NVIDIA's relationship with crypto mining has gone through several twists. Visit the Binance Official Site to check spot prices for major cryptocurrencies, download the Official Binance App, and iPhone users should check the iOS Install Guide first. The short answer: 2017-2022 was a "honeymoon period" between NVIDIA and crypto, with Ethereum PoW mining serving as the second-largest source of demand for NVIDIA's consumer GPUs; after the Ethereum Merge in September 2022, mining demand dropped to zero overnight; from 2023 to 2026, NVIDIA's relationship with the crypto space has shifted toward demand for AI training GPUs, which has indirectly fueled interest in AI-themed tokens like Bittensor and Render.
Before the Merge in September 2022, Ethereum ran on PoW consensus, with miners using GPUs to compute Ethash hashes. The stronger the GPU's compute power and the larger its memory, the higher the mining efficiency.
NVIDIA's consumer cards — the GeForce GTX 1060/1070, RTX 2060/2070, and RTX 3060/3070/3080 — were the workhorses of Ethereum mining.
| Period | Dominant mining GPU | Daily earnings per card (USD) |
|---|---|---|
| Q4 2017 | GTX 1070 | 5-10 |
| Q1 2018 | GTX 1080 Ti | 15-20 |
| Q4 2020 | RTX 3070 | 8-12 |
| Q3 2021 | RTX 3080 | 15-25 |
| Q1 2022 | RTX 3090 | 10-15 |
| Q3 2022 (pre-Merge) | RTX 3090 | 5-8 |
After the RTX 30 series launched in 2021, the mining boom caused severe GPU shortages. The RTX 3080's suggested retail price was $699, but it was scalped for $1,500-2,000 on the secondhand market. The "GPU tug-of-war" between gamers and miners lasted the entire year, and NVIDIA's revenue grew substantially as a result.
In May 2021, NVIDIA introduced LHR (Lite Hash Rate) technology, capping the Ethereum mining hash rate of new RTX 30-series batches at 50% of the full rate. LHR was meant to let gamers actually buy GPUs while pushing miners toward dedicated CMP (Cryptocurrency Mining Processor) cards.
But in May 2022, LHR was cracked by the mining community (via a security flaw in NVIDIA's official driver v512.59). Once cracked, LHR cards could recover 100% of their mining performance. NVIDIA's anti-mining effort largely failed.
In 2021, NVIDIA launched the CMP series — dedicated mining cards including the CMP 30HX, CMP 40HX, CMP 50HX, and CMP 90HX. These cards have no video output ports and don't support gaming; they were built purely for mining.
The CMP series generated roughly $500 million in sales during 2021-2022, according to NVIDIA's own financial disclosures. But after the Ethereum Merge, CMP cards became worthless overnight, and NVIDIA acknowledged the collapse of the CMP business in its Q4 2022 earnings report.
The Ethereum Merge completed on September 15, 2022, shifting consensus from PoW to PoS. GPU mining vanished from the mainstream instantly.
| Impact | Data |
|---|---|
| GPU miner daily earnings | Fell from $8 to $0.50 |
| Used RTX 30-series prices | Fell from $1,500 to $400 |
| Global GPU mining hash power | Down more than 80% |
| NVIDIA Q4 2022 revenue | Down 21% year over year |
After the Merge, some miners shifted to other PoW coins (Ravencoin, Ergo, Conflux), but these coins have small total market caps and mining profits nowhere near what Ethereum once offered. Large numbers of mining farms shut down, flooding the market with used GPUs.
NVIDIA's earnings reflected this turning point. In the second half of 2022, NVIDIA's gaming segment revenue fell more than 50% year over year, partly because miners stopped hoarding cards and the secondhand market undercut demand for new ones.
In 2023, ChatGPT ignited the AI boom, sending demand for NVIDIA's data center GPUs like the H100 and A100 soaring. NVIDIA's focus shifted from consumer GPUs and gaming toward data center GPUs and AI.
| NVIDIA business mix | FY2022 | FY2025 |
|---|---|---|
| Data center | 39% | 87% |
| Gaming | 33% | 9% |
| Professional visualization | 7% | 2% |
| Automotive | 2% | 1% |
| OEM | 19% | 1% |
NVIDIA's data center revenue grew from $15 billion in 2022 to $128 billion in 2025 — a 750% increase over three years. This "AI GPU revolution" pushed NVIDIA's market cap from $500 billion to $3.5 trillion (its 2024 peak), making it one of the most valuable companies in the world.
AI-themed tokens (Bittensor, Render, Akash, io.net) are highly correlated with NVIDIA's stock price.
| Period | NVIDIA stock | TAO price | RNDR price |
|---|---|---|---|
| Jan 2023 | $150 | $100 | $0.60 |
| Mar 2024 (peak) | $950 | $750 | $13 |
| Sep 2024 | $110 (post-split equivalent of $1,100) | $350 | $6 |
| Apr 2025 | $105 | $600 | $5 |
| Apr 2026 | $125 | $900 | $3.50 |
The correlation coefficient between NVIDIA's stock price and AI-themed tokens sits around 0.6-0.7, showing strong synchronization. This reflects the fact that both markets are driven by "AI investment sentiment" — AI-themed tokens are, in essence, a "high-beta asset" tied to NVIDIA.
