Best Short-Term Investments in 2026: The Grid by Horizon, Net of Taxes and Inflation

Short-term money has one enemy and it is not volatility; volatility at least announces itself. A savings account never prints a red number, which is precisely how it loses: quietly, in real terms, at whatever pace inflation sets. This page grids the short-term universe by horizon, under six months, six to eighteen, eighteen to thirty-six, and prices every vehicle the only way that counts: net of taxes, net of inflation, penalty terms included.

1. The variable that decides is the real rate, not the printed yield

Every short-term product advertises a nominal rate, and no saver spends nominal dollars. What a parking decision actually earns is the printed yield, minus the taxes the wrapper cannot avoid, minus the inflation the period delivers. The first two are known at purchase; the third is the gamble, and it is the entire gamble, because default risk on this shelf rounds to zero. The order of deduction matters as much as the amounts: taxes are proportional to the nominal rate, inflation is not, so a higher print taxed harder can keep less than a lower print taxed lightly, which is the entire section 3.3 in one clause. Short-term cash loses to inflation more often than to markets: that asymmetry, documented at length in cash vehicles measured in real yield, is the organizing fact of the page.

The 2026 configuration sharpens it. Policy sits at 3.50-3.75% after four consecutive holds (FOMC, June 2026), top-of-market savings rates pay about 4.00% against a 0.38% national average (NerdWallet, July 6, 2026), and inflation on the I bond’s semiannual CPI-U base runs at 3.34% annualized. The whole shelf therefore clusters within about a point of inflation: after tax, the sign of the real return is decided by basis points, wrapper rules and exit penalties, the exact matters this grid prices. Nothing on the shelf sits more than about a point from anything else before tax; after tax, vehicles regularly swap places, and a shelf that tight rewards reading fine print over chasing prints. The dominant reading runs the other way, and deserves naming: safe money is assumed to be money that cannot lose. The 2021-2023 record, walked through in section 5, shows the assumption failing at scale precisely when it was most trusted, on the least risky instruments in both currencies. Nothing about that episode was exotic; it was arithmetic executing. A nominal rate below inflation is a slow tax with excellent branding, and no deposit insurance covers it. The after-tax mechanics of the two default vehicles are compared line by line in comparing HYSAs and money market funds after tax.

2. The grid by horizon

The grid replaces the question “which vehicle is superior” with the answerable one: at this horizon, which vehicles are even in the field, and what does each charge to leave early. Three horizons, three different questions. Under six months, the question is availability: which vehicles return the money intact, on demand, with zero exit cost, because the money’s job is to exist on demand. Six to eighteen months, the question becomes locks: terms and penalties start paying, if the horizon is genuinely known. Eighteen to thirty-six months, rate risk enters: the reinvestment question, what rates will be at rollover, starts mattering more than today’s print. Past that point the question stops being one of parking and becomes one of allocation — the strategies that allocate a portfolio across market regimes.

The table’s quiet message is that the field grows with the horizon: at three months the menu has three lines, at three years it has six. That is the page’s title in mechanical form; the simulator in section 4 draws it, with curves that appear along the axis as vehicles become accessible.

The horizon question deserves one honesty check before pricing anything: most short-term money has no fixed horizon at all. An emergency fund’s horizon is “unknown, possibly tomorrow”, which locks it into the top row permanently; only money with a dated purpose, a tax bill, a purchase, a tuition payment, can honestly shop the lower rows. Misclassifying the first kind as the second is how savers end up paying early-withdrawal penalties, and it costs more than any spread on this page. The grid assumes the classification is honest; the penalty schedules of the products below assume, profitably, that it often is not.

3. Vehicle by vehicle, priced at July 2026

Five families cover the retail shelf; a sixth entry covers the design questions the shelf prefers not to raise. Each entry below carries its dated yield, its tax treatment at both levels of government, its exit terms and its regime signature; every figure is replaced at the January revision, and the date is part of the data. What the entries deliberately omit is a score: at the dispersion this shelf shows, a ranking would encode tax assumptions the ranker cannot know about the reader.

