The combination of post-halving block-reward compression and a Bitcoin price oscillating in the $90,000 to $100,000 range has pushed any mining rig older than Bitmain's S19 generation into structural unprofitability at average U.S. industrial power rates. Operators running pre-S19 fleets — including some Riot, Marathon, and CleanSpark legacy rigs, plus a substantial private-operator long tail — must now either negotiate below-average power rates, curtail aggressively to avoid running during low-margin hours, or accept ongoing operating losses on the older portion of the fleet. The threshold is moving steadily upward as difficulty climbs further.
The math is unforgiving. A typical pre-S19 rig — an S17, S15, T17, or older — operates somewhere in the 40-to-70 J/TH efficiency range. At an industrial power rate of 5 cents per kilowatt-hour and current network difficulty, those rigs produce mined Bitcoin at a marginal cost meaningfully above the spot price. Even operators with sub-3-cent power, common in stranded-renewable and methane-capture sites, are barely break-even on the oldest classes of hardware. The dusty back-row racks that many sites have kept running because the marginal hashrate was free-and-clear of capex are now actively destroying value, and operations teams have been quietly retiring sections of legacy hardware that until recently were treated as zero-marginal-cost contributors to total fleet output.
The dynamics have accelerated the secondhand ASIC market dramatically. Pre-S19 units are trading on platforms like Kaboomracks and the Compass marketplace at fractions of replacement cost, with many lots flowing to operators in lower-cost-of-electricity geographies — Pakistan, Russia, parts of Latin America, Ethiopia. The PCB-and-asic supply chain that once flowed primarily from Bitmain factories to North American hosting sites now has a vigorous secondary backflow channel, with thousands of units per week moving across borders and onto the books of operators willing to run on margins that public miners cannot accept. The logistics infrastructure around that flow — bonded-warehouse arrangements, payment escrow, condition certification — has matured into a small industry of its own.
For the public miners specifically, the pressure has accelerated decisions on fleet refresh. CleanSpark and Riot have both been aggressive in retiring legacy hardware and replacing it with S21 and (where available) S23-class rigs. MARA's parallel immersion-cooling conversion serves a related purpose: pushing the J/TH of existing S19 fleets lower through better thermal headroom, which extends the operational life of hardware that would otherwise be at the margin of profitability. The capex required for fleet refresh has been a meaningful drag on free cash flow across the cohort, and the funding mix between equity issuance, hardware financing, and operating cash flow has become a closely watched detail at every quarterly call.
The pattern is the textbook post-halving consolidation cycle that has played out across every previous Bitcoin halving, just compressed in time and accelerated by aggressive AI hyperscaler demand for the same power slots that older rigs used to occupy economically. The 2016 cycle saw a similar wave of S5 and S7 retirements; the 2020 cycle saw the S9 fleet pushed into the same value-destruction zone that pre-S19 hardware now occupies. What is different this time is the scale: the absolute number of pre-S19 rigs in the global fleet is several times what was in service during prior halvings, and the secondhand price discovery is happening in real time on transparent marketplaces rather than through opaque OTC channels.
For investors, the read-through is straightforward. Public miner cost curves should compress over the next two to four quarters as the slowest-to-refresh operators either retire legacy hardware or write it down. Realized hashrate per energized megawatt should rise as fleet vintages improve. Marginal production cost per BTC should fall. The miners that emerge from this consolidation phase will have meaningfully better unit economics; the miners that fail to refresh will be increasingly squeezed regardless of the BTC price tape. Watch the public-miner Q1 earnings prints for explicit fleet-age disclosures and any one-time impairment charges, which will signal whether managements are willing to take the cost of recognition rather than letting the older hardware quietly drag on margins.