Antminer S21 vs S19 XP — Hashrate, Efficiency & ROI Compared
The Antminer S21 vs S19 XP showdown — hashrate, J/TH efficiency, ROI math at current hashprice, and when each ASIC still makes sense in 2026.
The Antminer S21 delivers 200 TH/s at 17.5 J/TH, while the older S19 XP runs 140 TH/s at ~21.5 J/TH. At April 2026 hashprice, the S21 mines roughly 43% more BTC per month and still costs less per terahash to run. The S19 XP only makes sense at deep secondary-market discounts with sub-$0.04/kWh power.
Why this comparison matters in 2026
When the Antminer S19 XP launched in 2022, it was Bitmain's flagship — 140 TH/s at roughly 21.5 J/TH, the best air-cooled efficiency money could buy. Two years later the Antminer S21 (base, non-hydro) arrived with 200 TH/s at 17.5 J/TH, pushing the efficiency floor down another 18%. That gap is now the entire reason this comparison exists.
Secondary-market S19 XPs are flooding listing sites as large farms cycle them out. You can pick one up well below new-machine MSRP, and that price gap is tempting for home miners, small hosting customers, and anyone weighing a first ASIC purchase. The question is whether the sticker discount offsets two years of compounding efficiency losses — every extra joule per terahash is money handed to the power company instead of to you.
This piece runs the numbers in one direction: is the newer S21 worth the premium over a discounted S19 XP? We compare specs, electricity cost, revenue, and break-even math at April 2026 hashprice so you can decide which side of the trade makes sense for your power price.
Spec sheet — S21 vs S19 XP head to head
Both machines are air-cooled Bitmain flagships separated by one generation. They share the SHA-256 algorithm, 220–240V three-phase power input, and a ~75 dB noise profile — loud enough that neither belongs in a living room. The difference is in the chips.
| Specification | Antminer S21 | Antminer S19 XP |
|---|---|---|
| Launch year | 2024 | 2022 |
| Hashrate | 200 TH/s | 140 TH/s |
| Power draw at wall | 3,500 W | 3,010 W |
| Energy efficiency | 17.5 J/TH | ~21.5 J/TH |
| Chip process | 5 nm (Samsung) | 5 nm (TSMC) |
| Cooling | Air (4 fans) | Air (4 fans) |
| Noise level | ~75 dB | ~75 dB |
| Warranty (new) | 12 months | 12 months (if still new) |
| Retail price range (2026) | ~$3,000–$4,000 new | ~$1,200–$1,800 used |
The S19 XP wins only on raw wall wattage (it pulls 490 W less) and on used-market sticker price. Every other spec that matters for long-term profit — hashrate, J/TH, and projected useful life — tilts toward the S21. A lower power draw means nothing if your revenue per watt is worse.
What the efficiency gap costs you every month
Let's ground the J/TH numbers in actual money. Assume a residential-adjacent industrial rate of $0.05/kWh and 24/7 uptime for a 30-day month (720 hours).
Antminer S21 — electricity
3.5 kW × 720 hours × $0.05/kWh = $126/month. You produce 200 TH/s of work for that bill.
Antminer S19 XP — electricity
3.01 kW × 720 hours × $0.05/kWh = $108/month. Only 140 TH/s of work for that bill.
On the surface the S19 XP looks $18/month cheaper to power. But you have to normalize for work produced. Divide electricity by hashrate to get cost per terahash per month:
- S21: $126 ÷ 200 TH/s = $0.63 per TH/s/month
- S19 XP: $108 ÷ 140 TH/s = $0.77 per TH/s/month
The S21 buys you hashrate 22% cheaper per terahash than the S19 XP, even though its wall draw is higher. That's the entire point of efficiency — revenue scales with TH/s, not with watts. More joules per terahash means more of your electricity bill gets burned as heat rather than converted to block reward probability.
