Four HTB Vendors Is Not Redundancy. It Is Short-Side Survival.
- May 9
- 4 min read
Updated: 2 days ago
TL;DR — One HTB vendor is a single point of failure. When borrow tightens on a hot name — UPST sits at 32% short of float right now, BYND at 28%, AMC at 18% — a single vendor's supply runs dry within minutes of the pre-market open. The math: four independent vendor relationships don't reduce expected borrow cost by much, but they reduce the probability of being shut out of a setup at the wrong moment by an order of magnitude. The cost of running four desks is trivial. The cost of missing the trade because the one vendor is out of supply is the entire expected alpha.
This is a companion piece to our Hard-to-Borrow Mechanics overview. That article walks through what borrow rates, recalls and Reg SHO mean for the trade. This one is operational: why we run four HTB vendors, what changes when borrow tightens, and how a desk decides which vendor to call first on which name.
Where the single-vendor model breaks
Most retail and many smaller prop platforms run a single HTB relationship. The customer borrow inventory comes from one prime broker or one borrowing desk. That works as long as the universe of names you short doesn't significantly overlap with what every other customer of that vendor is also trying to short. The moment overlap happens — and it happens on every notable HTB name — the model fails for someone.
What the current HTB universe looks like
The chart above is a current snapshot of short interest as a percentage of float across a sample of HTB-prone US names. Read it as the current map of where vendor supply is most likely to fail:
UPST (32% short of float): consumer-lending name with persistent skepticism. Borrow has been stressed for most of the last 18 months. Single-vendor desks regularly run out of supply within 30 minutes of market open.
BYND (28%) and KSS (28%): the structural short universe — consumer brands with deteriorating fundamentals. Borrow rates oscillate between 15% and 200%+ depending on the news cycle.
AMC (18%) and TLRY (16%): names where retail sentiment can crowd the short in a single morning. Borrow can disappear faster than any institutional vendor can backfill.
GME (15%) and CVNA (13%): well-known meme/squeeze candidates. Borrow has been stressed-then-clearing-then-stressed in waves since 2021.
Every name above the 20% line has had multiple sessions in the past year where a single vendor's supply was exhausted before the cash open. Traders without redundancy were locked out of the setup during the window the chart was calling for participation.
Why four vendors changes the math
Each HTB vendor has different inventory sources. A long-only mutual fund lends through Vendor A. A pension fund lends through Vendor B. A wealth manager lends through Vendor C. A securities-lending agent lends through Vendor D. When demand spikes on a name, the four vendors' supply state moves independently — Vendor A may be exhausted while Vendor C still has 50,000 shares available at a workable rate.
The redundancy isn't about getting a better rate. It's about being in the trade at all when the name matters. The operational cost of running four desks — relationship maintenance, integration with the locate workflow, settlement reconciliation across clearing firms — is trivial compared to the cost of missing a setup with the entire desk's prep behind it.
The four-vendor workflow we actually run
Pre-market scan: the watchlist gets cross-checked against current locate availability across all four vendors. Names where 2+ vendors are already showing tight supply are flagged for tactical-short-only treatment.
Locate request order: we don't request from all four simultaneously — that signals demand to the lender community and accelerates rate hikes. We sequence requests starting with the vendor most likely to have inventory for that specific name, based on historical fill patterns.
Mid-session re-checks: if a vendor recalls or reprices, we check the other three before assuming the trade has to be exited. Sometimes the recall is vendor-specific and another desk still has term supply at a workable rate.
Term decisions: when one vendor offers term borrow at a high rate and another offers open borrow at a lower rate, we evaluate based on the thesis duration and recall sensitivity. Multi-vendor visibility makes this an actual choice rather than a forced acceptance.
When one-vendor desks miss the trade
The asymmetry shows up most clearly on pre-earnings shorts and on post-news squeeze attempts. A trader running a single vendor often discovers their locate has been denied at 09:25 — five minutes before the cash open — when the borrow desk reallocates inventory based on overnight demand surge. There's no time to source alternate supply, and the trade window closes before the order can be sent. We don't have that failure mode because by 09:25 we've already mapped which of our four vendors has the cleanest path to the locate, and we've staged the locate accordingly.
Connected pieces
This piece sits inside a broader execution framework. The Hard-to-Borrow Mechanics overview covers borrow rates, term/open, recalls, and Reg SHO. The DMA vs retail broker execution piece covers how short-side fills survive contact with the market. Together they describe the operational machinery that makes short selling at institutional quality actually work.
Joining the desk
Vortex Capital Group runs four HTB vendors and multi-clearing redundancy precisely because short-side reliability is a system property — not an outcome of any individual decision. Qualified short-biased traders applying to the desk can expect locate availability, controlled borrow rates, and recall response that match institutional execution quality. The trader application takes about ten minutes.
Comments