Your operating account and your reserve account are one account, and the FDIC says so in a sentence we had never read

We ran two accounts at one bank for eleven months. Operating on one, reserve on the other, about 180 thousand dollars parked in the second one where nobody touched it. I told two other founders to copy the arrangement. Two buckets, neither of them near the 250 thousand dollar insurance line, and I thought that was the whole trick.

I was wrong, and it was bad advice I repeated for the best part of a year, which I only found out because the sentence that disproves the whole arrangement sits in a free FDIC brochure I had never bothered to open, buried in a document written for consumers rather than for companies like ours. Nobody had ever pointed me at it.

I went looking for it after a founder asked me a version of the question in the headline, and I could not answer it with anything except my own arrangement. So how much cash should a startup keep in one bank? Working that out took an afternoon with the brochure and the failed bank list open side by side, and the answer turned out to have nothing to do with counting accounts.

The sentence that ended our arrangement

The brochure is called Your Insured Deposits, the section on companies runs to four paragraphs that nobody in our company had ever read end to end, and the second of those paragraphs does all the damage on its own by saying: “All deposits owned by a corporation, partnership, or unincorporated association at the same bank are combined and insured up to $250,000.”

The next line removes the escape route. “Accounts owned by the same corporation, partnership, or unincorporated association but designated for different purposes are not separately insured.” Then it uses our exact setup as its worked example. A company with an operating account and a reserve account at one bank gets both added together, insured to 250 thousand in total.

So our two accounts were one account. Splitting the money improved the bookkeeping, bought nothing else at all, left around 180 thousand covered and the rest exposed, and put us in exactly the position we would have occupied with a single account and half the paperwork.

Three things that do not help, and one that does

Once accounts stop working, you go hunting for something else that might. I tried three ideas and the brochure kills all of them.

Branches do not count for anything here. The brochure says that funds deposited in separate branches of the same insured bank “are not separately insured”, so a company banking in two towns with one charter behind both of them has exactly one limit, which surprised the founder who told me his two locations were covered twice.

Signatories do not count either. The number of partners, members, stockholders or account signatories “does not affect insurance coverage”. I had assumed adding my co-founder did something. It adds a co-founder and nothing else.

Divisions do not count either. A unit that is not separately incorporated gets folded back in with everything else the company holds there.

What counts is a separate charter. The limit is 250 thousand dollars per depositor, per insured bank, per ownership category, and a deposit at bank A is insured independently of one at bank B. That is the entire mechanism, and it turns on charters rather than on accounts, branches or the names written on the signature card.

The trap that catches you before you incorporate

There is a worse version for anyone still trading as a sole proprietor, and I nearly missed it because the paragraph sits under a heading about personal accounts. The FDIC insures deposits owned by a sole proprietorship as a single account of the business owner. The brochure treats that money as yours rather than as the company’s.

The business balance stacks on top of the founder’s personal savings at that bank and the two of them share one limit. The brochure works it through: four single accounts including one in the name of a sole proprietorship, 260 thousand dollars in total, 250 insured, 10 thousand not. Incorporating moves the money into its own ownership category with its own limit. I still find it odd that nobody lists that next to the tax reasons for incorporating, and I keep thinking about the founders who banked for two years before filing.

What actually happens on the Monday after

My picture of a bank failing was vague and slow, everybody waiting a long time for everything. The reality splits hard at the insurance line and the two halves behave nothing alike.

The insured half moves fast. The FDIC says it historically pays within a few days of a closing and “usually the next business day”, doing it either by opening an equivalent account for you at another insured bank or by sending a cheque, which means the covered part of your balance is a nuisance for a week rather than a problem for a year.

The uninsured half becomes a claim. You are a creditor of the receivership now, paid as assets are sold, and the FDIC says it “can take several years to sell off the assets of a failed bank”. Payments arrive on a pro rata cents on the dollar basis. Nobody promises the full amount and nobody promises a date.

Hold those two sentences next to a payroll run. Money under the line is available Tuesday. Money over it is a certificate and a wait measured in years. That gap is what should set your number rather than the odds of a failure, because the odds are genuinely low and the calendar is genuinely brutal, and a company that can cover eight weeks of payroll from insured balances survives a closing as an administrative nuisance while a company that cannot survives it as an existential one.

Fourteen banks since March 2023, five of them this year

I assumed 2023 was the anomaly and the list went quiet afterwards. The list never went quiet. The FDIC failed bank list, which I read on 31 August 2026, shows 14 closings since Silicon Valley Bank on 10 March 2023.

Signature followed on 12 March 2023 and First Republic on 1 May, after which the pace slowed without ever stopping: Heartland Tri-State in July 2023, Citizens Bank of Sac City that November, Republic First in April 2024, The First National Bank of Lindsay that October, Pulaski Savings in January 2025, The Santa Anna National Bank that June.

This year alone has produced five. Metropolitan Capital Bank and Trust in Chicago on 30 January, Community Bank and Trust of West Georgia on 1 May, Kentland Federal Savings and Loan on 10 July, Small Business Bank of Lenexa, Kansas on 17 July, and Tioga-Franklin Savings Bank in Philadelphia on 21 August, ten days before I wrote this.

Not one of those five made the front page. That is the part I would underline, because a closing needs no press coverage whatsoever to freeze your operating balance on a Tuesday morning, and the absence of a headline is what makes founders assume the problem ended in 2023. One of them is literally called Small Business Bank.

