Research
Big Tech Has Entered the Blockchain. The Decentralization Question Is Now Different.
July 14, 2026
9 mins read
AWS hosts 37% of Ethereum nodes. JP Morgan's Kinexys has settled $3 trillion on permissioned rails. This article examines what Big Tech's institutional blockchain push actually means for decentralization — and what it still can't solve.
The original argument for blockchain was simple: remove the intermediary. No bank, no cloud provider, no platform extracting rent from every transaction. A decade later, that argument faces a harder test. JP Morgan has settled more than $3 trillion on its own permissioned blockchain. Amazon Web Services runs validator infrastructure for Avalanche with FedRAMP-compliant node deployment. Google Cloud indexes blockchain data for Ethereum, Arbitrum, Polygon, and a dozen others. Visa has expanded stablecoin settlement to nine chains, with an annualized run rate of $7 billion as of Q1 2026.
Big Tech didn't take over the blockchain through hostile acquisition. It built the rails that institutional blockchain runs on.
TL;DR: Big Tech has moved from blockchain curiosity to active infrastructure provider and institutional participant. Enterprise adoption on cloud-hosted nodes creates a new centralization problem: 37% of Ethereum execution-layer nodes now run on AWS alone. The 2026 question is what "decentralized enough" means when the settlement layer is public but the infrastructure running it is private.
How Did Big Tech Get This Deep Into Blockchain?
The entry was gradual and institution-led. It didn't follow crypto hype cycles. It followed enterprise demand.
Amazon Web Services now offers Amazon Managed Blockchain (AMB), a production-grade service providing serverless access to Ethereum, Polygon, Bitcoin, and Hyperledger Fabric. The deeper play is AWS's partnership with Avalanche (Ava Labs): one-click node deployment via AWS Marketplace, support for custom subnets, and FedRAMP-compliant deployment in AWS GovCloud for government and enterprise use cases.
Google Cloud followed a different path. Its Blockchain Node Engine offers fully managed, dedicated node hosting, and its BigQuery datasets now index on-chain data across Ethereum, Avalanche, Arbitrum, Optimism, Polygon, and more. Google Cloud runs validator and seed nodes for select networks including Zilliqa and Somnia, and has formalized partnerships with Coinbase, Fireblocks (Confidential Space integration, 2024), and Deutsche Börse for digital asset infrastructure.
Microsoft Azure's approach is more conservative. Azure Confidential Ledger focuses on tamper-proof data integrity for enterprise records using trusted execution environments. The emphasis is on confidential computing and permissioned rails, not public validator services.
What's notable is the differentiation: AWS is aggressive on public chain infrastructure, Google on data indexing, Microsoft on enterprise confidentiality. Each is building the layer that institutions are most likely to need from them.
What Does $3 Trillion in Settlements Actually Mean?
JP Morgan's Kinexys platform (formerly Onyx) is the clearest signal of where enterprise blockchain stands. Cumulative volume has exceeded $3 trillion since inception, with average daily volume of $5–7 billion as of early 2026.
In November 2025, JP Morgan launched JPMD (its USD deposit token) on Base, Coinbase's Ethereum L2, in a pilot with B2C2, Coinbase, and Mastercard. Kinexys Fund Flow enables live on-chain investor data with JP Morgan Private Bank, Asset Management, and Citco. BMW, FirstRand, Mitsubishi, and Siemens are among the corporate clients running programmable payment flows through the network.
This is permissioned blockchain at scale. It proves the technology works for institutional settlement. But it also demonstrates the central tension of the Big Tech moment: Kinexys is not decentralized. It is a bank-operated blockchain network for bank clients. The counterparty risk is JP Morgan itself.
Visa's trajectory is structurally different. Its stablecoin settlement infrastructure now spans nine chains (Ethereum, Avalanche, Solana, Stellar, Polygon, Base, Canton, and others), with live USDC-denominated transactions reaching $7 billion annualized as of Q1 2026, up 50% quarter-over-quarter. Mastercard has expanded its Circle partnership for USDC and EURC settlement across EEMEA merchant and acquirer networks. These are public chain integrations, not proprietary rails. The centralization profile is fundamentally different from Kinexys.
Is 37% of Ethereum Running on AWS a Problem?
In October 2025, an AWS US-East-1 outage (DNS and DynamoDB failure, approximately 15–16 hours) disrupted execution-layer nodes on Ethereum proportional to their AWS concentration. At that moment, roughly 37% of Ethereum execution nodes ran on AWS.
Broader figures across major cloud providers (AWS, GCP, Alibaba) place 58–70% of all Ethereum nodes on centralized cloud infrastructure, according to node telemetry studies. RPC and indexing layers are even more concentrated at 60–70% estimates. These are the interfaces most users and protocols interact with, making the cloud dependency more operationally significant than raw node counts suggest.
By contrast, Solana's lower AWS stake concentration meant the same outage had minimal effect on validator uptime. This isn't a coincidence. It's the result of different validator incentive structures and geographic distribution.
The FSB and BIS have flagged cloud concentration risk in financial infrastructure broadly, though specific mandates targeting blockchain cloud dependency have not yet materialized. The regulatory attention is there. The enforcement framework isn't yet.
