Gate Institutional Monthly: Macro Liquidity Pivots, BTC And ETH Rebound

Summary

• In August 2026, global markets continued to recover amid weakening growth and employment momentum, although inflationary pressures remained unresolved. Policy expectations shifted from further tightening toward a pause in rate hikes and eventual easing, but elevated US Treasury yields, energy prices, and geopolitical risks continued to constrain the scope for liquidity improvement.

• Inflows into spot BTC and ETH ETFs strengthened significantly from July. BTC re-emerged as the core institutional allocation, while ETH continued to attract steady incremental buying. Capital shifted from selective positioning toward simultaneous allocations to both major assets, although it did not yet broaden meaningfully into small- and mid-cap crypto assets.

• Global equity markets returned to the AI and mega-cap technology theme. Nvidia’s earnings provided further evidence of sustained demand for computing power, with semiconductors emerging as a major catalyst toward month-end. Gold rose sharply on safe-haven demand and shifting policy expectations, oil retained a geopolitical risk premium, and copper remained relatively resilient.

• Activity across Meme trading front ends returned to a cyclical high, with liquidity rotating rapidly among ecosystems including Robinhood Chain, and Solana. GMGN regained market leadership through its multi-chain coverage, while trading front ends gradually evolved from single-chain tools into gateways for cross-chain speculative flows.

• Real transaction activity on x402 increased significantly. High-frequency, low-value on-chain micropayments began to validate the feasibility of AI agents autonomously discovering, purchasing, and settling external services. Transaction counts, the number of paid service providers, and call frequency are becoming key indicators of Agentic Payments adoption.

• The US Treasury’s expansion of long-dated Treasury buybacks drove Treasury yields and the US dollar lower. Combined with improving regulatory expectations, ETF inflows, and crowded short positioning, this helped propel BTC and ETH sharply higher. Whether the rally can continue will depend on whether spot demand can replace short covering as the primary driver.

1. Macroeconomic and Market Trends

1.1 Slowing Growth Shifts Policy Expectations Toward Easing, but Inflation and Energy Risks Limit the Scope

In August 2026, the central macroeconomic tension shifted further toward weakening employment and growth momentum while inflationary constraints remained unresolved. Data released during the month showed that the US Consumer Price Index rose 0.1% month over month and 3.4% year over year in July, while core CPI increased 0.2% month over month and 2.5% year over year. Headline inflation eased slightly from June, but housing costs continued to rise, while energy prices fell 1.5% month over month. This suggests that the improvement in inflation remained heavily dependent on lower energy prices and that underlying price pressures had not fully dissipated.

On growth, the second estimate of US real GDP for the second quarter remained at an annualized 1.5%, down from 2.1% in the first quarter. Notably, growth in real final sales to private domestic purchasers was revised up to 4.2%, indicating that consumption and private investment retained a degree of resilience despite the broader slowdown. At the same time, the core PCE price index rose at an annualized rate of 3.6% in the second quarter, reinforcing the combination of slower growth and persistently elevated inflation.

Employment data became an important catalyst for the shift in policy expectations during the month. US nonfarm payrolls fell by 23,000 in July, significantly underperforming market expectations, while June’s employment gain was revised down to 20,000. The unemployment rate declined from 4.2% to 4.1%, although this was partly attributable to a lower labor-force participation rate. Weak employment data prompted markets to reduce expectations of a September rate hike and to begin pricing in a policy pause and the possibility of subsequent easing.

The Federal Reserve, however, did not provide a clear signal that rate cuts were imminent. No FOMC rate meeting was held in August, and the federal funds target range remained at 3.50%–3.75%. At the Jackson Hole symposium, Federal Reserve Chair Kevin Warsh noted that year-over-year PCE inflation remained at 3.7% and that the recent run of softer data was insufficient to demonstrate a clear improvement in the underlying inflation trend. He maintained that price stability should remain the Federal Reserve’s primary policy priority.