3.1 High-yield savings accounts

About 4.00% at the top of the market, 0.38% at the average (NerdWallet, July 6, 2026), variable at the bank’s discretion, FDIC-insured, liquid same-day. Interest is ordinary income at both federal and state level, which is the detail that decides section 3.3’s comparison, and the one detail the rate tables never show. The rate follows policy with a lag and no contract: in a cutting cycle, the printed 4.00% is a snapshot, not a promise. The tenfold gap between top and average is the category’s structural feature rather than an anomaly: it persists because moving an account costs attention and the average holder does not, which makes inertia the highest-yielding product banks offer, to themselves. Both numbers are published; almost nobody reads the second.

3.2 Money market funds

Roughly 3.3% to 3.6% on the large funds in July 2026 (NerdWallet), no FDIC insurance, next-day liquidity, and one tax property savings accounts lack: the share of income earned on Treasury holdings escapes state income tax. Government funds, holding bills and repos, reprice with policy in weeks, which cuts both ways by construction. The family’s one historical blemish, a prime fund breaking its stable dollar in 2008 on failed commercial paper, produced the reforms that now wall government funds off from prime ones; the government variety has never repeated it, which is why it anchors most brokerage sweeps. The practical distinction on this shelf: a fund is where cash waits inside a brokerage; an account is where it waits inside a bank; and the yield gap between the two waiting rooms, run after tax, is smaller and more state-dependent than the stickers imply.

3.3 Treasury bills

The 4, 8, 13 and 26-week ladder prices in the immediate neighborhood of the policy floor of 3.50-3.75% (FOMC, June 2026), with the one-year at 3.75% in late April (auction data via Tipswatch). Two properties earn the family its line: full state and local tax exemption, and a weekly auction calendar that makes a self-rolling ladder trivial to run. The exemption is where rankings invert. A 4.00% savings account and a 3.60% bill, taxed at a 22% federal rate, land at 3.12% and 2.81% respectively for a saver with no state tax; add a state levy above roughly 8% and the order flips, the bill now keeping more than the account. Geography, not yield, decides the podium, which is one more reason this page does not print one. The arithmetic behind the 8% threshold is worth showing once, because it generalizes: the account keeps (1 minus federal minus state) of its rate, the bill keeps (1 minus federal) of its own, and the state rate at which the two nets cross equals the federal-adjusted yield gap divided by the account’s rate. Every saver can run it with their own bracket in one line; the simulator below runs it continuously. Further reading: the breakdown “Choosing investments in the light of the macro cycle”.

3.4 Certificates of deposit

Top one-year CDs pay about 4.40% (Fortune, July 2, 2026), the roughly 40-basis-point premium over top savings accounts being the price of the lock. The product is a rate wager in retail packaging: it wins if variable rates average below the locked print over the term, loses otherwise, and charges an early-withdrawal penalty, typically months of interest, for changing one’s mind. Fully taxable at both levels; the after-tax premium over a T-bill of equal term is therefore thinner than the sticker suggests, and turns negative in high-tax states. The wager framing is not rhetorical: a 4.40% one-year CD against a 4.00% variable account wins exactly when the variable rate averages under 4.40% across the year, a bet on cuts arriving. Four consecutive Fed holds make the bet’s terms unusually legible in mid-2026, which is different from settling it; the tool below prices the branches without picking. Laddering CDs across staggered terms softens the wager the way bill ladders do, at the cost of averaging away the premium that motivated the lock.

