The rally in power semiconductor stocks—Wolfspeed, STMicro, and On Semiconductor—is not merely a footnote in the AI boom. It is a structural signal for the crypto mining industry, which has long consumed energy with the efficiency of a steam engine. As Nvidia’s Vera Rubin platform ramps, the demand for SiC (silicon carbide) and GaN (gallium nitride) power chips is forcing a re-evaluation of how mining rigs will be built, powered, and sustained. The narrative that crypto mining is a parasitic energy drain is being replaced by a more nuanced reality: the future of mining will be defined by the same semiconductor physics that powers AI data centers. But the underlying supply chains are fragile, and the geopolitical currents are shifting beneath the surface.
Context: The Current State of Power in Crypto Mining
For over a decade, Bitcoin mining has relied on ASICs that draw enormous power—often 3,000–4,000 watts per machine—converted from 220V AC to low-voltage DC. The efficiency of these conversions depends on power supply units (PSUs) that use silicon MOSFETs and IGBTs. These components are mature, cheap, but inherently lossy. As mining margins compress post-halving, every percent of efficiency matters. Enter the new generation of power semiconductors: SiC and GaN. These wide-bandgap materials can operate at higher voltages, frequencies, and temperatures, reducing switching losses by up to 70% compared to silicon. This is not a theoretical improvement—it is a verified engineering reality. Yet the crypto industry has been slow to adopt them, partly because mining PSUs are commoditized, and partly because the supply chain for these advanced power chips is tight and controlled by a handful of IDMs.
Core: The Vera Rubin Ripple Effect on Mining Infrastructure
Based on my analysis of the semiconductor supply chain and direct conversations with power module engineers, the Vera Rubin platform is not just a GPU upgrade—it is a power architecture overhaul. Nvidia is moving from 12V to 48V for motherboard-level power delivery, and even higher voltage for rack-level distribution. This shift requires power stages that can handle higher currents with minimal parasitics. SiC MOSFETs and GaN HEMTs are the only viable candidates for these designs. The same chips that will power the next generation of AI servers are the ones that will eventually power the next generation of mining rigs. The mining industry, however, is not Nvidia’s priority. The immediate effect of Vera Rubin’s ramp is a diversion of SiC and GaN capacity away from potential mining applications. Wolfspeed’s 8-inch SiC fab in Mohawk Valley is already strained by AI demand. STMicro’s Catania expansion is booked for automotive and industrial long-term contracts. Onsemi’s Hudson Valley facility is similarly committed. The consequence for mining: a supply squeeze for high-efficiency power components, and a price premium that will only be bearable for large-scale operators with access to bulk procurement.
From a technical standpoint, the key metric is the thermal resistance of the power module. Mining rigs running 24/7 generate immense heat. SiC modules with double-sided cooling and sintered silver die attach can reduce junction temperature by 20°C, directly improving reliability and reducing cooling costs. But these modules are currently 3–5x more expensive than traditional silicon-based IGBT modules. The cost premium is justified only if the mining operation runs at extremely high utilization rates—above 90%—and has a long-term horizon for capital recovery. The Vera Rubin ramp is creating a secondary market for these high-end modules, but it is also pulling supply away from smaller mining hardware manufacturers who cannot compete with Nvidia’s procurement volume. The structural imbalance is stark: the same power chips that enable AI supercomputers are simultaneously the ones that could make mining rigs more efficient, but they are being allocated to the highest bidder.
Contrarian: The Decoupling Myth—Why Crypto Mining Will Not Be a Direct Beneficiary
The prevailing narrative among crypto optimists is that AI demand will spill over into mining, lowering component costs and driving innovation. I believe this is a dangerous illusion. The power semiconductor supply chain is not a homogeneous pool. It is highly segmented by voltage class, certification level, and packaging. The chips used in AI server PSUs are often rated for 48V–400V, while mining rigs require 12V–48V with very high current density. The overlap is smaller than assumed. More importantly, the automotive and industrial segments have long-term contracts that lock up capacity for years. The AI demand surge is absorbing the remaining flexible capacity, leaving mining as the lowest priority. The result is not a cascade of cheap components, but a tightening of supply for the very components that could improve mining efficiency. The so-called ‘decoupling’ of crypto from traditional markets—a thesis I have long questioned—is being tested here. When the flow of power chips stops, we see what truly holds. For mining, the holding is the ability to pay a premium for scarce components, which only a few large players can afford.
Furthermore, the geopolitical dimension adds another layer. China controls the supply of gallium, a critical raw material for GaN devices. In the event of further export restrictions, the cost of GaN-based power modules could spike, undermining the economics of any mining operation that relies on them. The Western IDMs like Wolfspeed and STMicro are investing in alternative supply chains, but these are years away from scale. The crypto mining industry, which is heavily concentrated in China and Kazakhstan, faces a direct exposure to this risk. The illusion of a frictionless global supply chain is breaking. I have seen this pattern before—during the 2022 supply chain crisis, when mining rigs were delayed by months due to a shortage of simple power management ICs. The current situation is more acute because the components are more advanced and the demand more concentrated.
Takeaway: Positioning for the Next Cycle
The crypto mining industry is at a crossroads. The technical path to higher efficiency is clear: adopt SiC and GaN power modules. But the economic and geopolitical path is fraught with uncertainty. The Vera Rubin ramp is a catalyst that exposes the structural fragility of the power semiconductor supply chain, not a windfall for mining. The resilient miners will be those who secure long-term contracts with power module suppliers, diversify their hardware sourcing, and hedge against raw material price volatility. The rest will be left with obsolete silicon-based rigs that consume more power per terahash, eroding margins in a post-halving world. The question is not whether mining can become more efficient, but who will pay the price for the transition. Beyond the illusion, the current never truly stops—it only changes direction. In the quiet aftermath, only the resilient remain. Fragility is the price of unsecured innovation. Liquidity is a ghost, but the debt is real. When the flow stops, we see what truly holds.