ROI math at April 2026 hashprice
Hashprice is how much revenue one petahash (1,000 TH/s) earns per day. It fluctuates with BTC price, network difficulty, and mempool fee levels. As of April 2026 hashprice sits near $40 per PH/s per day. For a single-machine comparison we convert that to per-TH/s per day: $40 ÷ 1,000 = $0.04/TH/s/day.
| Monthly economics | Antminer S21 (new) | Antminer S19 XP (used) |
|---|---|---|
| Gross revenue (30 days) | ~$240 | ~$168 |
| Electricity at $0.05/kWh | −$126 | −$108 |
| Net profit / month | ~$114 | ~$60 |
| Up-front cost | ~$3,500 | ~$1,500 |
| Break-even (months) | ~31 months | ~25 months |
| Profit over 36-month horizon | ~$604 | ~$660 |
The twist: over a static 3-year horizon at $0.05/kWh, the discounted S19 XP actually edges out the new S21 on total dollar profit — because the up-front discount is so steep. So why does anyone buy new?
Three reasons the table hides. First, hashprice is not static — as network difficulty grinds upward and block subsidies decay post-halving, less efficient miners cross the electricity-cost-of-production threshold and go unplug first. The S19 XP's margin is thinner and gets squeezed sooner. Second, the S21 carries a factory warranty and fresh chips; a used S19 XP may have 2+ years of thermal cycling already on the hashboards. Third, the S21 holds resale value far better over the next 12 months as the S19 XP keeps sliding.
Run the same math at $0.08/kWh (typical US residential) and the S19 XP's net profit collapses to roughly $-1/month — it becomes a paperweight. The S21 still makes about $74/month. The higher your power cost, the more the newer, more efficient miner wins.
Skip the S21 vs S19 XP dilemma entirely
The ROI math above assumes you're buying hardware. When you rent hashrate from NoxHash's Kazakhstan fleet, you plug into S21-tier machines running on $0.035/kWh industrial power — no $3,500 box to own, no warranty games, no obsolescence clock.
When the S19 XP still makes sense in 2026
We're not here to bury the S19 XP. It was the best air-cooled ASIC of its generation and it still mines Bitcoin competently if the conditions line up. The honest answer: it wins in a narrow band of scenarios.
Deeply discounted secondary market
If you can find a working unit under $1,200 with a functional hashboard warranty, the lower up-front cost can outperform a new S21 over a 24-month horizon, assuming cheap power.
Legacy deployments already amortized
If you bought an S19 XP new in 2022 and already broke even, you're now printing low-margin income. No need to panic-upgrade — run it until the next halving or until electricity costs catch up.
Sub-$0.04/kWh power sources
Cold-climate sites with flared gas, hydro curtailment, or off-peak industrial contracts can absorb the higher J/TH. If your marginal kWh is near-free, hardware efficiency matters less than capital spent per terahash.
Note what's missing from that list: typical home mining in most countries. At $0.08–$0.15/kWh residential rates the S19 XP is already a loss-maker in April 2026 and only gets worse as difficulty climbs. The remaining use cases are industrial-adjacent, not hobby-tier.
This is the same logic that keeps the S19 XP out of the NoxHash Kazakhstan fleet. We operate S21, S21 Pro, S21 Hyd, S21 XP Hyd, and WhatsMiner M60S/M66S machines — everything sits in the 12–18.5 J/TH band. At industrial scale, the efficiency gap between 17.5 J/TH and 21.5 J/TH compounds across thousands of machines into real money. We cycle out hardware when new generations deliver >15% efficiency gains, which is exactly why the S19 XP tier is gone.
What cloud mining users should take from this comparison
Here's the part most hardware comparison articles miss: if you rent hashrate instead of owning hardware, you don't buy a specific ASIC — you rent TH/s from a fleet. The NoxHash plans deliver a fixed hashrate (50 / 150 / 350 / 1,000 TH/s depending on tier) regardless of which specific rack your allocation resolves to on any given day.