3.5 I bonds

The May-October 2026 issue pays a composite 4.26%: 0.90% fixed for the bond’s life plus a 3.34% annualized inflation adjustment resetting semiannually (TreasuryDirect, May 1, 2026). On this shelf it is the only instrument whose real return is set by contract rather than by luck. The constraints are the price and they bind hardest at short horizons: nothing exits before month 12, three months of interest are forfeited before year five, and purchases cap at $10,000 per person per year. State-exempt, federally deferred until redemption, with an education-use exemption at the federal level under income conditions. As a short-term vehicle it is mediocre below eighteen months and increasingly singular beyond, which the simulator’s rising curve makes visible. The instrument’s defining moment was recent: buyers who filled the cap through 2021-2022 carried 9.62% composite rates through the inflation spike, the highest in the program’s history, with zero principal risk. That episode is the family’s use case in miniature: it pays for being early, not for being tactical.

3.6 The design questions behind the shelf

Two structural questions sit behind every line above and have their own treatments. What happens when a capped, tax-favored account is full, the overflow question, is examined in when a capped savings account overflows; and what a state is doing when it subsidizes saving for some households and not others is the subject of the means-tested savings account subsidy. Both read the American shelf against its European administered cousins, where caps and means tests are the explicit design rather than the exception. The comparison is not decorative: design choices, who gets subsidized rates, up to what amount, at whose expense, are the shelf’s invisible architecture, and they surface in every episode where administered rates and inflation diverge. France’s February 2026 reset, a formula pointing to 1.4% rounded up to 1.5% by ministerial decision, is the mechanism in miniature: an administered rate is a policy choice wearing a formula.

4. Your horizon decides: the simulator

The simulator prices the shelf at the horizon you choose, from three to thirty-six months, under your inflation assumption and generic declared tax rates. Curves appear along the axis as vehicles become accessible; locked products carry their penalties where they apply. Three inputs, amount, inflation assumption, horizon; the tax brackets are generic, fixed and printed in the source line. The output is a descriptive table, deliberately unranked: at most horizons the field sits within a few tenths of a point, and which tenth matters depends on tax facts the tool states rather than assumes away.

5. The safe asset on trial: 2021-2023, both sides of the Atlantic

The shelf’s worst documented episode is recent enough to have current account statements. Across 2022, rolling 3-month T-bills returned about 2% while US CPI averaged 8% (FRED TB3MS, CPIAUCSL): a purchasing-power loss near 6% in twelve months on the system’s reference safe asset, with savings accounts, then averaging a fraction of a percent, faring worse. France ran the same experiment with administered rates: the Livret A averaged 1.38% against 5.20% inflation in 2022 (Banque de France, INSEE). No market crash was required on either side; holding cash was sufficient. The pattern is not archived, either: in spring 2026 the Livret A at 1.50% slipped back under French inflation at 1.8% (INSEE, June), a small negative real gap reopening in real time, with press calculations pointing to a catch-up reset in August. The full episode, and what a nominal print conceals, is unpacked in the inflation haircut on cash returns and given historical depth in the nominal-real gap in yields. The instruments that escaped the episode shared one design feature, resets: rolling bills repriced upward within weeks and I bonds paid their record composites, while everything fixed and nominal absorbed the loss in proportion to its rigidity.

The slower version of the same loss ran through the previous decade and printed on no statement: deposits near half a percent against 1.5 to 2% inflation surrendered a cumulative double-digit share of purchasing power between 2010 and 2020, one invisible basis point at a time. Fast cut and slow bleed are one phenomenon at two speeds, and the shelf has no third mode: cash is always paying or charging in real terms, and the sign is set by the regime, not by the product. The episode reads differently by regime, which is the page’s analytical point. Today’s regime classification shows a transition state, mixed signals without a clear growth-inflation direction (Eco3min classifier, June 2026), and the transatlantic policy split sharpens it: the Fed held at 3.50-3.75% in June while the ECB hiked its deposit rate to 2.25%, its first increase in a year, against inflation it describes as slow to return to target. When administered or market rates sit below inflation for extended periods, the configuration has a name and a payer, documented in the mechanics of financial repression and, in its administered-rate form, in administered rates as a textbook repression case. What the transition state changes operationally is the price of waiting. With top nominal rates near 4% and measured inflation near 3.3% on the semiannual base, holding insured cash costs close to nothing in real terms for the first time since before the shock; the repression-era saver paid 4 to 6 points a year for the same parking spot. That is not a signal to hold cash, it is the removal of a penalty for doing so, and the distinction matters: a near-zero real rate rewards patience without rewarding cash itself. The base rates for cash across every configuration since the 1970s sit in historical returns under each regime, and the summary is one line: cash protects nominal balances in every regime and purchasing power only in some. Judging today’s 4% against its own regime rather than against memory is the discipline treated in what a nominal 4% buys in real terms.

