Charles Hoskinson, founder of Cardano, recently addressed concerns at Consensus in Hong Kong regarding the role of hyperscalers like Google Cloud and Microsoft Azure in decentralized computing. He argued that advanced cryptography, multi-party computation (MPC), and confidential computing mitigate risks, asserting that if the cloud cannot see the data, it cannot control the system.
MPC and Confidential Computing Reduce Exposure
Hoskinson’s argument leaned on technologies like multi-party computation (MPC) and confidential computing to shield data from hardware providers. While these tools are powerful, they do not eliminate the underlying risks associated with centralized infrastructure.
MPC distributes key material across multiple parties, reducing the risk of a single compromised node. However, this expands the security surface to include the coordination layer, communication channels, and governance of participating nodes. The system’s security then depends on a distributed set of actors behaving correctly and the protocol being implemented accurately, shifting the single point of failure to a distributed trust surface.
Confidential computing, utilizing trusted execution environments (TEEs), offers another layer of security by encrypting data during execution. This limits exposure to the hosting provider. However, TEEs rely on hardware assumptions and are vulnerable to side-channel and architectural attacks, as demonstrated in academic literature. The security boundary, while narrower than traditional cloud environments, is not absolute.
Crucially, both MPC and TEEs often operate on top of hyperscaler infrastructure. The physical hardware, virtualization layer, and supply chain remain concentrated. This concentration gives infrastructure providers operational leverage, allowing them to impose throughput restrictions, shutdowns, or policy interventions, even if cryptography prevents direct data inspection.
Advanced cryptographic tools can make specific attacks more difficult, but they do not remove infrastructure-level failure risks. They merely replace a visible concentration of risk with a more complex, distributed one.
The ‘No L1 Can Handle Global Compute’ Argument
Hoskinson also contended that hyperscalers are essential because no single Layer 1 blockchain can manage the computational demands of global systems, citing the trillions invested in data centers. Layer 1 networks are primarily designed for consensus, state verification, and data availability, not for intensive tasks like AI training or high-frequency trading.
While Layer 1 networks serve their intended purpose, modern crypto infrastructure increasingly relies on off-chain computation. The critical factor is the ability to prove and verify these results on-chain, a principle underlying rollups, zero-knowledge systems, and verifiable compute networks. The core issue is not the computational capacity of Layer 1, but rather who controls the execution and storage infrastructure behind the verification process.
If computation occurs off-chain but depends on centralized infrastructure, the system inherits centralized failure modes. While settlement may remain decentralized in theory, the pathway to producing valid state transitions becomes practically concentrated.
Cryptographic Neutrality vs. Participation Neutrality
Hoskinson’s argument for cryptographic neutrality—where rules cannot be arbitrarily changed and backdoors are impossible—is a powerful concept. However, cryptography relies on hardware, and the physical layer dictates participation. Throughput and latency are constrained by real machines and their infrastructure.
If hardware production, distribution, and hosting remain centralized, participation becomes economically gated, even if the protocol itself is mathematically neutral. In high-compute systems, hardware is paramount, influencing cost, scalability, and resilience against censorship. A neutral protocol running on concentrated infrastructure is theoretically neutral but practically constrained.
The priority should shift towards combining cryptography with diversified hardware ownership. Without infrastructure diversity, neutrality becomes fragile under pressure. If a few providers can impose rate limits, restrict regions, or enforce compliance, the system inherits their leverage. Rule fairness alone does not guarantee participation fairness.
Specialization Beats Generalization in Compute Markets
The competition with hyperscalers like AWS is often framed around scale, but this perspective can be misleading. Hyperscalers optimize for flexibility, serving numerous workloads with features like virtualization and elasticity, which come with associated costs.
Technologies like zero-knowledge proving and verifiable compute are compute-dense and reward specialization. A purpose-built proving network competes on efficiency metrics such as proof per dollar, proof per watt, and proof per latency. Vertical integration of hardware, software, and design can lead to significant efficiency gains by removing unnecessary abstraction layers.
Specialized networks, optimized for specific high-volume tasks, can outperform generalized cloud services in terms of sustained throughput and cost-effectiveness. The economic structure also differs, with protocol-aligned networks potentially amortizing hardware costs differently and tuning performance for sustained utilization rather than short-term rental models.
Use Hyperscalers, But Do Not Be Dependent on Them
Hyperscalers are valuable providers of efficient, reliable, and globally distributed infrastructure. The critical issue is dependence, not the providers themselves. A resilient architecture should leverage major vendors for burst capacity, geographic redundancy, and edge distribution, but core functions should not be anchored to a single provider or a small group.
Settlement, final verification, and the availability of critical artifacts must remain intact even if a cloud region fails, a vendor exits a market, or policy constraints tighten. Decentralized storage and compute infrastructure offer a viable alternative, ensuring that proof artifacts and historical records are not subject to a provider’s discretion.
Hyperscalers can serve as optional accelerators rather than foundational elements. While cloud services can offer reach and handle bursts, the system’s core ability to produce proofs and maintain verification data should not be gated by a single vendor. This approach fortifies crypto’s ethos of decentralization, ensuring that the network can continue to function even if a major hyperscaler were to disappear.
Fonte: CoinDesk
