI services from the very company that provided the funding. The money essentially makes a round trip, going from the tech giant to the startup and then back to the tech giant in the form of cloud spending. The tech companies then get to report this as cloud revenue, giving the impression of robust growth. The catch is, this growth might not necessarily stem from actual customer demand but rather from these cyclical investment arrangements.

Several high-profile deals have already raised eyebrows. For instance, Amazon Web Services (AWS) invested a staggering $4 billion in Anthropic, an AI startup, and in return, Anthropic committed to using AWS as its “primary cloud provider.” Google and Microsoft have engaged in similar tactics, with Google partnering with various AI firms and Microsoft investing in OpenAI, all while establishing themselves as these startups’ cloud service providers. Oracle didn’t want to be left out either and became Cohere’s cloud partner shortly after investing in the startup. Such partnerships, while seemingly symbiotic, have fuelled concerns that these revenue numbers are not entirely organic.

Business Insider was one of the first to highlight these issues last year, pointing out that some investors were uneasy about these arrangements. Notably, Jaluria, an analyst, questioned whether AWS’s revenue numbers included cloud services for training Anthropic’s models and whether Oracle’s numbers accounted for training Cohere’s models. He also pondered whether Google Cloud Platform (GCP) recognized any revenue from its internal AI projects. Jaluria’s concerns focused on the comparability between different cloud vendors and whether these inflated numbers painted an overly optimistic picture of a broad-based recovery in cloud workloads.

Amidst these widespread concerns, it’s worth noting that not all tech giants might be indulging in this practice. For instance, Microsoft recently disclosed that it does not recognize any revenue from OpenAI training its GPT models on Azure’s cloud infrastructure. This transparency might place Microsoft in a slightly different light compared to its peers. On the other hand, when approached for clarity, an Amazon spokesperson declined to confirm whether AWS’s revenue figures included cloud spending by Anthropic, leaving investors and analysts guessing.

The broader context here is that cloud spending growth has decelerated in recent years as businesses tighten their belts amidst a sluggish economy grappling with high inflation. If a portion of what appears to be a recovery in cloud spending is actually due to revenue round-tripping, then the narrative of a thriving cloud market becomes questionable. Investors and analysts will likely continue to scrutinize these numbers to determine the true health of the industry.