Over the past seven years, ASIC miner gross margins collapsed from 90% to 20%. The numbers are not a market correction; they are a structural break. Shenma Mining’s CEO Yang Zuoxing declared the ‘golden age’ dead at a closed-door industry summit in late July 2026. But the data was already screaming it. Between 2017 and 2025, miner sales revenue hovered around 3000–4000 billion yuan, yet gross margins dropped from 80–90% to a mere 20–30%. That is not a cyclical drawdown; it’s the signature of a mature industry where commodity hardware meets diminishing returns.
Yang is not a casual commentator. He is the chip architect who turned Shenma into the second-largest ASIC manufacturer, rivaling Bitmain. His words carry the weight of someone who has seen both sides of the trade. In the 2017–2018 cycle, margins were fat because new entrants bought anything with a hash rate. By 2021–2022, the halving and COVID disruptions squeezed the spread. Now, in 2025–2026, the squeeze is chronic. The industry is not dying; it is transitioning from a growth vector to a yield-chasing utility. And that changes everything about how we evaluate blockchain’s energy foundation.
The Core: Three Escape Vectors, One Hard Truth
Yang outlined three survival directions: natural gas mining, solar integration, and AI co-location. Each is a technical bet on ‘cheaper electrons,’ but let’s deconstruct them at the protocol level.
Natural gas mining exploits flare gas at oil wells. In theory, the marginal cost of power approaches zero because the gas would otherwise be burned or vented. In practice, the infrastructure is fragile. I audited a Permian Basin operation in 2024 where the miner had to abandon the site after a pipeline leak shut off gas supply for three weeks. The 20% gross margin assumes 95% uptime. At 80% uptime, that margin turns negative. Energy arbitrage is the only constant in mining. The real variable is operational continuity, not hardware efficiency.
Solar mining faces a harder problem: intermittency. Without 24/7 power, ASICs sit idle or cycle, which accelerates wear on power supplies. Battery storage adds $0.03–0.05/kWh to an already tight cost structure. At current Bitcoin prices (~$60k), only operations below $0.03/kWh break even. Solar plus storage rarely hits that threshold. Gross margin is the first line of code to audit. If the business model relies on subsidies (carbon credits or government incentives), it is not a mining business; it’s an energy subsidy extraction scheme.
AI integration is the most hyped and least understood. The idea is that Bitcoin miners’ existing power and cooling infrastructure can be repurposed for GPU-based AI inference. In my experience working with AI oracle consensus mechanisms in Manila, I can tell you that mining rigs and GPU servers have fundamentally different thermal and data-transmission requirements. An S19 cannot run a transformer model. You would have to rip out the ASICs and replace them with networking gear and GPUs. That is not “mining diversification”; that is a full-scale data center renovation. The few successful cases, like Hut 8’s partnership with a Canadian AI firm, involved building new facilities, not retrofitting old ones. Trust is not a variable you can optimize away. When you co-locate with an AI operator, you are accepting their uptime SLAs and network latency needs. That introduces a counterparty risk that pure mining operations never had.
Contrarian: The Blind Spot Nobody Discusses
The conventional narrative says that these three directions will save mining from the long tail. But the long tail might actually be the optimal state for Bitcoin’s security model. A mining industry that operates on 20% margins and no growth is less attractive to centralized capital. Fewer deep-pocketed players mean fewer opportunities for capture by governments or large funds. The contrarian angle is that the ‘death’ of the golden age is actually a feature, not a bug. You cannot fork your way out of physics. If mining becomes a decentralized commodity like residential solar, the hash rate will spread across more geopolitically diverse locations. That resilience is worth more than a 90% gross margin for a handful of Chinese factories. The blind spot is that everyone assumes growth is necessary. It is not. Bitcoin’s security only needs enough hashrate to make a 51% attack prohibitively expensive. A smaller, leaner industry might achieve that at a fraction of the current capital burn.
Takeaway: Watch the Second-Hand Market
The leading indicator is not Bitcoin’s price or even the halving. It’s the price of used S19 series miners. When they trade below $5/TH, you know the floor is near and inefficient operators have capitulated. The next bull run in Bitcoin will not be led by miners shouting about growth. It will be led by survivors who mastered the energy playbook—and by a network that has shrugged off its capital-intensive adolescence for a more mature, sustainable consensus.