USB Stick Miners: The Rung Between Toy and Bitaxe

Between a $25 ESP32 that will never find a block and a $66 board that hashes at a terahash sits a category most guides skip entirely. USB ASIC sticks are real miners running the same silicon as a Bitaxe, with odds measured in tens of thousands of years rather than billions. They are also not standalone, and the number on the listing is not what you will spend.

Our NerdMiner piece said the next rung up was a Bitaxe. That skipped a step. Here is the step.

Key takeaways

• Current USB sticks reach 600 GH/s at 10 W, 800 GH/s at factory settings, and 1 TH/s pushed hard — on the same BM1370 chip a Bitaxe Gamma uses.
• They are not standalone. A stick needs a powered USB hub and a host computer running CGMiner. A Bitaxe needs a wall socket.
• Real cost is about $140 before the host, against $61 to $72 for a Gamma delivering twice the stock hashrate.
• Hashrate per dollar: 4.29 GH/s for a stock stick, about 18 for a Gamma. More than four times better on the standalone board.
• Running cost is about 40% higher once the host’s own draw is counted, for less hashrate.
• Sticks still win on modularity — small increments, one host, and a natural fit beside a full node you already run.

The full ladder

Every figure below uses the same Bitcoin difficulty of 126.23 trillion, read early August 2026, so the rows are directly comparable.

Manufacturer and retailer figures read 7 August 2026, except the Gamma’s 14.3 W which is a measured wall figure. Expected block times calculated as difficulty multiplied by 2³² divided by hashrate.

Notice where the discontinuity is, and where it is not. Between the ESP32 and the smallest ASIC stick there are five orders of magnitude — that is a real cliff, and our NerdMiner piece covers why no firmware closes it. But from 25 GH/s upward the whole range is one continuous ladder. A USB stick and a Bitaxe are the same kind of thing at different points on it.

One chip or two? The arithmetic settles it

Worth a short detour, because sources disagree and it is a good illustration of how to check a spec sheet.

One retailer describes the current Compac A2 as using dual BM1370 chips. Another describes a single one. They cannot both be right, and you do not need to open the device to know which.

A single BM1370 in a Bitaxe Gamma produces about 1.2 TH/s. The Compac A2 tops out at 1 TH/s when pushed to its limit. Two of the same chip would land nearer 2.4 TH/s. The A2 has one chip, and the listing that says otherwise is wrong. When two sources conflict on a component count, the hashrate usually tells you the answer.

The cost nobody totals

This is the part that changes the decision, and it is structural rather than a matter of pricing.

A USB stick is not a miner. It is half of one. It has an ASIC and a USB interface, and nothing else: no controller that can talk to a pool, no network stack, no dashboard. It needs a host computer running CGMiner, and because a USB port cannot supply the current, it needs a powered hub as well.

Prices as listed 7 August 2026. Hub cost is an approximation and varies; the host is excluded from both columns because its price depends entirely on whether you already own one. Board prices in this market have fallen sharply — a Gamma listed above $100 as recently as February 2026 — so check current pricing rather than trusting any comparison table, including this one.

The running cost follows the same pattern. A stick at 10 W plus a small host board at roughly 10 W is about 20 W continuous. That is 40% more electricity than a Gamma for half the hashrate — around $2.34 a month at US rates against $1.67, or €4.23 against €3.03 at the EU average. Our European cost breakdown has the by-country figures.

To be fair to the sticks: if the host is a node you were running anyway, its power is already spent and the comparison narrows considerably. That is the scenario where they make most sense, and it is discussed below.

Efficiency, honestly compared

One caveat that matters and cuts against the stick. The A2 figures are manufacturer numbers for the device itself; the Gamma figure is measured at the wall socket. Comparing a nameplate against a measurement flatters the nameplate, and if the A2’s own draw were measured at the plug with its hub, the gap would widen rather than close. Our power measurement piece covers why that difference is typically around a fifth.

The other observation worth making: the A2’s efficiency barely moves across its whole operating range, 16.5 to 16.9 J/TH from 600 GH/s to 1 TH/s. That is unusual and it is a point in the design’s favour. On most boards, pushing harder costs efficiency noticeably.

When a stick is actually the right answer

Four cases, and they are real rather than consolation prizes.

1. You already run a node. If a machine is on 24/7 anyway, the host cost is zero and the powered hub is the only extra. A stick beside a full node is the classic use, and it sits naturally with running your own stratum layer — see our architectures piece.
2. You want to scale in small increments. Adding a stick to an existing hub costs $100 and no new power supply, no new IP address, no new device to update. Adding a Bitaxe means another of everything.
3. You want command-line control. CGMiner exposes clock, core voltage and fan through a scriptable interface. For anyone who wants to automate tuning across several devices, that is genuinely easier than driving several web dashboards.
4. The form factor matters to you. Some people want a rack of sticks on a hub. That is a legitimate reason and nobody needs to justify it further.

They have also won. Retailers document several instances of USB sticks taking full block rewards during the 6.25 BTC era. The odds were terrible and the wins were real, which is the same sentence that applies to every device at this scale — our wins tracker collects the documented cases.

When it is not

If your goal is maximum odds per dollar spent, the standalone board wins and it is not close: roughly 18 GH/s per dollar against 4.29, with no host computer and no hub.

The clearest illustration comes from a retailer that sells both. Their own guide notes that a four-stick combination priced near $600 delivers the same 1.2 TH/s as a single standalone board. That was written in February 2026 when the board cost just under $100; it now lists in the sixties, so the gap has widened rather than closed. A vendor arguing against its own higher-margin product is the kind of source worth trusting.

And if you are buying a stick because $100 sounds like an entry price, that is the comparison to avoid. It is $140 plus a computer, against $61 to $72 for twice the hashrate and nothing else needed.

The honest summary

USB ASIC sticks are not a scam and not a toy. They are the same silicon as the boards everyone recommends, in a form factor that trades integration for modularity, and they occupy a genuine rung on the ladder that most coverage jumps straight over.

What they are not is the cheap entry point they appear to be. The stick is one component of a mining setup, and the other components are not optional. Price the whole thing, including the machine that has to stay switched on, and then decide.

Either way, the arithmetic underneath is the same as it is at every scale: odds scale with hashrate and nothing else, and at 600 GH/s or 1.2 TH/s you are buying a lottery ticket with a calculable price. Our odds page has the full curve, and what is lottery mining covers why that is still a reasonable thing to buy.