APP
How to Decide When to Upgrade to a New ASIC Model
2026-09-09 09:32

Introduction

A newer ASIC miner with a higher nameplate hashrate or a lower J/TH rating is not, by itself, a reason to upgrade. The decision only makes economic sense when the expected incremental cash flow from the replacement—after electricity, hosting, deployment, downtime, and the resale value of the displaced unit—justifies the capital outlay under realistic Bitcoin price and network difficulty scenarios. This article sets out a measurement-based framework for making that judgment, and separates three categories of data that are often conflated: miner-side measurements (local hashrate, wall power, temperature, uptime), pool-side data (pool-estimated hashrate, submitted and rejected shares), and financial results (BTC revenue, fees, and realized net cash flow).

Step 1: Establish the Existing ASIC's Measured Baseline

Before comparing an installed machine to a candidate replacement, operators should document its actual performance over a representative period, using consistent time windows for every input:

  • Average local hashrate in TH/s, read directly from the miner or fleet-management software.
  • Measured wall power in watts, ideally from a branch-circuit meter or PDU rather than a manufacturer specification.
  • Uptime and any curtailment hours.
  • Pool-estimated hashrate over a comparable period.
  • Rejected-share percentage, and, where the pool reports it, the breakdown into stale, invalid, duplicate, and other rejection reasons.
  • Electricity price, demand charges, and any hosting fees.
  • Repair, labor, and spare-parts costs.
  • The current resale or redeployment value of the existing unit.

Local hashrate and pool-estimated hashrate are related but not interchangeable. ViaBTC's documentation notes that its real-time pool hashrate is calculated as an average over the prior ten minutes, while a mining machine's own display may refresh on a different interval (ViaBTC Help Center). A gap between the two readings does not automatically indicate a hardware problem; it may simply reflect different averaging windows.

For upgrade economics, these operating-cost inputs should also be separated into costs that actually change with the replacement and costs that remain fixed. Variable or avoidable costs, such as metered electricity consumption, may change materially when a more efficient miner is installed. Fixed or unavoidable costs should not be treated as upgrade savings unless the replacement genuinely changes the amount owed.

Step 2: Calculate Device Efficiency Correctly

Efficiency, expressed in joules per terahash (J/TH), should be calculated only from measured miner-side inputs over a steady-state period:

J/TH = measured miner power (W) ÷ average local hashrate (TH/s)

Because one watt equals one joule per second, dividing watts by TH/s yields joules per terahash directly. Pool-estimated hashrate should not be substituted into this formula. Pool data is useful for assessing connectivity, share rejection, and payout performance, but it is not a compatible input for a device-level efficiency calculation, since it is inferred from shares received by the pool over a defined averaging window rather than measured directly at the miner.

Step 3: Calculate Electricity Cost Separately

For a single miner over 24 hours:

Daily electricity cost = (power in W ÷ 1,000) × 24 × electricity price per kWh

For a fleet expressed in TH/s, power in kilowatts can be derived from hashrate and J/TH:

Power (kW) = (hashrate in TH/s × J/TH) ÷ 1,000

This relationship is why J/TH is a useful basis for comparing machines on an equal-hashrate footing, but it should always be paired with the site's actual electricity price and uptime rather than treated as a standalone efficiency score.

Step 4: Model Incremental, Not Gross, Economics

The relevant question is not whether a new ASIC generates positive gross revenue—almost any operating machine will, under most price conditions. The relevant comparison is the difference the upgrade makes relative to continuing to run the existing unit:

Incremental daily net cash flow = (new mining revenue value − old mining revenue value) − (new operating costs − old operating costs)

All terms in this cash-flow equation should use the same unit of account. If mining earnings are first projected in BTC, they should be converted into the model's chosen reporting currency using the BTC price assumption for that scenario before being compared with electricity, hosting, labor, or other fiat-denominated costs. BTC output can still be tracked separately as an operating metric.

Only operating costs that actually change because of the replacement should be included in the incremental comparison. For example, lower metered electricity consumption can produce a genuine saving, while a fixed hosting or site charge that remains payable under either option should not be counted as an upgrade benefit.

