How Much to Hold and Where to Park Idle Startup Cash

Where to park idle startup cash depends less on yield than on the exit. On 2 October 2026, Treasury bills from 4 to 52 weeks yielded 3.95% to 4.46%; safe instruments differ in how they return cash on a bad day. Hold runway through the next milestone and the raise after it, and fence the surplus.

Deciding where to park idle startup cash starts with sizing. A round holds two kinds of cash: runway up to the next milestone, which a startup should hold in quantity, and a surplus beyond it, whose cost lies in how it gets spent. Both then need homes, and the eight on offer look alike in a yield table but part ways on the day a funding market freezes, which is the day a reserve exists for.

How much cash should a startup hold?

Enough to reach the next milestone at a chosen probability and then raise on it; only cash beyond that line is surplus. By public-company ratios that is a large balance, and research on listed firms suggests it should be. Tim Opler, Lee Pinkowitz, René Stulz and Rohan Williamson found that US firms with strong growth opportunities and “riskier cash flows hold relatively high ratios of cash to total assets” (Opler et al., 1999). Firms with the best access to capital markets, by contrast, “tend to hold lower ratios of cash to total assets”. A venture-backed startup sits at the far end of both measures, so the direction of the finding carries over even where its size may not.

The precaution pays when markets shut. After the subprime crisis broke in 2007, corporate investment fell most at firms “that have low cash reserves or high net short-term debt” (Duchin, Ozbas & Sensoy, 2010). Venture markets shut too. Crunchbase counted $285 billion of global startup investment in 2023, 38% less than the $462 billion of 2022, with US funding down 37% to $138 billion (Crunchbase News, 2024).

A round is sized for a stage, so a startup’s line falls at its next milestone plus the months needed to raise on it. Whatever the round holds beyond that has no milestone attached, and the formula counts it in months of burn.

Surplus (months) = cash ÷ monthly net burn − months to the next milestone − months to raise

The startup in the table is hypothetical. It burns a net $0.4M a month, needs an average of 18 months to reach its next milestone (standard deviation 6) and another 6 months to raise on it.

How much of a $20M round a hypothetical startup’s next milestone needs, at four levels of confidence

Probability the round lasts	Months of burn funded	Round size ($M)	Surplus in a $20M round
50%	24.0	9.6	26.0 months ($10.4M)
80%	29.0	11.6	21.0 months ($8.4M)
90%	31.7	12.7	18.3 months ($7.3M)
95%	33.9	13.5	16.1 months ($6.5M)

Source: hypothetical company, computed by CX Cash on the assumption that months to the milestone are normally distributed. Months funded = 18 + z × 6 + 6, where z is the standard normal value at the target probability.

Even at 95% coverage the milestone needs $13.5M, so a $20M round carries 16 months of burn that no milestone has asked for (our arithmetic). Moving from 80% to 95% adds about 4.8 months, because the normal curve thins in the tail; a timeline that skews late argues for the higher rows.

Does a cash surplus cost anything while it sits in bills?

Little in yield; the larger cost is in how a surplus gets spent. Aswath Damodaran of NYU Stern writes that “an investment in treasury bills is a zero net present value investment, earning exactly what it needs to earn, and thus has no effect on value”. His charge of waste is narrower: “cash left in a checking account, earning no interest, is wasting cash” (Damodaran, 2005). At the 13-week bill’s 4.11% on 2 October 2026, the table’s $6.5M surplus would earn about $267k a year, two-thirds of one month’s burn.

Spending is where a surplus loses value. In Jarrad Harford’s study of US acquisitions, cash-rich bidders “destroy seven cents in value for every excess dollar of cash reserves held” (Harford, 1999). A startup seldom buys companies; its channel is hiring. Paul Graham told founders in 2013 that “The more you raise, the more you spend”, because “the main way to spend money is people” (Graham, 2013). Marc Andreessen urged founders in 2007 to raise “as much as you can”, since “the funding window may not be open when you need more money”. In the same post he told them to “bank as much of the money as possible, and save it for a rainy day” (Andreessen, 2007). Both are practitioners’ observations rather than data, and they point to the same fence: a separate account, outside the operating budget, drawn only on a board decision.

