The AI-Crypto Convergence: When BlackRock Meets 40M TPS

Crypto investors often measure adoption through human activity: active wallets, exchange volumes, ETF flows, stablecoin balances, and retail transaction counts. Yet the industry’s next major demand engine may operate largely outside those familiar metrics.

On October 5, 2026, BlackRock published research arguing that autonomous AI agents could become an underappreciated source of crypto demand. Bitcoin Magazine covered the thesis the following day. The argument is straightforward: software that independently purchases data, computing resources, services, or digital assets needs payment and settlement infrastructure built for continuous, low-value, machine-speed transactions.

On October 7, Sui reported that its Tunnels system processed more than 40 million transactions per second in a live, high-stress demonstration. The result involved programmable offchain channels rather than 40 million conventional Layer 1 transactions individually reaching global consensus each second. Even so, it illustrates the type of architecture that may be required if machine-driven commerce reaches scale.

Together, these developments suggest that the convergence of AI and crypto is more than a speculative narrative. BlackRock is describing a potential source of demand, while emerging high-throughput architectures are demonstrating one possible way to support it.

BlackRock’s Machine-Native Crypto Thesis

BlackRock’s crypto thesis begins with a fundamental mismatch. Traditional payment systems were designed around people and institutions. They generally assume account-opening processes, identity checks, checkout interfaces, payment minimums, operating schedules, intermediaries, and dispute processes involving humans.

Autonomous software has different requirements. An agent might pay a fraction of a cent for an API query, compensate another agent for specialized analysis, purchase seconds of computing capacity, or execute thousands of conditional payments during a single workflow. These actions require programmable authorization, rapid and reliable settlement, global availability, and economics that remain viable at very small transaction sizes.

BlackRock’s October 5 research describes AI as machine-native intelligence and digital assets as potential machine-native money. Its report argues that agentic commerce could increase demand for programmable settlement infrastructure, particularly stablecoins and blockchain networks. BlackRock also noted that stablecoin market capitalization had exceeded $300 billion by September 2026, while adjusted stablecoin transaction volume surpassed $11 trillion in 2025—evidence that digital settlement rails have already achieved meaningful scale (BlackRock, October 5, 2026).

The key investment insight is that crypto adoption may no longer depend exclusively on persuading more people to use wallets. Instead, one person or business could authorize hundreds of agents, each generating transactions continuously. That dynamic changes the potential relationship between users and network activity.

Why Agents May Prefer Blockchain Rails

Blockchains give software direct access to programmable assets without requiring every transaction to pass through a human-facing checkout process. Smart contracts can define spending limits, approved counterparties, permitted assets, expiration times, and conditions for execution.

Stablecoins are especially relevant because agents need predictable units of account. Volatile tokens may still provide network security, collateral, or fee utility, but an agent purchasing $0.002 of data generally benefits from knowing that its payment asset will remain close to its intended value.

This does not mean blockchains will replace cards or bank transfers. Conventional APIs, bank rails, and centralized ledgers may remain efficient for many machine payments. Blockchain rails are most likely to have an advantage when parties need interoperable, programmable, always-on settlement without relying on a shared intermediary.

The likely outcome is a hybrid payment environment, with blockchain rails gaining relevance in digitally native, machine-to-machine transactions.

Why Legacy Blockchain Capacity Was a Bottleneck

Earlier blockchain designs typically required validators to agree on a largely ordered sequence of transactions. That model offers security and consistency, but it can create congestion when every interaction competes for the same blockspace.

Human users can tolerate occasional delays or bundled actions. A machine-to-machine economy cannot make the same assumption for time-sensitive payments, permissions, or service delivery. If millions of autonomous AI agents negotiate services, stream payments, update permissions, and settle obligations simultaneously, a serial execution model can become a practical and economic bottleneck.

Not every agent interaction needs immediate global settlement. Many routine exchanges can be handled locally or offchain. But when value must be transferred, permissions enforced, or disputes resolved, the system still needs a low-cost and verifiable path to settlement.

Fees are equally important. A payment rail may be technically capable of processing a microtransaction but still be commercially unusable if the fee exceeds the value of the payment itself. Infrastructure for agentic commerce therefore needs high capacity, predictable costs, rapid confirmation, programmable permissions, and verifiable records.

Parallel processing helps by identifying transactions that do not depend on one another and executing them concurrently. Two agents exchanging data should not necessarily wait for an unrelated DeFi trade or gaming transaction. Architectures that separate independent state changes can use computing resources more efficiently than systems that force every operation through a single global queue.

One approach is to move frequent, low-value interactions into offchain or local channels while retaining a blockchain as the settlement and dispute-resolution layer. That is the model illustrated by Sui’s recent demonstration.