That abstracts away the S21-vs-S19-XP choice for renters, but the hardware comparison still matters to you indirectly. It sets the floor on cost structure. A fleet built on 17.5 J/TH machines has roughly 22% more BTC per electricity dollar to work with than one built on 21.5 J/TH machines. That headroom is what lets a platform stay solvent during BTC price drawdowns and difficulty spikes — both of which hit the least efficient miners first.
When you evaluate any cloud mining service, ask which generation of ASICs powers their fleet. Providers running on S19-tier or older hardware are mining on thinner margins; providers on S21-tier hardware have a buffer. Ask for a specific machine model, not just a brand name — "Antminer" covers everything from 2020-era S19 to 2024 S21 XP Hyd, and the difference in economics is enormous.
Key takeaways
Efficiency beats raw power draw
S19 XP pulls less wattage, but S21 gets 22% more hashrate per electricity dollar. Revenue scales with TH/s, not with watts saved.
Power price is the tiebreaker
At $0.05/kWh the S19 XP still pays. At $0.08/kWh it flips to unprofitable while the S21 keeps earning. Check your rate before buying.
Used hardware carries hidden risk
A discounted S19 XP may have 2+ years of thermal cycling on the hashboards. Factor in replacement costs and no warranty when modeling ROI.
Renting sidesteps the decision
Cloud mining lets you tap S21-tier efficiency without a $3,500 purchase. You pay per TH/s — the fleet operator absorbs hardware obsolescence.
Frequently asked questions
Is the Antminer S19 XP still worth buying in 2026?
Only at deeply discounted secondary-market prices — roughly under $1,500 — and only if your electricity costs stay under $0.05/kWh. At typical US residential rates of $0.10–$0.15/kWh, the S19 XP loses money in April 2026 and will keep losing money as difficulty climbs.
Can an S19 XP mine profitably anywhere in 2026?
Yes, in three niches: sub-$0.04/kWh industrial contracts, flared-gas sites in Texas or North Dakota, and curtailed hydro in places like British Columbia or Paraguay. Outside those niches the 21.5 J/TH efficiency leaves too little margin once difficulty adjusts upward.
How long until the S21 becomes obsolete like the S19 XP?
Historically Bitmain releases a new flagship generation every 18–24 months with a 15–20% efficiency gain. The S21 shipped in 2024, so expect the S22 or successor line around 2026–2027. That likely gives S21 a useful-life tail of three to four years before new hardware dominates economically.
Does NoxHash use any S19 XPs in its fleet?
No. Our Kazakhstan fleet runs S21, S21 Pro, S21 Hyd, S21 XP Hyd, and WhatsMiner M60S/M66S — all in the 12–18.5 J/TH efficiency band. We cycled out S19-tier hardware once 17.5 J/TH machines became available, because at industrial scale the efficiency gap becomes a solvency issue.
What comes next after the S21 generation?
Bitmain has already shipped variants pushing the S21 family down to 12 J/TH (S21 XP Hyd) and 15 J/TH (S21 Pro). A full next-gen S22 series is expected in 2026–2027 and will likely target sub-12 J/TH using 3 nm or 4 nm nodes. MicroBT's WhatsMiner M70 generation is on a similar roadmap.
Should I buy an S21 new or wait for the S21 price to drop?
If you have cheap power and want to own hardware, buying S21 now is reasonable — every month you wait is lost hashrate income. If you don't have sub-$0.06/kWh power, skip ownership entirely and rent hashrate from a fleet that does. The hardware arbitrage only works where electricity is cheap.
How does hashprice affect this comparison?
Hashprice scales revenue linearly for both machines. If hashprice doubles to $80/PH/s/day, both miners print money; if it halves to $20/PH/s/day, the S19 XP flips unprofitable first. Track current hashprice on Hashrate Index before making any purchase decision.
Not investment advice
All ROI math in this article uses April 2026 hashprice (~$40/PH/s/day) and example electricity costs — actual results vary with BTC price, network difficulty, mempool fees, and your power rate. Mining income is not guaranteed. Past hashprice performance does not predict future earnings. Only commit capital you can afford to lose, and check current hashprice data before making any hardware purchase.