In 2017, a top-tier ASIC mining rig delivered an 80% gross margin to its manufacturer. By 2025, that same margin had collapsed to 20–30%. The sales revenue of the entire industry—covering Bitmain, Shenma, and the rest—stayed flat at 300–400 billion yuan across three distinct cycles. The golden age is not a feeling. It is a data point.
I have seen this pattern before. In 2020, when I built a real-time monitoring dashboard to manage a $150,000 DeFi leverage strategy, the moment protocol yields began to compress, the smart money rotated. The same structural compression is now under way in the mining hardware business. Yang Zuoxing, the founder of Shenma Mining, stated this clearly in a recent industry address: the easy profits are gone. His credibility rests on over a decade of chip design experience and a front-row seat to every boom and bust since the first ASIC. But the data was already visible.
Context: The Three Cycles
Mining hardware sales have hovered around 300–400 billion yuan per cycle: 2017 (the ICO-fueled bull run), 2021 (the DeFi and NFT mania), and 2025 (the current post-halving landscape). The headline numbers look stable, but the underlying economics have deteriorated. Gross margins for mining rig manufacturers dropped from 80–90% in 2017 to 50–60% in 2021, and then to 20–30% in 2025. This is not a short-term correction. It is a secular decline driven by three forces: the Bitcoin halving (which cut block rewards by 50% in 2024), the physical ceiling on ASIC chip efficiency (J/TH is approaching the thermodynamic limit), and a new competitor for capital and electricity—AI.
Yang pointed out three potential escape routes: natural gas flaring capture, AI data center integration, and solar-powered mining. All are attempts to lower the variable cost that now dominates the P&L: energy. But none of these are scalable today. Natural gas mining works only near oil fields. Solar mining is intermittent. AI integration requires turning a single-purpose machine into a multi-purpose one—a feat of engineering that remains unproven in deployment.
Core: The Mechanics of Margin Compression
From my audit background, I know that when a system’s core parameter—in this case, gross margin—compresses by 60–70 percentage points over eight years, it is not a random walk. It is a structural failure. The mining hardware business is a proxy for the entire PoW mining industry. If the manufacturers cannot sustain their R&D spending on next-generation chips, the pace of efficiency gains will slow further. That creates a feedback loop: slower gains → higher power costs → lower profitability for miners → weaker demand for new machines.
I saw the same mechanics in the Terra/UST collapse. The algorithmic stablecoin’s design looked elegant on paper, but when I ran my Rust-based validator node to track the oracle feeds in real time, the fragility was obvious. The mint-redeem mechanism assumed infinite demand for arbitrage; the moment that assumption failed, the peg broke. Mining hardware manufacturers face a similar assumption: that there will always be enough marginal electricity at low enough prices to support the next generation of machines. Yang’s data suggests that assumption is eroding.
AI is the most disruptive variable. Yang explicitly stated that AI is competing for both capital and power. In the 2021 cycle, mining farms could bid up power prices because Bitcoin was above $60,000. In 2025, with Bitcoin trading lower and AI data centers offering predictable long-term contracts, utilities favor the AI tenant. Miners are left with the scraps—off-peak hours, curtailled energy, or stranded natural gas.
Contrarian: The New Directions Are Not a Panacea
The market narrative will spin Yang’s three new directions as a lifeline. I am skeptical. In 2021, I executed a bot-driven arbitrage on Bored Ape Yacht Club NFTs, buying five at a $150,000 average floor and selling during the FOMO peak for a 300% markup. But when the market corrected in late 2022, I liquidated at a 60% loss. The lesson: liquidity is an illusion during stress. The same applies to natural gas mining, solar mining, and AI integration. These are not liquid, proven revenue streams; they are experimental capital expenditures.
Consider the AI integration path. Turning an ASIC miner into a co-located AI accelerator requires redesigning power delivery, cooling, and compute logic. Even if it works technically, the economic viability is unknown. The mining hardware companies have no moat in AI—they are competing with NVIDIA and AMD, who have years of software stack maturity. The solar mining route depends on battery storage costs, which remain high. Natural gas flaring capture is the most mature, but its scale is limited to oil-producing regions.
The contrarian truth is that the golden age ended because the industry’s core value proposition—providing cheap, specialized computation for PoW—has been commoditized. The three new directions are attempts to escape commoditization, but they will likely take years to bear fruit, if they ever do.
Takeaway: Trade the Structure, Not the Story
Yang’s message is not a call to abandon mining. It is a warning to those who treat mining hardware as a safe store of value. The data on margins is clear. The competition from AI is structural. The three escape routes are long shots.
I trade the structure, not the story. Trust is a variable I solve for, never assume. The market doesn’t owe you an exit, only a price. If you are a miner, hedge your power costs or prepare for a long winter. If you are an investor, examine the power purchase agreements, not the keynote slides. The golden age is over, but the long tail will persist. The survivors will be those with the lowest cost of electricity, the highest chip efficiency, and the discipline to ignore hype.