Does yield matter when choosing where to park idle startup cash?

Less than the exit does. The table below sets the eight homes side by side, and the consolidated cash-position tracker further down has a column for each one’s stressed-day exit.

Eight places to park a round: close on yield, far apart on the exit

Where the cash sits	What the company owns	Normal-day exit	Bad-day exit	Dated yield
Bank deposit	A claim on one bank	Same day	Insured to $250,000; the rest is a receivership claim	Set by the bank
Repo sweep	Treasury collateral, owned or pledged	Per contract	FDIC recognises the collateral interest	SOFR 3.87% (1 Oct 2026)
Eurodollar sweep	A foreign-branch deposit	Per contract	Unsecured general creditor	Set by the bank
Government money fund	Fund shares, 99.5% in cash, government debt and repo	On demand, at $1	No mandatory fee	4-week bill 3.95% (2 Oct 2026), a proxy
Institutional prime fund	Fund shares priced to four decimals	On demand	Mandatory fee above 5% daily outflows; 1% by default	Financial paper 4.16% (1 Oct 2026)
Treasury bill at a broker	A US government obligation	Sale settles next business day	Sold at market; investors buy bills in a run	3.95% to 4.46% (2 Oct 2026)
Treasury bill at TreasuryDirect	The same obligation	At maturity	No sale inside; 45-day hold before transfer	As above
Brokered CD	A share of an insured bank deposit	At maturity	Sold at a discount if rates have risen	None published by the Fed

Source: 17 CFR 270.2a-7; 12 CFR 360.8; TreasuryDirect; SEC; New York Fed; US Treasury; Fed H.15. Bill yields coupon-equivalent; none is a current rate.

The spread a founder can choose is narrow. On 2 October 2026 the Treasury quoted the 4-week bill at 3.95% and the 52-week bill at 4.46% on a coupon-equivalent basis, 0.51 points apart (US Treasury). The day before, the Fed’s H.15 release put 3-month AA financial commercial paper, a staple of prime money funds, at 4.16% against 4.00% for the 3-month bill, both on a discount basis (Federal Reserve, H.15). Overnight Treasury repo (SOFR) cleared at 3.87% on $3,067 billion of trades (New York Fed).

The level of rates, which nobody chooses, moves income far more. The same 3-month bill yielded 0.02% on 1 June 2021 and 5.35% on 2 October 2023: about $2,000 against $535,000 a year on a $10M reserve. What a founder controls is how each dollar comes back when funding markets are stressed: whether it returns, how many days that takes, and what it costs. On such a day, government funds, maturing bills and insured deposits pay at par; bills sold early and brokered CDs pay a market price; prime funds can charge a fee, and Eurodollar sweeps rank as unsecured claims.

Which money market funds can a company buy?

Government and institutional prime funds; retail funds are closed to it. Rule 2a-7 defines a retail fund as one whose policies are designed “to limit all beneficial owners of the fund to natural persons” (17 CFR 270.2a-7). The SEC’s 2023 release adds that institutional funds “can be held by a wider range of investors, such as corporations, small businesses, and retirement plans” (SEC, 2023). A government fund holds at least 99.5% of its assets in cash, government securities or fully collateralised repo; an institutional prime fund can also buy commercial paper and bank CDs.

Their exits differ. Since October 2024, an institutional prime fund whose daily net redemptions exceed 5% of net assets must charge a liquidity fee on every share redeemed that day. The fee is the fund’s good-faith estimate of what selling a slice of its portfolio would cost, or 1% if it cannot make one. A board may add a discretionary fee of up to 2%. Government funds face neither rule unless they opt in. In the SEC’s summary, they “generally offer greater safety of principal but historically have paid lower yields”.