One-time costs should be modeled separately from ongoing operating costs:

Net upgrade investment = ASIC purchase price + shipping and import costs + deployment and infrastructure costs + installation downtime cost − net proceeds from old ASIC resale

Here, installation downtime cost should reflect the lost net operating contribution during the period in which the existing unit or fleet would otherwise have been running. That means accounting for the mining revenue value forgone during deployment while also subtracting variable costs that are avoided because the displaced equipment is offline.

A simple payback calculation can serve as a planning tool:

Simple payback period = net upgrade investment ÷ expected incremental daily net cash flow

This should be presented as a scenario-based planning calculation rather than a guaranteed return estimate. It is sensitive to Bitcoin price, network difficulty, transaction fees, uptime, curtailment, and realized hardware performance—all of which change over time.

Step 5: Test Multiple Scenarios

Because Bitcoin's block subsidy is fixed by protocol rule at 3.125 BTC per block since the halving at block 840,000 in April 2024, and will reduce again at the next scheduled halving (Bitcoin.org), an upgrade evaluated over a multi-year horizon should not assume current BTC revenue extends indefinitely. It is generally useful to test at least three internally defined cases rather than relying on a single projection:

  1. Base case: a defined BTC price and difficulty path based on the operator's baseline assumptions, together with expected operating conditions.
  2. Downside case: lower BTC price, higher difficulty, lower transaction fees, or more downtime.
  3. Operational-stress case: reduced uptime, higher cooling costs, longer deployment time, or delayed energization of the new fleet.

No single scenario or payback period should be treated as a universal threshold; the acceptable payback horizon depends on an operator's financing terms, risk tolerance, and site conditions.

Efficiency Gains Are Real, but Infrastructure Fit Matters

Efficiency comparisons are easiest to reason about on an equal-hashrate basis. The Bitmain ANTMINER S19 XP is rated at 21.5 J/TH (141 TH/s, 3,031.5 W, measured at 25°C), while the ANTMINER S23 Hyd is rated at 9.5 J/TH (580 TH/s, 5,510 W, measured at 35°C) (Bitmain S19 XP specification; Bitmain S23 Hyd manual).

At 1 PH/s of hashrate, the S19 XP specification implies roughly 21.5 kW of power draw and 516 kWh of energy per day. The S23 Hyd specification implies roughly 9.5 kW and 228 kWh per day—a difference of about 288 kWh per day per PH/s. At an illustrative electricity price of $0.06/kWh, that difference is approximately $17.28 per day per PH/s in electricity cost alone.

This is an illustrative calculation using rounded manufacturer specifications, not a site-specific profitability forecast, and it assumes identical uptime for both machines. It also compares figures measured under different stated thermal conditions (25°C versus 35°C), so it should not be read as a controlled like-for-like test. More importantly, the S23 Hyd is a hydro-cooled unit that requires three-phase 380–415V input and a compatible coolant system—it is not a drop-in replacement for an air-cooled deployment. A lower J/TH figure can meaningfully reduce electricity consumption, but a more efficient hydro-cooled ASIC is not automatically the better replacement for an air-cooled fleet if the site cannot support its electrical and cooling requirements at a reasonable cost.

Rule Out Correctable Problems Before Attributing Underperformance to the Hardware

An apparent performance gap does not always mean the existing ASIC is obsolete; it may reflect configuration, connectivity, thermal, or firmware issues that an upgrade would not by itself resolve. Before concluding that replacement is necessary, it is worth checking:

  • Whether local hashrate and pool-estimated hashrate diverge after accounting for their different averaging windows.
  • The rejected-share percentage, and, where the pool reports it, whether stale, invalid, or duplicate shares make up the largest share of rejections.
  • Whether rejection rates increased after a firmware change, overclocking, a temperature change, or a change in connection endpoint.
  • Whether the site's ambient temperature, voltage, cooling capacity, and network conditions support reliable operation.

