US frees up billions for banks while quietly admitting SVB’s core failure never went away

Washington is in a generous mood with its banks. In March, federal regulators unveiled a sweeping overhaul of capital requirements (the financial cushions that banks must hold to absorb losses in hard times), and the headlines wrote themselves: deregulation, relief, billions freed up for lending and buybacks. The proposal would cut the required capital for the largest Wall Street firms by nearly 5%.

The Federal Reserve estimated that roughly $20 billion in capital could be released for the eight largest banks alone. Former Fed Vice Chair for Supervision Michael Barr put the figure even higher, warning the total could reach $60 billion once all related changes were factored in.

Why this matters: Bank stability depends less on reported capital and more on what markets believe is actually there. If unrealized losses are still sitting on balance sheets, confidence can break faster than regulation can react, turning a technical accounting issue into a liquidity crisis.

But something unexpected surfaces when you read the fine print. Regulators carved out one specific exception: certain large regional banks would have to begin accounting for unrealized losses on their books, a change directly tied to the collapse of Silicon Valley Bank in 2023. That provision, largely overlooked in coverage of the broader rollback, amounts to a regulatory admission.

To understand why, you need to understand what an “unrealized loss” actually is for banks. Imagine you buy a ten-year government bond for $100. Interest rates then rise sharply, new bonds now pay more, making yours less attractive as its market value drops to, say, $80.

Even though you sold nothing and lost no cash, this means that you're now sitting on a $20 loss, unrealized and invisible to most financial scorecards.

For years, midsize banks were allowed to exclude those paper losses from the capital figures they reported to regulators, as though the gap between market value and book value didn't exist.

How Silicon Valley Bank's unrealized losses triggered a bank run in 2023

Silicon Valley Bank's collapse resulted from something far more mundane than fraud or reckless lending: a portfolio of perfectly legal long-term bond investments that shed much of their value as interest rates climbed.

We began seeing the first signs of a crisis in early March 2023, when SVB announced a $1.8 billion loss on the sale of securities, a direct consequence of those unrealized losses, alongside a plan to raise $2 billion in fresh capital.

Shares fell 60% the following day as uninsured depositors began withdrawing their assets en masse; by that evening, $42 billion had left the bank, with another $100 billion staged for withdrawal by morning.

Nearly 30% of its deposits evaporated in a matter of hours. SVB was killed by panic, and the panic was caused by the losses that had been there for quite a while, suddenly becoming visible.

The bank's capital looked substantially more adequate than it was, given that almost none of its supervisors, depositors, or investors could gauge the true size of the unrealized securities losses.

Under the rules then in place, SVB had exercised a legal and widely available option, simply opting out of including those losses in its reported capital figures, a decision that turned out to be catastrophic.

Banks that were required to reflect unrealized losses in regulatory capital, meanwhile, managed their interest rate risk considerably more carefully. The lesson of SVB is that hiding losses of this magnitude guarantees that no one will act until it's too late.

Why the new bank capital rules still require regional banks to report unrealized losses

Which brings us back to the current proposal. The change requiring large regional banks to account for unrealized losses will increase their capital requirements by 3.1%, although their total capital is still expected to fall by 5.2% when all pending changes are considered.

Banks with assets below $100 billion face no such requirement, and their capital is projected to fall even further. The message we get from this is clear: the problem was real, and it was real at a specific scale. The carve-out is Washington saying, in its characteristically bloodless bureaucratic language, that SVB's collapse was due to bad regulation.

Barr, who left his vice chair role earlier this year rather than face removal by the Trump administration but retained his seat on the Fed board, has been vocal about his unease with this. In a formal dissent, he warned that capital requirements are being significantly reduced, that liquidity requirements could also be reduced, that Federal Reserve supervisory staff have been cut by over 30%, and that banking is built on trust.

That final phrase deserves attention. A bank can survive deteriorating accounting right up until the moment the people whose money sits inside it stop believing it.

Supporters of the broader rewrite have a reasonable case. The original 2023 Basel proposal was widely seen as overcalibrated, a blunt instrument that pushes risk out of the regulated system into the shadows instead of actually reducing it. Fed Governor Michelle Bowman said that capital will remain robust and that the new framework now better aligns with requirements and actual risk.

But the unrealized-loss carve-out survives even inside the loosened framework. If the problem were truly solved, if duration risk and depositor confidence were no longer the market's concerns, there would be no reason to keep the provision. Regulators don't impose expensive requirements out of nostalgia.

The temptation is to see the new proposal as straightforward deregulation. But the more accurate interpretation is also the more interesting one. Even as Washington hands banks relief, it's quietly preserving a single hard lesson from SVB: that when rates jump and losses pile up, what a bank is actually sitting on still matters, whether the rules say so or not.

The post US frees up billions for banks while quietly admitting SVB’s core failure never went away appeared first on CryptoSlate.