Hook

NVIDIA just dropped a quarter of nearly $100 billion in revenue – and growth is accelerating. But if you’re a crypto miner running a rig of RTX 4090s, this isn’t a victory lap. It’s a tombstone.

The same silicon that drove the 2021 mining frenzy is now being hoovered up by hyperscalers building AI clusters. The ledger does not lie, but the CEOs do – Jensen Huang isn’t selling you GPUs for mining; he’s selling dreams to enterprises. I’ve watched this pivot live, from the 2018 ETC hash rate sprint to the Uniswap liquidity blitz. Now the game has changed: NVIDIA is no longer a pickaxe seller for the crypto gold rush. It’s a sovereign weapon supplier for the AI arms race.

Context

Let’s rewind. From 2017 to 2022, NVIDIA’s gaming GPU division inadvertently powered the Proof-of-Work ecosystem. Ethereum alone consumed millions of cards. When the Merge killed PoW on Ethereum, the narrative shifted: miners would flock to altcoins like Ravencoin or Kaspa. But a quieter, more powerful trend was already underway – the AI boom.

By 2023, NVIDIA’s Data Center revenue had overtaken Gaming. Today, it accounts for well over 90% of total revenue. The H100, B200, and soon Blackwell are designed for training large language models, not Sha-256. The architecture is different: Tensor Cores, high-bandwidth memory (HBM3e), and CoWoS packaging are optimized for matrix multiplication, not hashing. Speed is the only hedge in a zero-latency market, and NVIDIA is moving at AI speed, leaving miners in the dust.

Core

The Numbers That Matter

• $100B run-rate: NVIDIA’s quarterly revenue is on track to exceed $100B annualized. That’s Google + Amazon + Microsoft combined in AI infrastructure spending.
• Growth acceleration: The company guided for sequential growth, implying supply bottlenecks have been cleared – specifically CoWoS advanced packaging at TSMC.
• Gross margin ~75%: This isn’t a commodity business. NVIDIA holds near-monopoly pricing power.

What This Means for Crypto Mining

The same CoWoS capacity that enables B200 dies is also the bottleneck for any future GPU-based miner. TSMC can only produce so many interposers. Every wafer allocated to NVIDIA’s AI chips is a wafer not available for consumer-grade GPUs that could be used for mining. Even the RTX 5090 (expected 2025) will face allocation constraints because the H100 family consumes the most advanced nodes.

I’ve been tracking hardware availability since the 2020 Uniswap V2 liquidity mining days. Back then, I deployed my own $5,000 to test yields. Now I run automated bots that scrape distributor inventory for GPU prices. The data is clear:

• Retail GPU prices remain elevated despite the mining downturn. A used RTX 4090 is still $1,500+ – because AI researchers and small-scale inference shops buy them up.
• New mining-specific GPUs (e.g., CMP series) are dead. NVIDIA doesn’t want to be seen as pro-mining; they even lobbied for PoW restrictions.
• PoW coins that rely on GPU mining (Ravencoin, Beam, Ergo) have seen hash rate drop 30-50% since 2022. The remaining miners are running on cost-sunk hardware.

The Supply Chain Reality

From the semiconductor analysis: NVIDIA’s “growth acceleration” is a direct signal that TSMC’s CoWoS capacity has expanded beyond previous fears. But this capacity is spoken for. Intermediaries are just slow nodes in the network – TSMC, SK Hynix, Samsung – all aligned to serve NVIDIA first. Any spare capacity is negligible.

For proof, look at the HBM market. HBM3e is sold out for 2024-2025. GPU dies need HBM; without it, even if you had dies, you can’t make a finished product. This vertical integration locks out mining.

Contrarian

The prevailing view among crypto optimists is: “AI demand will eventually saturate, GPUs will flood back to mining, and we’ll have a cheap second-hand market.” I think that’s wishful thinking.

First, the AI industry is still early. Enterprise AI, sovereign AI (nation-state clusters), and edge AI are just beginning. The total addressable market dwarfs crypto mining by orders of magnitude. Volatility is the price of admission, not the exit – but in this case, the volatility is on the side of AI, not crypto.

Second, even if AI demand slows, NVIDIA will repurpose that capacity for automotive, robotics, or metaverse – not for mining. Huang has explicitly de-prioritized crypto. The company’s investor relations materials no longer mention blockchain. Consensus is fragile until it becomes irreversible – and here, consensus is that NVIDIA is an AI company.

Third, mining is moving to ASICs. For PoW coins that matter (Bitcoin, Litecoin), ASICs already dominate. For GPU-mineable coins, the future is either niche or Proof-of-Stake. New projects like Kaspa use custom algorithms that favor ASICs anyway. The window for profitable GPU mining is closing.

Takeaway

Crypto builders need to stop assuming GPUs will always be abundant and cheap. If you’re designing a protocol that relies on GPU computation (ZK proofs, AI inference on chain, FHE), you must plan for hardware scarcity. Action precedes analysis in the eyes of the mover – start experimenting with FPGA or specialized accelerators now.

NVIDIA’s quarter isn’t just a financial milestone. It’s a clear signal that the compute layer of the global economy is being rewired for AI. Crypto either rides on top of that layer or gets crushed by it. The block explorer reveals what the headline hides – and this headline hides a slow death for GPU mining.