New research work led by a top MIT researcher suggests that decentralization isn’t just a design choice but a principle of efficiency, where control breaks down as systems scale.
In the crypto industry, decentralization is loosely defined as the distribution of power and control across multiple independent participants, rather than a single central authority.
It means that no single entity—such as a company or government—can decide how the system operates, alter the rules, or halt transactions on its own.
“The first rule of control is observability. You can’t control something you don’t observe—and observability doesn’t scale,” Muriel Médard, co-founder and CEO of decentralized memory infrastructure firm Optimum, told Decrypt in an interview at TOKEN2049 in Singapore.
The question is no longer whether one agrees with or likes decentralization, but more about how “centralization doesn’t work once a system gets large enough,” Médard explained.
These ideas were first explored in Médard’s recent MIT study on wireless transmitters, which showed how distributing functions instead of centralizing them can make communication systems far more energy-efficient.
But Médard’s claim that decentralization operates more efficiently by default is less about what nature is and more about how it behaves.
Any statement about nature “is already subtracted from a fundamental lack of natural frames, such as in the equation, n-1,” Virgilio Rivas, professor of philosophy at the Polytechnic University of the Philippines, told Decrypt.
Rivas explains that systems organize themselves by removing any single center or fixed frame. In simple terms, nature works through connections and differences and “behaves in such a way that it can be observed” in its “real default.”
Still, Médard’s team at Optimum is now applying that same principle to blockchain networks, turning her research on distributed efficiency into code.
Their new network layer, tested on Ethereum’s Hoodi testnet, spread blocks in about 150 milliseconds—around 6.5 times faster than Gossipsub, the system Ethereum uses to share data between validators.
Dubbed mumP2P, Optimum’s system builds on mathematical principles Médard first developed over two decades ago in U.S. military–funded research on reliable communication networks.
“What we’re building is the memory layer,” which works “like a computer’s operating system,” she explained. Memory helps systems move data, and that’s where most inefficiencies in blockchains happen, Médard claims.
Optimum says its testnet shows faster data sharing can make blockchains like Ethereum or Solana run more efficiently, with quicker transactions and lower fees that could affect how users trade and interact on-chain.
But while lower latency could help “narrow price gaps,” major gains like those shown by Optimum won’t immediately “erase” existing bottlenecks, said Kanny Lee, founder and CEO of decentralized exchange protocol SecondSwap.
“Even with a six-times improvement over Ethereum, blockchain still operates much slower than traditional finance,” Lee told Decrypt. What this does signal, however, is a “more efficient on-chain environment,” where arbitrage “gets harder and markets react faster to information,” he added.
Given these benefits, blockchain systems could behave “less like a relay and more like an integrated trading network,” Lee said.
He noted that as infrastructure improves, the edge shifts from speed to access as seen in markets for locked tokens, vesting allocations, and structured distributions that “trade on timing and access, not latency.”