6. The layer before the yield: sizing the cushion

Every comparison above assumes the short-term pocket is correctly sized, which is the decision that actually dominates outcomes: an emergency fund earning a mediocre rate beats a brilliant rate on a fund that is too small to absorb a shock. Stress-testing that sizing is the job of stress-testing your financial resilience, and calibrating the monthly flow that fills it belongs to how much to set aside each month. The yield questions of this page start after those two are answered. The ordering is not pedantry: the penalty section of every locked product above is, functionally, the price of having skipped it.

7. FAQ

How are T-bills taxed compared with HYSAs?

T-bill interest is subject to federal income tax but fully exempt from state and local tax; savings account interest is taxed at both levels. The gap is decisive in high-tax states: above roughly 8% state income tax, a 3.60% bill outearns a 4.00% savings account after tax. In no-tax states, the savings account’s higher print survives intact.

How does a T-bill ladder work?

A sum is split across maturities, for instance 4, 8, 13 and 26 weeks, and each bill is rolled into a new one at expiry. A quarter of the capital reprices roughly every month, no single decision commits the whole amount, and the blended yield tracks the policy rate with a one-to-two-month lag. Auctions run weekly, so the structure maintains itself once set; the cost of running it at TreasuryDirect or through a brokerage is zero commissions and a few minutes a month.

What determines I-bond rates?

Two components. A fixed rate, set at issuance and locked for the bond’s 30-year life, currently 0.90% for May-October 2026 issues; and an inflation adjustment recalculated every six months from CPI-U, currently 3.34% annualized. Together they compose to 4.26%. The fixed component is a guaranteed real rate; the variable one tracks whatever inflation does.

How did cash returns compare with inflation in 2021-2023?

Badly, on both continents. US 3-month T-bills returned about 2% across 2022 against 8% average CPI; American savings accounts averaged well under 1%. France’s Livret A served 1.38% against 5.20% inflation in 2022, then 2.92% against 4.90% in 2023. The reference safe assets lost roughly 4 to 6 points of purchasing power with no market event involved.

What distinguishes money market funds from savings accounts?

Legal nature and insurance. A savings account is a bank deposit, FDIC-insured, with a rate set at the bank’s discretion. A money market fund is an investment holding short government or corporate paper, not insured, with a yield that follows market rates mechanically within weeks. Treasury-heavy funds also pass through a state tax exemption that deposits never carry.

8. Where the short shelf ends

Past roughly three years, this page’s logic inverts, and the inversion is worth dating: it is drawdown recovery time, not preference, that draws the line. drawdown risk stops disqualifying the growth families, and the comparison reopens to the full field, run in the all-horizons investment panorama. Two adjacent decisions frame the shelf on its own terms: the account holding the cash sets its sweep rate and its exit costs, the grid of where short-term cash actually sits; and the rules of thumb that beginners inherit about cash buffers hold only under conditions worth knowing, catalogued in the beginner rules and when they hold. The shelf itself will be repriced each January, and mid-2026 suggests the revisions will not be symmetric across currencies: a Fed on hold and an ECB tightening put the two shelves on different clocks for the first time since the inflation shock. The structure, availability first, then locks, then rate risk, is the part that survives every revision.

Last updated — 15 September 2026