Decentralized GPU networks such as Render Network, Akash Network, and io.net don't threaten NVIDIA's core business. There are two reasons.
First, the total GPU capacity of these networks is a tiny slice of the global AI compute market (under 1%). NVIDIA's main customers are hyperscale data centers run by Microsoft, Google, Meta, and Amazon.
Second, the hardware powering these decentralized GPU networks still comes from NVIDIA. Nodes on Render run NVIDIA RTX 4090s, and nodes on Akash run NVIDIA H100s and A100s. If anything, these networks are just another sales channel for NVIDIA GPUs.
NVIDIA leadership has mentioned "decentralized GPU networks" as an interesting direction within crypto in several public talks between 2023 and 2025, but without committing to any investment or formal partnership.
NVIDIA's Inception program supports its AI startup ecosystem, covering more than 15,000 AI startups worldwide. Some crypto-plus-AI projects are included — Bittensor, for example, gained early access to NVIDIA Inception.
However, NVIDIA Inception is a free incubator program and does not involve investment or token purchases. NVIDIA holds no crypto positions and runs no mining business.
CUDA is NVIDIA's parallel computing API and the de facto standard for GPU programming. Crypto mining, zero-knowledge proofs, and AI model training all rely on CUDA.
This "software moat" gives NVIDIA a special position in crypto infrastructure. Even without mining demand, CUDA remains foundational for prover nodes on ZK rollups (zkSync, Polygon zkEVM) and for the hardware layer behind AI-crypto projects like Bittensor and Render.
In 2024, NVIDIA CEO Jensen Huang introduced the "AI Factory" narrative — framing data centers as the factories of the 21st century, with GPUs as their "machine tools." This narrative moves entirely away from crypto, repositioning NVIDIA's story as "the compute supplier for the AI industrial revolution."
Even so, the crypto market still treats NVIDIA's stock price as a bellwether for the AI sector: NVIDIA up means AI-themed tokens up, NVIDIA down means AI-themed tokens down, and that correlation has held throughout 2024-2026.
If you're bullish on AI's long-term growth, buying NVIDIA stock (NVDA) is the most direct option. NVDA is a compliant, deeply liquid, fundamentally supported asset (with a P/E ratio of roughly 30-50) in the traditional markets.
AI-themed tokens are a "high-beta alternative" to NVDA. When NVDA rises 10%, AI-themed tokens might rise 30-50% — and the same applies on the downside. That high beta makes AI-themed tokens suitable for short-term speculation, but not for a long-term core holding.
Q: Can I still mine Ethereum with a GPU?
No. After the 2022 Merge, Ethereum moved fully to PoS, so GPU mining is no longer possible on it. GPU miners can only mine other PoW coins (Ravencoin, Ergo), but the returns are far below what Ethereum once offered.
Q: Does NVIDIA hold Bitcoin or Ethereum?
According to public filings, NVIDIA holds no crypto positions at all. NVIDIA's stance is "provide the tools but don't participate," which contrasts sharply with Tesla (which once held BTC) and MicroStrategy (which holds a large BTC position).
Q: Which is a better buy — NVIDIA stock or crypto?
They serve different roles. NVDA is a compliant stock with deep liquidity and cash-flow support. Crypto is high-beta, highly volatile, and highly elastic. The two can be combined in a portfolio; the right choice depends on individual risk tolerance.
Q: Will decentralized GPU networks replace AWS?
Not in the near term. AWS and Azure hold significant advantages in scheduling efficiency, network bandwidth, and technical support for large-scale AI training. Decentralized GPU networks mainly serve a price-sensitive, moderate-performance long-tail market.
Q: What's the difference between the NVIDIA H100 and the consumer RTX 4090?
The H100 is a data-center-only GPU priced at $30,000-40,000, optimized specifically for AI training with HBM3 memory and high-bandwidth NVLink. The RTX 4090 is a consumer GPU priced at $1,500-2,000, used mainly for gaming and small-scale AI inference. Their performance gap is 5-10x.
Q: How do you estimate an AI token's "NVIDIA content"?
Look at the actual GPU types and quantities the project uses. Bittensor subnets train large models on NVIDIA H100s (high NVIDIA content), Render Network uses consumer RTX 4090s for rendering (moderate NVIDIA content), and SingularityNET is mostly an AI services marketplace (low hardware content).
Q: Does NVIDIA have plans to enter the crypto space?
Nothing public suggests it does. NVIDIA's long-term positioning is as an "AI compute supplier" and it does not directly engage in crypto business.
Q: How much did the Ethereum Merge affect NVIDIA?
The near-term impact was significant (Q4 2022 revenue fell 21%), but it was quickly offset by AI demand. Starting in 2023, the boom in NVIDIA's data center business fully covered the decline in its gaming segment.
AI-themed tokens are highly correlated with NVIDIA's stock price, and a pullback in NVIDIA shares can spill over into the AI token market. This article is compiled from public data and does not constitute investment advice. Please read our Disclaimer.
Data sources: NVIDIA earnings reports (Q1 FY22 - Q4 FY26), CoinGecko, and official data from various AI projects (April 2026).