Author: Paul Veradittakit

Translation: Deep Tide TechFlow

Deep Tide Readings: When AI agents begin to possess identity, budget, and settlement capabilities, they are no longer just tools, but new clients. Paul Veradittakit outlines a real map of the intersection of AI and blockchain today from the four perspectives of currency settlement, credit capital, identity control, and computing power sovereignty, specifically naming projects that have already positioned themselves. For investors and practitioners focused on the agent economy and crypto infrastructure, this piece serves as a practical reference for judging the competitive landscape.

Summary:

• Last month, I wrote that when prices no longer rise, founder-market fit continues to compound. This month, Franklin Bi's "The Battle of 8 Billion Customers" outlines the target market we need to build. It encompasses 8 billion people, several agents owned by each individual, and corporate agent teams responsible for programming, finance, procurement, and sales.
• The product is actually quite simple: wallets with limits, stablecoin settlements that machines can complete, proofs that agents can present, computing power that you truly own, and proof without public ledgers.
• Pantera already has a batch of invested companies building at various levels of this intersection, including: currency and machine settlement (Circle, Coinflow, OpenFX); credit, capital, and trading (Morpho, Ondo); identity, credentials, and control (World, TransCrypts, Alchemy); and computing power sovereignty and real-time proof (B3IQ, Orthogonal, Accountable).
• Founders and investors often ask me what is really happening at the intersection of AI and blockchain. I intend to start writing about it here. If you are building in any of the layers mentioned in this issue, please contact me or the Pantera team.

In "Founder-Market Fit," I made a point: the pairing of specific founders with specific markets is the only thing that can continue to compound in a downturn. Simply put, it’s depth, proactivity, networks, and obsession. Examples I cited include Offchain Labs, Ondo, Morpho, Circle, and Alchemy.

Last week, my colleague Franklin Bi wrote an important article titled "The Battle of 8 Billion Customers," outlining the markets that founders are entering. We are now faced with and building for 8 billion people, several agents owned by each, and corporate agent teams responsible for programming, finance, procurement, and sales. This means hundreds of billions of new decision-making and transaction endpoints. When agents possess identity and memory, budget authority, selection and settlement capabilities, and responsibility to their owners, they become clients. Gartner predicts that by 2030, agents' impact on procurement will reach $30 trillion. Visa, Mastercard, and Coinbase's x402 have been issuing credentials, making payments below one cent feasible. Cloudflare also indicated that bots now account for over half of HTTP requests.

8 Billion Customers

8 billion people, each operating a set of personal and corporate AI agents for programming, finance, procurement, sales, and logistics. This represents hundreds of billions of new independent decision-making and transaction endpoints.

AI models become economic agents the moment they possess the following three things:

Identity and memory: cryptographic credentials anchored to the subject.

Budget authority: programmable limits, rate limits, and session keys.

Autonomous settlement: the ability to discover, evaluate, and pay for services on-chain.

Franklin posed a crucial question to builders: who owns these agents? Is it the humans and companies they represent, or the platforms they operate on?

If centralized cloud vendors own the identity, memory, and learning cycles of your agents, switching vendors means firing your digital workforce and letting a group of amnesiacs start over. Blockchain provides property rights at the underlying level: portable identities, bounded delegation, and settlement mechanisms that no model provider can alter.

What AI x Blockchain Looks Like in Practice

1. Currency and Machine Settlement

AI agents do not fill out KYC forms, wait three days for ACH transactions, or manage monthly credit card subscriptions. They need frictionless, below-one-cent, 24/7 payment channels.

Coinflow: Seamlessly settles card and bank payments as stablecoins in over 170 countries, without exposing users to the underlying chain.

OpenFX: Processes stablecoin settlements with an annualized volume of hundreds of billions of dollars, explicitly built for software clients rather than human end users.

Circle (USDC): Launched during the last bear market, it has since become the default accounting unit for machine-to-machine microtransactions and corporate agent settlements.

2. Credit, Capital, and Trading

When agents need to authorize loans or deploy capital based on configurations, they require a programmable liquidity layer.