Overall, the macro trade in August shifted from a singular focus on the risk of further rate hikes toward a more nuanced view: weak employment reduced the probability of additional tightening, but high inflation delayed the prospect of easing. For crypto markets, the tail risk associated with tighter policy diminished, but long-dated Treasury yields and oil prices remained elevated, meaning liquidity conditions did not improve across the board. As a result, although risk assets rallied significantly during the month, capital remained concentrated in highly liquid assets such as BTC, ETH, and mega-cap technology stocks.

1.2 ETF Flows: BTC Regains the Lead as ETH Continues to Attract Steady Inflows

From an ETF-flow perspective, institutional allocation activity was considerably stronger in August 2026 than in July. Based on the 21 trading days covered by Farside Investors, US spot BTC ETFs recorded approximately $3.539 billion in cumulative net inflows during the month, while spot ETH ETFs attracted approximately $1.837 billion. Net inflows into BTC were roughly 1.9 times those into ETH, marking a clear reversal from July, when ETH inflows exceeded those of BTC.

In terms of timing, BTC ETFs recorded consecutive inflows at the beginning of the month before briefly moving into net outflows from August 10 to 14. Capital returned rapidly after August 17, including a single-day net inflow of approximately $606 million on August 20. The products then recorded positive inflows for seven consecutive trading days from August 19 to 27. Although approximately $202 million flowed out on August 28, a net inflow of around $217 million was recorded again on August 31, indicating that the month-end outflow did not develop into a sustained withdrawal of capital.

ETH ETF inflows were smaller in absolute terms but more consistent. All 11 trading days from August 17 to 31 recorded non-negative flows, with daily net inflows approaching or exceeding $200 million on August 20, 26, and 27. BlackRock’s product remained the primary source of incremental demand, indicating that institutional appetite for ETH did not disappear even as BTC regained the lead.

Overall, the ETF market moved from selective replenishment in July to simultaneous increases in BTC and ETH allocations in August. The hierarchy of capital, however, remained clear: BTC returned to its position as the core holding, while ETH functioned more as a higher-beta supplementary allocation. This structure supported a recovery in liquidity across major crypto assets, but it was not sufficient to demonstrate a broad rotation into small- and mid-cap tokens.

1.3 Global Capital-Market Trends

1.3.1 Global Risk Assets: US Equities Rebound Broadly, but Macro Disruptions Drive Intra-Month Volatility

Global risk assets generally recovered in August 2026, with all major US equity indices posting positive returns. Based on closing levels from July 31 to August 31, the S&P 500 rose from 7,440.75 to 7,686.14, gaining approximately 3.3% during the month. The Nasdaq Composite advanced from 25,118.17 to 26,370.89, an increase of approximately 4.99%, while the Dow Jones Industrial Average rose from 52,201.42 to 53,185.90, gaining around 1.89%. Technology stocks regained the lead, although value stocks and small caps also attracted some inflows.

Risk appetite during the month was driven primarily by three factors. Employment data came in below expectations, reducing the urgency of further rate hikes; earnings at major technology companies continued to grow, easing some concerns about returns on AI investment; and ETFs and other institutional channels resumed allocations to highly liquid risk assets. At the same time, rising Treasury yields, persistently high oil prices, and renewed tensions between the United States and Iran prevented markets from establishing a sustained one-way advance.

Notably, after the United States struck Iranian military facilities near the Strait of Hormuz on August 31, Brent crude climbed back above $90 per barrel, the S&P 500 fell 0.3% on the day, and the 10-year US Treasury yield rose to approximately 4.75%. This demonstrated that geopolitical risks could still affect asset valuations rapidly through oil prices and inflation expectations.

Overall, risk appetite improved further in August compared with July, but this did not represent the complete removal of high-rate risks. Institutional investors continued to favor large companies with greater earnings visibility while using gold, energy, and highly liquid crypto assets to manage macroeconomic uncertainty.

