BlockBeats News, June 21st - JPMorgan Chase recently released a research report stating that the forecast for AI total capital expenditure by 2030 has been raised from $51 trillion to $55 trillion, with accompanying debt financing scaled up to $41 trillion. The report points out that the surge in AI demand and the ongoing scarcity of computing power have jointly driven the acceleration of this round of expenditure, with Google's monthly AI token processing volume reaching 3.2 trillion times. Meanwhile, companies such as Microsoft and Uber have exhausted their annual AI budgets within months.

Regarding mega-scale cloud providers, the four major American supercomputing giants (Google, Amazon, Microsoft, Meta) are expected to have a combined capital expenditure guideline of around $700 to $725 billion in 2026, a year-on-year increase of about 75%, and is expected to exceed $1.1 trillion in 2027. JPMorgan Chase predicts that their combined operating cash flow will exceed $900 billion in 2027, but it will still be insufficient to cover the massive expenditure plan, making debt and equity financing a norm.

In terms of chip beneficiaries, JPMorgan Chase predicts that Broadcom will achieve over $150 billion in AI-related revenue (including ASIC/XPU and AI networks) by 2027. The management has disclosed that the 2027 order backlog exceeds $100 billion, far exceeding the bank's expected approximately $60 billion in 2026, which is seen as a conservative estimate. Broadcom's AI chip financing platform (AI XPV) jointly formed with Apollo and Blackstone is also viewed in the report as a landmark structural innovation in the GPU financing field.

Regarding financing structure, the top five supercomputing companies have completed approximately $240 billion in external financing since the beginning of the year, with the high-grade bond market still being the main channel. It is estimated that the issuance of AI-related high-grade bonds in the next five years will reach $2.1 trillion.

On the power front, JPMorgan Chase has raised its data center capacity growth forecast from 122GW to 138GW, but electricity remains the most critical bottleneck, with supercomputing manufacturers actively exploring alternative solutions such as self-built power sources.