Morpho: As a credit backend, embedded into Coinbase, Robinhood, Societe Generale, and Apollo. It provides the default lending infrastructure for agents to programmatically query loans.

Ondo Finance: Converts tokenized U.S. treasuries and stocks into productive and yield-bearing collateral. Nathan Allman left Goldman Sachs to solve the problem of institutional asset tokenization, which directly intersects with the budget issues of agents today.

FalconX (via the acquisition of bloXroute): Assembles a high-speed prime brokerage and execution stack that serves continuously operating markets—that’s the only operating timetable recognized by AI agents.

3. Identity, Credentials, and Control

In a digital environment saturated with synthetic content, proving human intent and agent authorization is paramount.

World: Establishes foundational personality proof primitives; uses cryptography to prove the existence of a unique human actor in production environments, combating bot networks and witch attacks.

TransCrypts: Places employment, education, and legal credentials onto a user-controlled track, allowing agents to verify entitlement claims without relinquishing sensitive underlying data.

Alchemy: Provides a core developer platform supporting agent-specific wallets and session key models. Developers no longer hand the master private key over to agents, but issue fine-grained limits, including counterparty limits, expiration timestamps, and instant revocation capabilities.

4. Computing Power Sovereignty and Real-time Proof

B3IQ: Offers sovereignty as a service through a power leasing model. To maintain true agent capability, institutions must ensure that model weights and execution paths remain outside of locked vendor environments.

Orthogonal: Operates as the orchestration and discovery layer for agent services. It is one of the leading platforms offering metered access and native billing in decentralized networks.

Accountable: Enables financial institutions and autonomous funds to cryptographically prove real-time solvency without disclosing private asset balance sheets.

What We Look For in Founders Defining Categories

Having sovereignty alone is not enough to become a value proposition. Winning products leverage decentralized infrastructure to offer lower transaction costs, stricter privacy guarantees, faster customization, or execution reliability that closed platforms cannot match.

When evaluating teams building at this intersection, we look for four qualities:

Deep domain expertise: You have been in the problem rather than just read about it. (For example, Ed Felten left Princeton and the White House to build Offchain Labs / Arbitrum).

Strong proactivity: You have a clear enough view of the market structure that large institutions are building in your lane. (For example, Paul Frambot founded Morpho at twenty to create DeFi's default credit engine).

Unfair networks: Distribution partnerships give you space to maintain sales during market downturns. (For example, Jeremy Allaire partnered Circle with Coinbase to turn USDC into a global settlement standard.)

Obsession: Even when the heat shifts elsewhere, you still have the faith to build across cycles. (For example, Nikil Viswanathan and Joe Lau made Alchemy the default developer platform for Web3.)

If you currently work at Goldman Sachs, Citadel, Stripe, Block, or leading AI laboratories: technical skills that previously seemed adjacent to digital assets are now the job descriptions needed to build the economy of 80 billion customers.

If you are building the core layer for the agent economy, please contact me or the Pantera Capital team. We look forward to building with you.

Business

Bitcoin Records Best August Since 2017

Bitcoin trading prices approached $78,000, increasing about 24% for the month, marking the strongest August in nine years. It briefly broke above $81,000 near the weekend. U.S. spot Bitcoin ETFs absorbed approximately $3.3 billion in August. This is the best month since October 2025 and at one point pushed the asset class back above $100 billion. This surge was led by regulated U.S. demand flowing back through the front door.

Tokenized Spot Stocks Surpass $2.5 Billion

Tokenized spot stocks surpassed $2.5 billion in August, with a month-over-month increase of about 8%, and an increase of over 260% year-to-date. BNB Chain, Ethereum, and Solana split the market. DTCC is also conducting real-time simulations with about 40 firms. Participants include JPMorgan, Goldman Sachs, Invesco, and Citadel, using tokenized stocks and treasuries as collateral. On-chain stocks are beginning to be settled like collateral rather than packaging.

USDC Supply Jumps, Circle Reclaims Trading Share

USDC supply increased by approximately $2 billion within a week after stagnating for six months. On-chain trackers also indicated several billion dollars in issuance that week, with circulation nearing $74 billion. Bernstein maintains an "outperform" rating for Circle with a target price of $140. The agency noted that by 2026, USDC's share of adjusted trading volume has risen to over 60%. Only dollar tokens that can stay within the boundaries of U.S. banks are the ones institutions will expand.

