Author: Hu Tao, ChainCatcher
On September 15, 2022, Ethereum completed the largest and riskiest technical upgrade in cryptocurrency history—"The Merge"—officially transitioning from the Proof of Work (PoW) consensus mechanism to Proof of Stake (PoS). This upgrade not only reduced Ethereum's energy consumption by approximately 99.95%, but also completely changed the issuance of ETH and the network security mechanism.
However, nearly four years have passed since this upgrade, which many regard as a historic turning point for Ethereum, and discussions about whether it was worth it have recently heated up again.
The core idea is not complicated: What would it look like if Ethereum had not completed The Merge back then, but instead continued along the PoW route to 2026?
Today, when GPU computing power has become a core strategic resource in the AI era, a highly imaginative counterfactual scenario is: Could those millions of GPUs originally used for ETH mining be reorganized to form a global distributed computing network? Would PoW instead give Ethereum another strategic position in the AI era? And for ETH, would continuous miner selling pressure and higher issuance become unavoidable costs?
1. A Potential Global Distributed GPU Computing Network
If Ethereum had stuck with PoW, one of the most direct assets left behind would be the massive GPU computing infrastructure built around mining machines.
Unlike Bitcoin's highly specialized ASIC miners, Ethereum's historical GPU miners have stronger general-purpose computing capabilities. Theoretically, if Ethereum had continued using PoW and gradually established mechanisms for computing scheduling, task verification, and revenue distribution at the protocol layer or in ecosystem applications, these GPUs would have the opportunity to transform from devices solely securing the network into infrastructure for AI, rendering, HPC, and other computing tasks.
Jademont, co-founder of Waterdrip Capital, stated clearly on X that he has always believed the transition from PoW to PoS was a severely overrated decision. It did reduce energy consumption, but the cost savings are almost negligible compared to the development opportunities lost.
"If Ethereum had continued evolving along PoW and persistently promoted the development of computing infrastructure, it could have grown into the world's largest blockchain-based AI computing network, occupying a more important strategic position in the AI era. The shift to PoS prematurely ended that possibility."
Crypto KOL Enheng also said: "If ETH had continued with PoW, it might have one of the largest GPU computing pools in the world. Imagine if millions of ETH mining GPUs had not exited back then, but instead were connected to the AI computing market through protocols. The Ethereum ecosystem could have formed a decentralized AI cloud powered by global miners, similar to a decentralized version of AWS GPU Cloud."
Logically, this is not far-fetched. Bitcoin's ASIC miners are highly specialized and difficult to repurpose; Ethereum's GPU miners have general-purpose computing potential. If the protocol layer could design incentive mechanisms to allow miners to rent out computing power during idle time or beyond block production, or to marketize computing power through sidechains or application layers, Ethereum could have built a unique bridge between "blockchain security" and "AI infrastructure." The decentralized computing projects, rendering networks, and distributed training attempts we see today might have emerged earlier and on a larger scale on the Ethereum mainnet or within its ecosystem.
Of course, practical constraints also exist: miners' interests do not naturally align with AI customers' needs, and issues such as network latency, task scheduling, data privacy, and verification mechanisms need to be solved. But starting with "one of the world's largest GPU pools" would indeed leave more room for imagination in Ethereum's AI strategy.
2. Direct Impact on Coin Price and Selling Pressure
From a purely supply-demand and price perspective, sticking with PoW is not without costs.
0xTodd, co-founder of Ebunker, said that in the late PoW era, Ethereum mined approximately 13,000 ETH per day. At a common price of $2,000 at the time, that would be about $26 million in daily selling pressure. After the transition to PoS, production dropped sharply to about 3,000 ETH, reducing selling pressure to about $6 million, only a quarter of the original.
This "super halving" is seen by some as one of the true core motivations behind the Merge—the Foundation was more concerned with reducing the flow of new coins to miners and mining machine manufacturers than with electricity consumption alone (especially the surplus hydro and thermal power in southwest and northwest China at the time). The difficulty bomb was designed as a "red button" as early as 2013–2014, precisely to prevent inertia from keeping Ethereum on PoW forever.