The yield gap is small. Commercial paper’s premium over the 3-month bill, a rough proxy for what a prime fund earns over a government fund, had a median of 0.09 points over the 100 days of 2026 with both rates published. A 1% default fee on one stressed redemption therefore costs about 11 years of 2026’s median premium (our arithmetic): on a $5M redemption, $50,000 against $4,500 a year of extra yield.

Have investors run from prime money funds before?

The fee matters only if the bad day comes, and it has come twice. In the week of 15 September 2008, investors withdrew about $300 billion, 14% of the assets, from taxable prime funds. The Reserve Primary Fund, holding $785 million of Lehman Brothers debt, suspended redemptions as of 17 September (SEC, 2010). Its investors had more than $0.98 on the dollar back by early 2010; the wait was the cost.

March 2020 repeated the pattern with a different trigger. During the two weeks from 11 to 24 March, “publicly offered institutional prime funds had a 30% redemption rate (about $100 billion)”, while government funds took in a record $838 billion that month (SEC, 2023). The rules then in force made the run worse. Since 2016 a prime fund had been allowed to impose fees or gates once its weekly liquid assets fell below 30%. No fund did, yet a peer-reviewed study found that “such liquidity restrictions exacerbated the run on prime MMFs during the COVID-19 crisis” (Li, Li, Macchiavelli and Zhou, 2021). The 2023 reform removed gates and replaced the threshold with the fee. New York Fed economists note that in money funds “run risk has historically materialized only in prime funds” (Anadu et al., 2023).

H.15 shows the price of that risk. The premium of commercial paper over the 3-month bill rose from 0.21 points on 2 March 2020 to 2.57 on 25 March, when investors bought bills until they yielded −0.04%. On 17 September 2008 it reached 2.63. In both episodes the funding freeze and the prime-fund run came together. Nothing guarantees they will coincide again, but if they do, a startup will need its reserve just when a prime fund’s exit is at its worst and a bill’s at its best.

Should a startup buy Treasury bills at TreasuryDirect or through a broker?

Through a broker, because a bill held at TreasuryDirect cannot be sold there. Bills come in terms of 4, 6, 8, 13, 17, 26 and 52 weeks, from $100 (TreasuryDirect), and corporations may open a TreasuryDirect account (TreasuryDirect). The catch is the exit: a bill held there can be sold only through a bank, broker or dealer, after a 45-day hold. In TreasuryDirect’s words, “This means you can’t sell or transfer a 4-week bill from TreasuryDirect because it matures in less than 45 days” (TreasuryDirect). The government’s own platform is the one window where the Treasury’s shortest bill cannot be sold.

At a broker the exit is a sale that settles the next business day, by market practice (SEC, 2023). The price risk is small: if yields rise a full point, a bill with 91 days left loses about 0.25% of face value (our arithmetic). The tax edge is narrower than TreasuryDirect’s summary, “No state or local taxes”, suggests. The federal exemption, 31 U.S.C. § 3124, excepts “a nondiscriminatory franchise tax or another nonproperty tax instead of a franchise tax, imposed on a corporation” (31 U.S.C. § 3124). California’s franchise tax uses the exception and takes about 0.39 points of a 4.46% bill yield at its 8.84% rate (Franchise Tax Board, 2025).

Brokered CDs carry deposit insurance up to $250,000 per depositor per bank, but the SEC warns that “if you buy a brokered CD and need to get your money back early, you may lose some of your principal” (SEC).

What does a bank sweep or a treasury account hold?

Whatever the contract bought, which can differ from what the balance on the screen suggests. The FDIC’s rule for failed banks treats each kind of overnight sweep by what the transfer bought (12 CFR 360.8). In a repo sweep where the customer owns the Treasury collateral or holds a perfected security interest in it, the FDIC recognises that interest. Funds swept to a Eurodollar account at a foreign branch become “an unsecured general creditor’s claim against the receivership” if the bank fails that day. Banks must disclose in writing whether swept funds are deposits.