What Sui’s 40 Million TPS Result Actually Means

On October 7, Sui announced that its Tunnels system processed 40,614,180 transactions per second during a live, high-stress demonstration, exceeding a 20 million TPS target and its previous result of roughly 6 million TPS (Investing.com, October 7, 2026).

The technical distinction matters. This was not 40 million conventional Layer 1 transactions individually reaching global consensus every second. The activity occurred through Sui Tunnels, programmable offchain state and payment channels. Tunnels are opened and closed through mainnet transactions, while intermediate interactions occur offchain and can be cryptographically verified when settlement or enforcement is required.

That design illustrates an important scaling principle: a blockchain does not need to place every machine interaction directly into shared global state. It can instead serve as the trust, settlement, and dispute-resolution layer beneath much larger volumes of local or offchain activity.

More than 10,000 Tunnels were reportedly opened during the demonstration. CertiK analyzed the test in real time, although a detailed independent report, including complete proofs, logs, and methodology, was still pending as of October 7. Investors should therefore distinguish between a reported live demonstration and a fully documented, independently reproducible result.

Parallelism and Selective Settlement

The broader significance of Sui’s TPS result is not simply the headline number. It is the combination of parallelizable execution, programmable channels, and selective onchain settlement.

Agents can establish rules when opening a channel, conduct large numbers of interactions without paying mainnet fees for each step, and settle or enforce the resulting state onchain when needed, including at closure or in a dispute. This approach could support chat services, games, compute markets, data exchanges, and streaming payments without forcing every intermediate event through expensive global consensus.

For AI workloads, that may be more valuable than maximizing raw base-layer TPS. The objective is not to put every software action onchain. It is to make economically important actions enforceable and auditable.

The Next Billion Users May Be Software

Retail adoption traditionally follows a one-user, one-wallet mental model. Autonomous systems break that assumption. A single company might deploy separate agents for procurement, cybersecurity, treasury management, cloud optimization, advertising, and data acquisition. Those agents could also hire specialized agents operated by other organizations.

The resulting network would resemble an economy rather than a single application. Agents would discover counterparties, negotiate terms, pay for resources, verify delivery, and maintain records. Crypto infrastructure could provide a shared economic layer between systems that do not fully trust one another.

This is why BlackRock’s thesis may be more structurally significant than another forecast about human adoption. According to the firm’s analysis, blockchains are well suited to high-frequency, sub-cent payments for API calls, on-demand information, and consumption-based computing. If that thesis proves correct, transaction demand could grow with software activity rather than population size.

The addressable market is therefore not limited to the number of people willing to manually sign transactions. It includes the number of economically active software processes that those people and their organizations authorize.

How Investors Should Evaluate the Opportunity

The AI narrative alone is not an investment thesis. A disciplined analysis should separate technological capacity, real economic demand, and sustainable token value capture.

First, examine production usage rather than benchmark peaks. Investors should track recurring transaction volume, active applications, stablecoin liquidity, fee generation, retained developers, and the share of activity produced by genuine economic demand.

Second, assess whether the token captures value. A network can support enormous offchain volume without generating proportionate fees or token demand. Investors should understand which actions require the native asset, how validators and other infrastructure providers are compensated, and whether greater usage increases demand for staking, collateral, transaction fees, or governance participation.

Third, evaluate security boundaries. Agent-controlled wallets create risks involving compromised models, malicious prompts, faulty permissions, oracle manipulation, and automated losses at machine speed. Strong infrastructure needs spending caps, revocable authorizations, simulation tools, monitoring, clear recovery mechanisms, and practical frameworks for identity and accountability.

Finally, avoid treating TPS as a universal comparison. Investors must distinguish among base-layer transactions, batched operations, state-channel interactions, theoretical capacity, test results, and sustained mainnet performance. For high-throughput blockchains, methodology and economic relevance matter as much as the largest published number.

The strongest projects at the intersection of AI and crypto will likely combine scalable execution with stablecoin depth, dependable developer tooling, robust security, predictable fees, and measurable value capture. Capacity is necessary, but capacity without demand is merely unused infrastructure.

Conclusion

BlackRock’s machine-native economy thesis and Sui’s 40 million TPS demonstration represent two sides of the same potential structural shift. Autonomous AI agents could create demand for constant, programmable microtransactions, while parallel execution and offchain channels offer a path beyond the cost and capacity constraints of earlier blockchain systems.

The opportunity is substantial, but investors should remain disciplined. Sui’s result illustrates what emerging architecture may support; it does not, by itself, prove production adoption, fully documented independent verification, or token value accrual. The critical metrics will be verified performance, real agent-driven volume, security, settlement demand, and sustainable economics.

If those elements converge, crypto’s next adoption cycle may be defined less by another billion people opening wallets and more by billions of autonomous software agents using blockchain infrastructure as part of their native financial system. The investable beneficiaries, however, will be the networks and assets that convert that activity into durable economic value.