We Analyzed 30 Days of IBKR Borrow Fees: 5 Names Never Left Triple Digits

TL;DR: We recomputed 30 days of nightly borrow-fee data — 2,804 observations of 470 symbols across 20 nightly IBKR files, August 11–September 5, 2026. Sixteen names printed a triple-digit annualized borrow fee at least once; five — JZ, GCTK, INM, RGC, EZGO — never printed below 100% on any night they were tracked. The triple-digit club grew from 2 names to 9.

The borrow fee is the annualized carry a short seller pays to keep a position open, and it is the purest public read on how scarce a stock is in the lending pool. Tapeboard ingests the IBKR stock-loan file nightly for every name in its short-side coverage universe and archives each snapshot. This study re-ran every figure below directly from that archive — 20 nightly files dated August 11 through September 5, 2026, covering all 30 calendar days ended September 5 — so every number in this article is reproducible from the same rows the terminal serves.

The 30-day picture

The first thing the series shows is how calm the typical name is. The median fee across the coverage universe sat between 0.41% and 0.42% on all 20 nightly snapshots. Of the 2,804 symbol-day observations, 2,249 — 80.2% — were below 2% annualized, the general-collateral band where borrow is effectively free.

The action lives in the tail. Sixteen distinct names printed 100% or higher at least once during the window, and the size of that club trended up all month: 2 names on August 11, 7 by August 22, and 10 on September 4 — the widest nightly count of the window — before settling at 9 on September 5.

The persistence board, ranked by nights spent at or above 100%:

Source: Tapeboard's archived nightly IBKR stock-loan snapshots, August 11 – September 5, 2026. No name was tracked on all 20 nights (coverage rotates — see Limitations), and an untracked night breaks a streak; streaks count consecutive snapshots with a triple-digit print.

Five names — JZ, GCTK, INM, RGC, EZGO — were tracked for at least ten of the 20 nights and never once printed below 100%. JZ is the standout: it spent the first two weeks of the window between 106.9% and 120.3%, then repriced 709 percentage points overnight — 106.9% on August 25 to 815.9% on August 26, the largest single-night fee jump in the dataset — and still closed the window at 606.1% on September 5. GCTK is the mirror image: it opened the window as the most expensive name at 364.6%, slid to 176.8% by August 22, and ended the window at 231.2% — seventeen tracked nights and it has never printed below 176.8%.

One texture worth noting: these rates don't wobble much night to night once a desk sets a level. EZGO printed exactly 209.1% on seven of its eleven tracked nights; INM printed exactly 161.0% on six of fifteen. Stock-loan desks move rates in steps, not continuously — the repricing events, when they come, arrive all at once, as JZ's August 26 print shows.

What "never left triple digits" means

A 100% annualized borrow fee means a short pays the full notional value of the position per year just in carry — roughly $274 per day on a $100,000 short. At JZ's August 26–27 print of 815.9%, that same short accrues over $2,200 per day. Fees at that level mean the broker's lendable inventory is effectively exhausted and the desk is rationing the remaining shares through price.

That is a statement about scarcity, not about direction. Triple-digit fees cluster in names that are distressed, very-low-float, recently reverse-split, or newly public — the fee tells you shorts are trapped in an expensive seat, not that a short squeeze is coming. High fee is one input; Tapeboard's composite squeeze score pairs it with short interest, days to cover, and volume factors before calling anything a setup.

What this list does capture is persistence of scarcity. A one-night 400% print can be a thin-offer artifact in a single nightly file; nineteen tracked nights above 100%, as JZ posted, is a structural condition — the name has printed triple digits in every nightly file that tracked it since August 11, through the biggest repricing of the window.

The turnover board

The other eleven names that ever touched triple digits show how fast the door swings both ways:

• GOVX is the boundary case — pinned right at the line. It printed 90.2% and 92.3% on August 18–19, crossed to 128.8% on August 20, dipped back to 91.5% for a night, then re-crossed on August 22 and stayed between 101.7% and 109.1% through September 5.
• VVOS spent all of August under 11%, stepped up to 25.4% on September 1–2, tagged 130.9% on September 4 — and was back at 87.4% by September 5. One night in the club.
• DRMA printed between 103.0% and 113.0% on four nights from August 18–22, then faded back to the 93.8–98.6% band just under the line.
• One-night spikes: BAOS hit 429.0% on August 13, OFAL 382.1% on August 14, CYCU 105.9% on August 15; none printed triple digits a second time.
• Late arrivals: ATPC appeared on August 28 already at 654.4% and held a 642.2–663.6% band for seven straight nights; BIAF arrived August 25 at 208.3%, cooled toward 119.6%, then repriced to 689.6% on September 3 and held it through September 5 — the window's closing number-one. XPON jumped 279.1 percentage points in a night (96.4% to 375.5% on August 28), peaked at 530.6% on September 1, and was still at 340.2% on its last tracked night of the window. NCPL printed 640.2% twice (September 2–3) and 476.6% on September 5.

Seven of the sixteen names that ever touched 100% held the level for four or fewer tracked nights. Persistence at triple digits is rare; churn is the norm.

Where to watch it live

The highest borrow-fee leaderboard refreshes nightly from the same IBKR file this study used, with the full ranking and per-ticker fee history. Our daily snapshot post — the latest is the September 4 hard-to-borrow table — archives each night's top 25. For how borrow fee weighs against short interest, days-to-cover, and the other factors, see the squeeze score methodology; borrow fee carries a 25% weight.

Limitations

• One venue. These are IBKR-published rates. Prime brokers and other retail brokers run their own loan desks, and their rates for the same name on the same night can differ materially. IBKR is a large, real benchmark — not the whole market.
• Coverage, not census. The archive tracks the symbols in Tapeboard's short-side coverage universe (roughly 115–166 tracked per night, 470 distinct over the window), rotated to concentrate on the most-shorted names. A name absent from a night's file row is a coverage gap, not evidence the fee fell — 97 symbols in the window were tracked on exactly one night. The five-name persistence claim covers every night each name was tracked, and no name was tracked on all 20 nights.
• Archive gaps. No nightly files were archived for August 7–10, inside the 30-calendar-day window billed in the headline. The study's universe is the 20 nightly files we do hold (August 11 – September 5); persistence claims extend no further back than that.
• Nightly cadence. Fees are end-of-day file prints, not intraday. A squeeze that reprices the loan market at 10:30 AM shows up in the next file, not this one.
• Fees are gross of rebate. The quoted annualized fee does not net the rebate on short proceeds; true carry differs modestly.
• Not a squeeze call. High borrow fees often mark fundamentally weak names. Nothing here is a prediction of price direction, and a triple-digit fee is not a buy signal.

This post is for educational and informational purposes only and is not investment advice. Borrow fees reflect securities-lending conditions reported in IBKR's daily data; they are not real-time intraday rates and may differ from rates at other brokers. Short selling carries unlimited downside risk. Editor: Marcus Reilly.