Mercury, which sells a treasury account, advises founders to “find something you can just set and forget so that you have more time to build and run your company” (Mercury). For a seed company with a year of runway, that advice holds, provided the fund inside is a government fund. A treasury account is a wrapper, and the fund it holds may be government, prime or short-term bond.

Where should payroll, the reserve and the surplus sit?

Keep the next two months of burn in insured deposits, and hold the reserve in a government money fund and in Treasury bills bought through a broker, their maturities set against the burn plan. The fenced surplus can sit in the same instruments, in an account of its own. We worked the ladder for a hypothetical Series A startup holding $12.0M on 2 October 2026, with 8- to 52-week bills matched to a 24-month burn plan. Across flat, falling and rising rate paths, its first-year interest varied by about $17k, against $132k for a government fund alone, and it needed no early sales. A burn plan is a forecast, so a rung that matures before the cash is needed simply waits in the fund.

Notes and sources

• Anadu, K. E. et al. (2023). Runs and Flights to Safety: Are Stablecoins the New Money Market Funds? Federal Reserve Bank of New York Staff Reports, no. 1073.
• Andreessen, M. (2007). How much funding is too little? Too much? The Pmarca Guide to Startups, part 6 (collected at pmarchive.com).
• Board of Governors of the Federal Reserve System. H.15 Selected Interest Rates, release of 2 October 2026 and historical data package.
• California Franchise Tax Board (2025). Form 100 Booklet: California Corporation Franchise or Income Tax Return.
• Code of Federal Regulations: 12 CFR 360.8; 17 CFR 240.15c6-1; 17 CFR 270.2a-7.
• Crunchbase News (2024). Global Startup Funding In 2023 Clocks In At Lowest Level In 5 Years, 4 January 2024.
• Damodaran, A. (2005). Dealing with Cash, Cross Holdings and Other Non-Operating Assets: Approaches and Implications. Stern School of Business, New York University.
• Duchin, R., Ozbas, O. and Sensoy, B. A. (2010). Costly external finance, corporate investment, and the subprime mortgage credit crisis. Journal of Financial Economics 97(3), 418–435.
• Federal Reserve Bank of New York. Secured Overnight Financing Rate, 1 October 2026.
• Graham, P. (2013). How to Raise Money. paulgraham.com.
• Harford, J. (1999). Corporate cash reserves and acquisitions. Journal of Finance 54(6), 1969–1997.
• Li, L., Li, Y., Macchiavelli, M. and Zhou, X. (2021). Liquidity restrictions, runs, and central bank interventions: evidence from money market funds. Review of Financial Studies.
• Mercury. How to adopt an effective cash management strategy for your startup.
• Opler, T., Pinkowitz, L., Stulz, R. and Williamson, R. (1999). The determinants and implications of corporate cash holdings. Journal of Financial Economics 52(1), 3–46 (NBER Working Paper 6234).
• SEC (2010). Money Market Fund Reform, 75 FR 10060; (2014) Money Market Fund Reform, 79 FR 47736; (2023) Shortening the Securities Transaction Settlement Cycle, 88 FR 13872; (2023) Money Market Fund Reforms, 88 FR 51404.
• SEC Office of Investor Education and Advocacy. High-Yield CDs: Protect Your Money by Checking the Fine Print (page dated 20 September 2022).
• TreasuryDirect. Treasury Bills; Buying a Treasury Marketable Security; Selling Marketable Securities.
• US Code, 31 U.S.C. § 3124.
• US Department of the Treasury. Daily Treasury Bill Rates, 2 October 2026.

Rates come from the Treasury, the Federal Reserve’s H.15 release and the New York Fed, each dated to the day it was quoted; none should be read as a current rate. Spreads, medians, interest sums and the milestone table carry no citation where they are calculated from those rates and the other cited figures, and the ladder figures come from a script, ladder.py, kept with the research pack. The cash-holding studies cover listed companies, so applying them to startups is our reading. Both worked examples, the milestone startup and the Series A ladder, are hypothetical.