Mega-Deals, AI Capex, and the Limits of Bank Balance Sheets

According to Bloomberg, banks entered 2026 working on roughly $100 billion of leveraged buyout debt, with approximately three-quarters already underwritten and in the process of being sold to investors. After several muted years following 2022's market dislocation, the leveraged finance departments are once again operating at scale.

The take-private of Electronic Arts is emblematic. The $55 billion transaction is reportedly backed by roughly $36 billion of equity and $20 billion of committed debt from JPMorgan Chase, with approximately $18 billion expected to be funded at closing. JPMorgan has already begun selling down portions of the financing, including a $3 billion Term Loan A, ahead of a broader syndication of dual-currency Term Loan B and high yield bonds. A $20 billion single-bank commitment still represents substantial underwriting exposure. That debt must ultimately be placed into leveraged loan and/or high yield markets at levels that clear institutional demand.

The difference this cycle is the competitive backdrop. Private credit providers have spent the past several years marketing themselves as a faster, more certain alternative to the syndicated loan market — essentially a one-stop solution with less syndication risk. In response, large banks appear increasingly willing to demonstrate that the traditional bank-led model can provide the same certainty at scale. Bloomberg notes banks have been ultra-competitive on both pricing and documentation to win mandates, effectively outmaneuvering private credit on most large deals. Fully underwritten bridges of $20 to $60 billion are as much strategic statements as financing commitments.

However, financing certainty to the issuer translates into warehousing risk for the bank. When a bank commits the full bridge, it accepts the timing risk between signing and syndication. Banks are eager to shift exposure quickly — in some cases selling down debt well before the deal closes — because the margin for error is narrow. If spreads remain tight and investor demand is robust, the model works as intended. But if spreads gap wider, even modestly, the bank must either flex pricing materially or temporarily hold risk on the balance sheet.

The experience of 2022 demonstrated how quickly underwriting economics can deteriorate when sentiment shifts. Several large LBO financings became "hung deals" as rates rose and liquidity dried up, forcing banks to syndicate at discounted levels and absorb losses. The lesson was not about weak credits; it was about the fragility of distribution assumptions.

Layer onto this the AI capex super-cycle. Even if partially funded through operating cash flow, the proposed spending implies sustained and significant debt issuance from the largest issuers in the market. Investment grade investors will have substantial AI-related supply to digest at the same time that M&A-driven bond financings and bridge takeouts are hitting the market. Bloomberg reports bankers are confident markets can absorb even larger deals, including rumored financings north of $25 billion — but confidence does not always translate to capacity.

The key risk, therefore, may not necessarily be recession or credit deterioration, but market congestion. When structural AI issuance, strategic M&A bonds, leveraged loans, and bridge refinancings all compete for investor capital, clearing spreads become more sensitive to marginal demand. In that environment, underwriting assumptions made in a tight-spread world can prove optimistic. A 100-to-150bps widening between commitment and pricing does not require a macro shock; it can result from only a temporary imbalance between supply and demand.