Ethereum completed The Merge on Sept. 15, 2022, shifting from proof-of-work to proof-of-stake in what was widely seen as one of the largest and riskiest technical upgrades in crypto history. The move cut the network’s energy consumption by about 99.95% and changed both ETH issuance and Ethereum’s security structure.
Now, nearly four years later, debate over that decision has flared up again. The question is simple on its face: if Ethereum had never completed The Merge and had remained on PoW through 2026, what would the network look like today?
A global distributed GPU network that never formed
The article’s central counterfactual starts with hardware. Unlike Bitcoin’s ASIC miners, which are highly specialized, Ethereum’s historical mining base relied on GPUs with broader computing utility. In theory, if Ethereum had stayed on PoW and gradually built mechanisms for compute scheduling, task verification and revenue distribution at the protocol or application layer, those machines could have moved beyond securing the chain and into AI, rendering and high-performance computing.
Waterdrip Capital co-founder Jademont said on X that he has long viewed Ethereum’s move from PoW to PoS as a severely overrated decision. In his view, the switch did lower energy consumption, but the savings were almost negligible compared with the development opportunities that disappeared with it.
He wrote: 「If Ethereum had continued evolving along the PoW path and kept building out its compute infrastructure, it could have become the world’s largest blockchain-based AI compute network and taken a much more important strategic position in the AI era. Moving to PoS ended that possibility early.」
Crypto KOL 嗯哼 made a similar point, arguing that if ETH had remained on PoW, it might have retained one of the largest GPU compute pools in the world. He described a scenario in which millions of GPUs that once mined ETH did not exit the market, but instead connected through protocol-level rails to the AI compute market, creating a decentralized AI cloud supplied by miners around the world, something akin to a decentralized version of AWS GPU Cloud.
The logic rests on the flexibility of the hardware. ASICs are hard to repurpose. GPUs are not. If Ethereum had been able to design incentives that let miners rent out idle compute outside block production, or if sidechains and application-layer systems had turned compute into a market, Ethereum might have built a bridge between blockchain security and AI infrastructure. The article suggests that decentralized compute projects, render networks and distributed training efforts that exist today might have appeared earlier, and at greater scale, inside Ethereum’s ecosystem.
That said, the piece does not treat this as a solved roadmap. Miner economics do not naturally line up with the needs of AI customers, and issues such as latency, scheduling, data privacy and verification would still need to be addressed. Even so, it argues that starting with one of the world’s largest GPU pools would have given Ethereum much more room to compete for strategic relevance in the AI era.
Lower issuance under PoS versus continued miner sell pressure under PoW
The article then turns to ETH supply and price pressure. Ebunker co-founder 0xTodd said Ethereum was producing roughly 13,000 ETH per day near the end of the PoW era. At a commonly referenced ETH price of $2,000, that translated into about $26 million in daily sell pressure. After the move to PoS, issuance dropped to around 3,000 ETH per day, bringing daily sell pressure down to about $6 million, roughly one quarter of the earlier level.
That steep decline in new issuance is presented by some as one of the real motivations behind The Merge. Under that view, the Ethereum Foundation was more focused on reducing the flow of newly issued ETH into the hands of miners and mining hardware vendors than on power consumption alone, especially at a time when surplus hydropower and thermal power existed in parts of southwest and northwest China. The article also notes that the “difficulty bomb” had been built into Ethereum’s design as early as 2013 to 2014, effectively as a red button meant to stop the chain from remaining on PoW indefinitely through inertia.
If Ethereum had stayed on PoW, the network would likely have carried higher inflation and more persistent miner selling pressure over a longer period. In bull markets, hash power and price could have reinforced each other, making miners more willing to hold. In bear markets, that same structure might have intensified downside pressure. By contrast, lower issuance under PoS fits more easily with a “digital gold” narrative and is more compatible with long-term capital and institutional staking. The article adds that long-run deflation expectations would have been weaker if PoW had continued.
Crypto KOL yyy pointed to Ethereum Classic as a rough market comparison. He wrote: 「Even now, plenty of people still think Ethereum moving to PoS was its biggest mistake. Look at its big brother ETC. ETC, which kept using PoW after the hard fork, has fallen from around the top 20 in market cap near the time of The Merge to 64th now, a drop of more than 80%.」
The article immediately cautions that ETC and ETH are too different in ecosystem depth, developer base and brand strength to support a direct one-to-one comparison. Still, it says the case at least shows that simply staying on PoW is not enough on its own to guarantee stronger market-cap performance.
Scaling constraints and the “world computer” path
If GPU infrastructure and commodity-like characteristics were the possible upside of staying on PoW, scaling may have been the biggest structural cost.
Ethereum has long aimed to become an open “world computer.” In practice, its development path has shifted from raising L1 throughput alone to building a scaling framework across L1, L2 and data availability. The article says ongoing work around blobs, rollups and later protocol upgrades is all part of reducing compute and data costs across the broader Ethereum ecosystem.
PoW does not make scaling impossible, but it places tighter constraints on protocol design. A PoW network relies on hash power competition for security, which means block propagation, block times and state growth all have to account for miners’ hardware thresholds and network conditions. PoS, by comparison, allows validator sets, finality and data-availability mechanisms to coordinate more tightly at the protocol level, giving Ethereum more room to pursue further scaling.
The article also argues that PoS lowers the real-world resource cost of maintaining security, leaving more room for the protocol to allocate resources toward L2, data availability and user experience. A PoW Ethereum might have preserved a large GPU industry around the chain, but likely at the price of lower efficiency in scaling.
Energy use and institutional adoption
Energy remains another unavoidable part of the trade-off. Even if a PoW version of Ethereum had leaned heavily on renewable energy or low-cost electricity in certain regions, high energy consumption itself could still have created complications in conversations with regulators and institutions.
For a public blockchain that wants to host stablecoins, real-world assets and traditional financial assets, PoS is presented as a cleaner fit for institutional concerns around energy, ESG and compliance risk. In that sense, continuing with PoW might have given Ethereum more room to retain a large GPU-centered industrial base, but it also could have imposed a higher cost in scaling efficiency, energy optics and institutional uptake.
Two different futures, not one better label
The article closes by framing Ethereum’s move from PoW to PoS as both a technical and economic choice, and as a path embedded early through the difficulty bomb. What Ethereum gave up was the possibility of a large-scale distributed compute network. What it gained was lower inflation, greater capital efficiency and a smoother scaling path. The reason the debate has resurfaced now is that AI has turned compute itself into a strategic resource, and the GPUs that once mined ETH happen to be general-purpose machines.
If Ethereum had never switched to PoS, the article says, it might indeed have ended up with a massive distributed GPU network and found a very different place in the AI compute market. ETH might also have had a stronger PoW-style “digital commodity” character. But the costs would have remained: higher issuance, ongoing miner sell pressure, scaling constraints, and unresolved questions tied to energy use and institutional adoption.
In that framing, the real issue is not whether PoW or PoS is “more advanced.” It is whether Ethereum would rather have become a digital commodity network with smart-contract capabilities, or a candidate for global settlement and digital economic infrastructure. The arrival of the AI era has simply made that old decision interesting again.
History cannot be replayed, and there is no way to prove that the road not taken would have been better. What can be said, the article argues, is that Ethereum’s move to PoS was never just a technical upgrade. It was a long-range strategic bet involving energy, security, capital efficiency, compute power and the future direction of the industry. The question now is no longer where those GPUs went, but whether Ethereum can extract enough economic value from the path it did choose across AI, stablecoins, RWAs and global on-chain finance.