The mechanism that gets you past 250 without more logins

Opening one account per 250 thousand dollars is a real answer and a bad one. Four banks means four onboardings, four sets of credentials, four reconciliations a month. A seed stage company with one finance hire will not keep that up past the first quarter, and I have watched two try.

The alternative is a reciprocal deposit network. Your bank keeps the relationship and places your balance across other member banks in slices under the limit, so one agreement covers a multiple of 250 thousand. IntraFi is the best known operator and plenty of mid sized banks resell it, which means the question to put to your own bank is not whether such a thing exists but whether that particular bank is already a member of a network, because joining one is a decision taken at the bank rather than by the customer asking for it.

This corner of the rulebook moved four days before I wrote this. On 27 August 2026 the FDIC board approved an interim final rule implementing section 902 of the 21st Century ROAD to Housing Act, which rewrote the statutory framework for reciprocal deposits. The rule raises how much an “agent institution” may keep out of the brokered deposit bucket, on a new tiered calculation based on liabilities, up to a ceiling of 30 billion dollars. It widens which banks qualify as agent institutions. Comments run for 30 days from publication in the Federal Register.

None of that changes your coverage arithmetic by a dollar. It changes how easy the product is to buy. I suspect that matters more, because for a company with one finance hire the deciding factor is never the arithmetic, it is whether anyone has time to open the thing.

So what number do I use now

The question has an answer and it is not 250 thousand dollars.

I size the operating balance at one payroll run plus one month of payables and keep that at the primary bank whatever it comes to. That money has to move on schedule. No coverage rule is worth a missed salary, and I would rather carry uninsured cash for a week than explain a late payday. Everything above that goes where the wait does not apply: a reciprocal network, a second charter, or Treasuries.

Take a company holding 3.2 million dollars with a 260 thousand dollar monthly burn. One payroll run at 140 thousand plus a month of payables at 120 thousand gives an operating float of 260 thousand, which stays at the primary bank and sits 10 thousand over the insured line on its own. That leaves 2.94 million to place, which is either twelve slices of 250 thousand across twelve charters, or one reciprocal agreement, and choosing between those two is a staffing decision rather than a finance one.

The honest framing is that concentration is a timing problem dressed as a credit problem. Your bank probably will not fail this quarter. If it does, the cost is not a haircut, it is a calendar.

Questions we get

Five arrive every time. I have answered them in enough founder chats that the wording below is close to what I paste. They are shaped by what founders actually get wrong rather than by what the brochure chooses to emphasise.

Is there a treasury policy template worth copying? The templates circulating in founder groups are fine on structure and wrong on this exact point, because they set limits per account rather than per bank. A policy saying “no more than 250 thousand per account” describes protection it does not provide. Write yours per charter and the rest can stay as it is.

How should deposit concentration limits be written at our size? As a share of total cash at any single insured institution, counting every account there as one number, because that is how the FDIC counts it. We use a ceiling per bank and a floor of one payroll plus a month of payables at the primary, and the reason the pair works better than a single rule is that the ceiling protects you from the receivership calendar while the floor protects you from the much more common failure of being so careful about coverage that payroll goes out late, which is the one mistake a board actually notices. Those two lines do most of the work.

Do counterparty limits for startups need to cover anything besides banks? Yes, and the analogy breaks here. A money market fund or a Treasury holding is not a deposit and has no 250 thousand line at all, so one policy has to describe two different risks. Treating a brokerage sweep as an insured deposit is the mistake that usually follows this one.

How does a reciprocal deposit network actually work? Your bank places your balance in sub 250 thousand pieces across member banks and takes matching deposits back, so you keep one relationship and one statement. The FDIC rewrote the framework on 27 August 2026, raising the tiered exclusion ceiling to 30 billion dollars for qualifying institutions, and the practical effect for a company your size is not the ceiling itself, which no seed stage balance will ever approach, but the fact that a wider set of mid sized banks now qualifies to offer the product at all. Ask your own bank whether it does.

How many banks should a startup use? Two relationships rather than four, unless the balance genuinely forces more. One primary for operations, one alternative fully opened and tested instead of merely applied for, and coverage above that solved with a network or Treasuries rather than another login.

A short digression about the industry

A piece built on a failure list reads darker than the data deserves. The FDIC Quarterly Banking Profile for the second quarter of 2026, published 25 August, has insured institutions returning 1.37 per cent on assets and 90.1 billion dollars of aggregate net income, up 9.7 billion or 12.0 per cent on the quarter. That is a profitable industry, not a queue of candidates. Fourteen closings in three and a half years is small against thousands of charters, and the reason to act is the length of a receivership rather than the odds of entering one. Anyway, back to the number.

What I could not establish

What share of startup deposits sits above the line. Aggregates for insured and uninsured balances are published for the industry, and I have not found a breakdown that isolates venture backed companies, so I cannot tell you whether our 180 thousand dollar mistake was typical or careless.

Whether the March 2023 rescue repeats. Depositors at Silicon Valley Bank and Signature were made whole above the limit under a systemic risk exception, and I have watched founders quietly treat that as precedent. It is decided case by case. I do not know of any FDIC material that describes it as available on request, and I would not build a treasury policy on it.

What these networks cost the end customer. Pricing lives inside individual bank agreements and neither figure is published. My guess is that the spread is small enough that most founders would accept it if anyone quoted it, but that is a guess, and the two banks I asked would not put a rate in writing.