Therefore, without the transition to PoS, Ethereum would have long suffered from higher inflation and selling pressure from miners. In a bull market, hashrate and coin price could form a positive feedback loop, with miners more willing to hold; in a bear market, it could exacerbate the decline. In contrast, the low issuance under PoS is closer to the "digital gold" narrative and more likely to attract long-term capital and institutional staking. While coin price volatility may be higher, long-term deflationary expectations would also be weaker.
"Even now, many smart alecks still think Ethereum's transition to PoS was its biggest blunder. Look at its big brother ETC. After the hard fork, ETC, which has continued to use the PoW mechanism, has seen its market cap drop from the top 20 before Ethereum's Merge to 64th now, a decline of over 80%." said crypto KOL yyy.
Although ETC and ETH differ greatly in ecosystem, developers, and brand, and cannot be simply compared, it at least shows that merely "continuing PoW" is not sufficient to guarantee better market cap performance.
III. Scaling, the World Computer Narrative, and Technical Evolution Difficulty
If GPUs and coin price are the potential advantages left by PoW, then scaling might become its biggest structural constraint.
Ethereum has long aspired to be an open "world computer," but today's development direction has shifted from simply increasing L1 transaction throughput to a scaling system composed of L1, L2, and data availability. In recent years, the continuous advancement of Blob, Rollup, and subsequent protocol upgrades has essentially been about reducing the computational and data costs of the entire Ethereum ecosystem.
PoW is not absolutely incapable of achieving scaling, but its consensus mechanism would make some protocol designs more constrained. PoW networks rely on hashrate competition to ensure security, and block propagation, block time, and on-chain state growth must all consider miners' hardware thresholds and network conditions. In contrast, PoS allows for closer protocol-level coordination among validator sets, finality, and data availability mechanisms, providing greater design space for further Ethereum scaling.
More importantly, PoS reduces the real-world resource costs required to maintain Ethereum's security, allowing the protocol to allocate more resources to L2, data availability, and user experience.
The energy issue is also unavoidable.
Even if PoW Ethereum could use a large amount of renewable energy or low-cost electricity in some regions, its high energy consumption itself could still become an issue that needs to be explained in terms of institutional adoption and regulatory concerns. For a public blockchain that aims to carry stablecoins, RWA, and traditional financial assets, PoS is clearly more compatible with institutional considerations regarding energy, ESG, and compliance risks.
Therefore, if Ethereum had continued with PoW, it might have retained a massive GPU industry, but it might also have had to pay a higher price in terms of scaling efficiency, energy costs, and institutional adoption.
IV. Conclusion: History Has No Ifs, but Hypotheses Can Illuminate Reality
Ethereum's transition from PoW to PoS was both a proactive technical and economic choice and a path predetermined by the "difficulty bomb" in its early design. It gave up the possibility of a certain distributed hashrate network in exchange for lower inflation, higher capital efficiency, and smoother scaling space. The reason this discussion has resurfaced today is that AI has made "hashrate" itself a new strategic resource, and those GPUs from back then happen to be versatile.
If Ethereum had not transitioned to PoS, it might indeed have possessed a large-scale distributed GPU network and, after the rise of the AI computing market, obtained a completely different development path; ETH might also have had a stronger PoW "digital commodity" attribute.
But at the same time, higher issuance, miner selling pressure, scaling constraints, and energy and institutional adoption issues could also become the ceiling for its long-term development.
Therefore, this counterfactual game is ultimately not about "which is more advanced, PoW or PoS," but about two completely different development paths:
Continuing with PoW, Ethereum might have become a digital commodity network with smart contract capabilities; transitioning to PoS has allowed it to gain the potential to become a global open settlement and digital economy infrastructure.
And the arrival of the AI era has simply made this already concluded choice interesting again—because the large number of GPUs that were eliminated back then are now precisely among the world's scarcest computing resources.
History cannot be replayed, nor can it be proven that another path would have been better. But at least it is certain that Ethereum's choice of PoS back then was not a simple "technical upgrade," but a long-term strategic bet on energy, security, capital efficiency, computing power, and the future direction of the industry.
Now the real question is no longer "where did the GPUs go back then," but whether Ethereum can leverage the path it has chosen to capture enough economic value in the development of AI, stablecoins, RWA, and global on-chain finance.