Regulation

Senate Set to Vote on CLARITY Bill on September 15

Majority Leader Thune submitted a motion to advance the initiative to end the debate. Following the August recess, procedural votes have been scheduled for the calendar on September 15. The House has passed the bill; the Senate has a short window before its next recess, with ethics, DeFi, and Democratic support remaining key variables. The market structure will either obtain codified law this fall or have rules determined by institutions for the remainder of this cycle by default.

After Bill Stagnation, SEC Proposes Rules for Crypto Asset Regulation

On August 18, the SEC proposed a tailored issuance system: $5 million startup exemptions, a financing tier up to $75 million, conditional Howey safe harbor, and state registration preferential applicability. Chairman Atkins emphasized that this proposal cannot replace legislation. While Congress is paused, agencies continue to write market rules.

Treasury Proposes Rules Defining Stablecoin Issuance and Sale

The U.S. Treasury released the third proposal regarding the issuance and offer of payment stablecoins, setting a 60-day comment period. The previous final rule's statutory deadline of July 18 was missed. The OCC is now aiming to complete this before November; the backup effective date for the law is January 18, 2027. This framework is no longer a debate. It is a compressed implementation clock.

New Products and Transactions

Crypto M&A Hits Record High in First Half of 2026

Mastercard completed its acquisition of BVNK on August 3, integrating stablecoin payment stacks into a global card network. Total crypto M&A in the first half of the year has reached a record $9.7 billion, despite a decrease in the number of transactions. Huge capital flow is directed toward licensed payments rather than permissionless experiments. Buyers in the digital asset space are now legacy payment companies.

Coinbase Establishes Tokenization Center in Abu Dhabi

ADGM granted Coinbase a license to facilitate investments and custody of tokenized securities backed by underlying stocks. This authorization expands Project Diamond into a regional issuance and custody center. Transfers remain within a regime of sanction screening and wallet-level controls. Tokenized stocks are departing the demonstration phase and gaining regulated footholds outside the U.S. listing queue.

BlackRock Introduces Tokenized Money Market Product Built for GENIUS Reserves

BlackRock launched an on-chain share class of a treasury liquidity fund, alongside a multi-chain daily reinvestment reserve tool. Both aim to meet the qualified reserve requirements of permitted U.S. payment stablecoin issuers. The company manages approximately $60 billion in Circle reserves and aims to become the default reserve manager for this category. The growth of stablecoins is turning into an asset management task.

Pantera News

Bitcoin Shifts Position from Short to Long After Recovering 200-Day Moving Average

Pantera General Partner Cosmo Jiang pointed out that since nearing $126,000 in October 2025, there has been about a ten-month, approximately 50% retracement. With Bitcoin holding around the $69,000 200-day moving average, traders are shifting from net short or cash to long exposure. He noted that friendlier U.S. policies and larger government bond buybacks are direct catalysts, with $80,000 being the next resistance level.

Fortune Reports on B3IQ's Power Rental Buy Model

Pantera portfolio company B3 Labs launched B3IQ. Researchers and businesses can put down 30% to own U.S.-hosted NVIDIA machines instead of renting ultra-large computing power locked until 2030. The first wave has ramped up quickly, with the team reporting eight-figure GPU sales within six days, and capacity expanding for startups, labs, and data-sensitive industries. Crypto-native operators are becoming the ones who can truly deliver scarce computing power.

On Stateful: Why AI Builders Need to Own Their Computing Power

Watch the show · Listen on Stateful

Franklin Bi sat down with B3 co-founders Daryl Xu and Viktoriya Hying along with Yorke Rhodes from New York University to discuss this launch. The cost of renting H200 nodes for two years is roughly the same as purchasing directly. Idle cycles will be matched to acquirers, allowing machines to recoup costs. Closed models have prevented research teams from working on human trafficking and war-zone evacuation-related projects. Open-source models are catching up to the forefront. Owning hardware is the key to making them usable.