For protocols handling tokenized real-world assets or institutional settlement, cloud concentration is an operational risk, not just an ideological concern. A 15-hour execution-layer disruption on a network settling billions daily is a material event.
What Did Meta, Apple, and Nvidia Actually Do?
Meta's story is instructive precisely because of what it abandoned. After shutting down Diem and Novi in 2022 (assets sold to Silvergate), Meta has no proprietary blockchain infrastructure. Its 2026 move is distribution-oriented: rolling out USDC stablecoin payouts for creators on Facebook, Instagram, and WhatsApp, with pilots in Colombia and the Philippines, using Stripe's infrastructure for custody, settlement, and tax reporting.
The lessons from Libra's regulatory collapse are evident in the design: no self-issued token, no proprietary ledger, third-party custody. Meta's blockchain play is now about distributing compliant stablecoins it doesn't control.
Apple has no blockchain infrastructure. Its relevant moves are App Store policy: updated rules following antitrust and Digital Markets Act rulings now permit crypto wallets, exchanges, NFT minting, and external payment links without the mandatory 30% in-app cut on certain transaction types. Mining remains prohibited on-device. The change enables broader crypto app distribution, but it is regulatory compliance, not an infrastructure play.
Nvidia's involvement is near-zero on the blockchain-specific side. Its GPU architecture has historically been used for crypto mining, but Nvidia's 2024–2026 strategic focus is entirely on AI infrastructure (Blackwell, Rubin GPUs, hyperscaler deals with Meta, Oracle, and the U.S. Department of Energy). No dedicated Web3 partnerships or enterprise blockchain node hardware initiatives were announced in this period.
What Is the Actual Decentralization Status in 2026?
The 2021 narrative — decentralization as an absolute value — has been replaced by something more pragmatic and more useful: "effective decentralization" as a measurable property of stake distribution, geographic spread, and upgrade feasibility. Institutions don't need every node to be on a Raspberry Pi in someone's basement. They need the network to remain operational, verifiable, and resistant to single-point failure.
The tension is structural. Enterprise adoption creates cloud co-location incentives: deploying validators near existing infrastructure reduces latency and simplifies compliance. That's rational behavior for every institutional actor. The aggregate effect is concentration.
The ecosystem responses are multi-chain strategies, DePIN-style distributed compute experiments, and L2/rollup designs that minimize single-cloud RPC reliance. But the honest assessment in 2026 is that these are directionally correct and not yet at the scale needed to offset institutional cloud pull.
BlackRock's BUIDL fund illustrates where institutional tokenization sits: tokenized US Treasuries on Ethereum via Securitize, AUM in the $1–2 billion range, on-chain yield and real-time P2P transfers, integrated with DeFi protocols including Uniswap. Franklin Templeton's OnChain U.S. Government Money Fund (FOBXX/BENJI) has approximately $700M AUM across Stellar and other chains.
These are real-money, regulated products running on public blockchains. They prove institutional blockchain works. They also surface the next-layer problem: when the settlement layer is transparent but the data feeding it (NAV calculations, reserve compositions, collateral positions) comes from off-chain sources with no cryptographic verification, the trust assumption hasn't moved. It has just been relocated.
The verifiable infrastructure gap isn't in the settlement layer. It's in the data layer underneath it.
Public blockchains can prove a transaction happened. They can't prove the underlying data that triggered it was accurate. That's the distinction that will define the next infrastructure cycle, and it's the gap that verifiable data infrastructure is being built to close.
Why Data Integrity Becomes the Next Battlefield?
When JP Morgan settles $5 billion daily on Kinexys, the settlement record is the easy part. The hard part is the data that flows into it: trade confirmations, collateral valuations, fund NAVs, compliance status. Those inputs come from off-chain systems. The blockchain records what it's told.
The same applies to BlackRock's BUIDL: the on-chain yield distribution is provable. The underlying Treasury composition, the repo rates, the intraday collateral positions are reported, not proven. This isn't a critique of these products. It's a description of the current infrastructure boundary.
As tokenized RWA markets scale toward and beyond $36 billion in on-chain value, the gap between verifiable settlement and unverified input data grows into a compliance and operational risk. Regulators are already looking at this. Stablecoin regulatory requirements in 2026 increasingly focus on continuous reserve transparency, not just monthly attestations.
The infrastructure layer needed isn't another chain. It's a proof layer that sits between off-chain data sources and on-chain contracts, generating cryptographic verification at the data level before anything reaches the settlement network. This is what zkDatabase is designed to provide: Zero-knowledge proofs at the source, so the blockchain isn't just recording what it's told. It's recording what's been proved.
Oracle networks confirm data came through authorized sources. zkDatabase confirms the data was real.
Conclusion
Big Tech's blockchain moment is real. AWS, Google Cloud, JP Morgan, Visa, BlackRock — these aren't test-case participants. They are production infrastructure, live settlement volumes, and regulated financial products. The technology has been validated by the most risk-averse institutions in the world.
The decentralization question has shifted. The 2021 argument was ideological. The 2026 argument is operational: 37% of Ethereum running on a single cloud provider isn't a philosophy problem, it's a resilience and regulatory problem. The enterprise blockchain layer is scaling. The data verification layer underneath it is not keeping pace.
That gap is where the next infrastructure cycle will be decided.
Related toResearch
