AI boom drives record bank fees in the US

US banks are booking record investment banking fees as capital pours into the infrastructure needed to support artificial intelligence, from data centers and power to memory and compute capacity.

Goldman Sachs said it generated $3.4 billion in investment banking fees in Q2 2026, up 55% from a year earlier and the highest total in the bank’s history. CEO David Solomon described the moment as a “supercycle” of AI capital spending, saying companies across regions and industries are looking for funding to build the computing infrastructure required for AI.

Within Goldman’s results, securities offerings grew especially fast. Revenue from equity underwriting jumped 130% to $985 million, while debt underwriting rose 75% to a record $1.03 billion. Similar demand also led 8 banks to participate in financing for SoftBank’s $40 billion deal tied to OpenAI.

Other major Wall Street banks reported the same pattern. JPMorgan posted $3.3 billion in investment banking fees, up 30% year over year and its best result since 2021. Morgan Stanley increased the figure 58% to $2.44 billion, while Bank of America rose 50% to $2.14 billion.

Ted Pick, CEO of Morgan Stanley, called the current phase the start of a large investment cycle. According to the bank’s research, spending on data center construction could reach about $850 billion in 2026, rise to $1.3 trillion in 2027, and hit $1.5 trillion in 2028. He also cautioned that the cycle is still in its early stages and its eventual trajectory remains uncertain.

Recommended reading

Netflix slips 9% after soft Q3 outlook despite Q2 beat

Bank executives say the surge is not limited to AI companies themselves, but extends across the surrounding stack, including power supply, memory, compute, and data center construction. Citigroup CEO Jane Fraser said AI has become a central topic in client discussions, with demand rising across the chain from technology to energy.

Data center power limits are starting to matter

Banks are also becoming more selective. Jeremy Barnum, CFO of JPMorgan, said the bank passed on some data center deals because of concerns about power availability and the future tenants for those facilities. None of the biggest banks has yet disclosed the precise volume of its AI infrastructure lending.

Across the five largest US banks, total net income in Q2 came to about $49 billion, roughly 39% higher than a year earlier. But most of that growth came from capital raising and securities issuance, while traditional advisory work expanded much more slowly.

Solomon warned that the current cycle will not last forever.

“The market will eventually face a reset and a decline in activity, after which another phase of growth is possible.”

For banks, one of the biggest open questions is now whether data center economics — and the energy grid behind them — can sustain the next wave of AI expansion.