Hyperscaler RPOs Hit More Than USD 2 Trillion; Amazon Alone at USD 627bn
The remaining performance obligations (RPOs) across Alphabet, Amazon and Microsoft now stand at more than USD 2 trillion, according to Iain Fulton at Amova Asset Management, citing company reports and @business July 2026 data. RPOs are contracted future revenues for services not yet delivered, so they represent hard commitments already on the books, not projections.
The chart breaks it down: Amazon is at USD 627.0bn, Microsoft at USD 519.5bn, and Alphabet at USD 364.0bn. All three lines show steep upward curves from 2022 onward.
The issue Fulton flags is that OpenAI and Anthropic account for a significant portion of that expected demand. Those LLM companies are not the hyperscalers, they do not generate the same cash flow, and they will need significant financing and substantial revenue growth over the next few years to honor those commitments. @GoldmanSachs research, cited in the note, puts the revenue requirement at USD 1 trillion, which Fulton calls an extremely ambitious assumption.
The practical implication: the RPOs are helping underpin earnings expectations across the AI hardware supply chain right now. If the LLM companies cannot fund themselves adequately heading into 2027 and 2028, the credit-risk transmission runs straight into the hyperscaler revenue line, and from there into the supply chain multiples built on those RPO-backed earnings.
The remaining performance obligations (RPOs) across Alphabet, Amazon and Microsoft now stand at more than USD 2 trillion, according to Iain Fulton at Amova Asset Management, citing company reports and @business July 2026 data. RPOs are contracted future revenues for services not yet delivered, so they represent hard commitments already on the books, not projections.
The chart breaks it down: Amazon is at USD 627.0bn, Microsoft at USD 519.5bn, and Alphabet at USD 364.0bn. All three lines show steep upward curves from 2022 onward.
The issue Fulton flags is that OpenAI and Anthropic account for a significant portion of that expected demand. Those LLM companies are not the hyperscalers, they do not generate the same cash flow, and they will need significant financing and substantial revenue growth over the next few years to honor those commitments. @GoldmanSachs research, cited in the note, puts the revenue requirement at USD 1 trillion, which Fulton calls an extremely ambitious assumption.
The practical implication: the RPOs are helping underpin earnings expectations across the AI hardware supply chain right now. If the LLM companies cannot fund themselves adequately heading into 2027 and 2028, the credit-risk transmission runs straight into the hyperscaler revenue line, and from there into the supply chain multiples built on those RPO-backed earnings.