ViaBTC's support documentation describes a rejection rate within 3% as within its normal operating guidance, and associates abnormally high rejection rates with factors such as network conditions, elevated miner temperature, or firmware issues (ViaBTC Help Center). This figure reflects ViaBTC's own practical guidance rather than a universal standard applicable to every pool, coin, or network path, but it is a reasonable starting point for diagnosing whether an underperforming machine has a correctable issue rather than a genuine efficiency shortfall.

For operators using ViaBTC, hashrate and rejection-rate alerts can be configured with email, app push, or Telegram notifications. This is useful for establishing a pre-upgrade baseline and for monitoring a newly installed machine during its initial operating period, though an alert identifies a condition that warrants investigation rather than diagnosing a hardware fault automatically.

Upgrading Does Not Always Mean Buying New

Replacing hardware can also mean acquiring used equipment that is newer than the existing fleet without the cost of the newest model. In its Q1 2026 shareholder letter, MARA Holdings disclosed the acquisition of 2.4 EH of what it described as next-generation used ASIC miners to improve fleet efficiency. The filing does not specify unit models, per-unit pricing, warranty terms, or realized fleet-level returns, so the figure should be read only as evidence that operators sometimes pursue efficiency gains through used-equipment purchases rather than exclusively through newly manufactured units. Whether a used machine is a reasonable substitute for a new one depends on its delivered cost, verified condition, expected remaining service life, and measured energy profile—the same underlying factors that apply to any upgrade decision.

It is also worth distinguishing individual ASIC specifications from fleet-level reporting. Publicly reported fleet efficiency figures, such as a company's disclosed peak efficiency across its deployed miners, reflect a mix of models, operating modes, and deployment status, and should not be compared directly to a single machine's manufacturer rating without understanding how each figure was measured.

Conclusion

An ASIC upgrade is justified when the replacement machine's expected incremental net cash flow remains compelling after realistic electricity, infrastructure, uptime, and difficulty assumptions—and when the site can actually support the new model without deployment cost or operational disruption eroding those gains. Before committing capital to new hardware, it is worth confirming that the existing fleet's underperformance, if any, is not attributable to correctable issues such as poor cooling, unstable power, firmware misconfiguration, or elevated share rejection. A disciplined comparison of measured miner-side data, pool-side data, and realized financial results—modeled across more than one price and difficulty scenario—provides a more reliable basis for the decision than comparing nameplate specifications alone.

FAQ

Is a lower J/TH rating always worth paying for?

Not automatically. A lower J/TH figure reduces electricity consumption per unit of hashrate, but the purchase only pays off if the incremental net cash flow—after accounting for deployment cost, any required infrastructure changes, and financing—exceeds the net upgrade investment within a timeframe the operator finds acceptable.

Can I use my pool's hashrate estimate to calculate my miner's efficiency?

No. J/TH should be calculated from measured local hashrate and measured wall power over the same time window. Pool-estimated hashrate uses a different averaging methodology and is not a compatible substitute in a device-level efficiency calculation.

Does a high rejection rate mean I need a new ASIC?

Not necessarily. Elevated rejection rates are often linked to network conditions, high operating temperature, or firmware issues rather than to the hardware's underlying capability. ViaBTC's guidance treats a rejection rate within 3% as within its normal range; a higher rate generally warrants investigation before hardware replacement is considered.

Should I always buy the newest ASIC model available?

Not necessarily. The newest model may require infrastructure the site does not have, such as three-phase power or hydro-cooling. A used but still meaningfully more efficient machine, or a smaller infrastructure investment paired with an air-cooled model, may produce a better outcome depending on site constraints.

How many scenarios should I model before deciding?

There is no fixed number required, but modeling a base case alongside a downside case (lower BTC price or higher difficulty) and an operational-stress case (lower uptime or delayed deployment) provides a more realistic view than relying on a single projection.

References

Prev
How to Compare Mining Profitability Across BTC, LTC, DOGE, ZEC and KAS
Next
Zcash Mining Is Getting More Profitable: Is It Time to Mine ZEC Again?