1.3.2 Equities: AI Earnings Delivery Regains Recognition, with Semiconductors Driving the Late-Month Rally

At the individual-stock level, markets returned to the AI and mega-cap technology theme in August, but the investment case became increasingly dependent on demonstrated earnings delivery. Nvidia rose 8.7% in a single session following its earnings release, its largest one-day gain since April 2025, helping lift the Nasdaq by approximately 1.6% that day. This indicated that investors remained willing to pay a valuation premium for rapid earnings growth backed by clear AI demand.

Large cloud-computing and platform companies continued to benefit from investment in AI infrastructure. Capital expenditure by Microsoft, Amazon, Alphabet, and Meta remained elevated, but investor assessments of that spending became increasingly differentiated. Companies capable of rapidly converting investment into revenue through cloud services, advertising, enterprise software, and demand for computing power received greater recognition. Those with longer payback periods and more pronounced pressure on free cash flow continued to face valuation constraints.

The performance of Apple and Tesla was driven more by product cycles, supply-chain developments, and capital-expenditure expectations, and did not move fully in line with AI compute stocks. Semiconductor companies such as Micron and Broadcom continued to offer higher upside sensitivity, although their performance was more vulnerable to inventory cycles, order guidance, and valuation levels. The healthcare sector provided only limited defensive benefits, as the market favored technology and communication-services companies with clearer earnings growth.

The central change in the equity market during August was therefore that earnings delivery once again became a prerequisite for valuation expansion. Technology remained the dominant theme, but index gains continued to depend heavily on earnings reports and a small group of large companies. Institutional portfolios were therefore better served by maintaining a balance among platform companies, semiconductor stocks, and other sectors.

1.3.3 Gold: Strong Monthly Advance Followed by Rapid Profit-Taking at Month-End

Gold strengthened significantly in August, gaining approximately 10% during the month. It broke above its previous trading range in early August and subsequently rallied sharply amid a weaker US dollar, softer employment data, and rising geopolitical risks, at one point approaching $4,700 per ounce.

The rally reflected three primary factors. First, weak employment data reduced the probability of further Federal Reserve rate hikes. Second, developments in the Middle East and risks surrounding the Strait of Hormuz increased demand for safe-haven assets and inflation hedges. Third, stronger buying through gold ETFs provided incremental capital that supported the breakout. The World Gold Council also noted during the month that broad-based gold ETF purchases helped prices break above the preceding technical consolidation range.

Gold nevertheless pulled back sharply from its highs around August 28, indicating that short-term positioning had become crowded. Rising long-dated Treasury yields also increased the opportunity cost of holding gold. Gold therefore moved from a period of high-level consolidation in July into a trend-driven recovery in August, although volatility rose markedly toward month-end.

For institutional portfolios, gold’s value as a hedge strengthened again. Following such a rapid one-month advance, however, its near-term risk-reward profile was less attractive than it had been at the beginning of the month. Its subsequent direction will continue to depend on inflation data, real interest rates, and risks to Middle Eastern energy supplies.

1.3.4 Commodities: Energy Trades in a High Range, Precious Metals Lead, and Copper Remains Relatively Resilient

Commodity performance remained highly differentiated in August. Precious metals were the strongest segment, with gold gaining approximately 10% and silver and platinum posting double-digit or near-double-digit returns. In addition to its safe-haven and monetary characteristics, silver benefited from expectations of industrial demand. Platinum was supported by supply constraints and rotation within the precious-metals complex.

Oil experienced substantial volatility during the month. In early August, markets priced in easing tensions in the Middle East and recovering supply, briefly pushing oil prices lower. Renewed conflict around the Strait of Hormuz, however, drove Brent crude back above $90 per barrel by month-end. Oil broadly remained range-bound at elevated levels. High oil prices continued to represent a major external constraint on inflation expectations and global risk-asset valuations.

Copper remained relatively resilient, posting a modest monthly gain or trading sideways at elevated levels. The energy transition and capital expenditure on power grids and data centers continued to support the medium- and long-term demand outlook, but slowing global growth, Chinese domestic demand, and high inventories limited the upside. Copper therefore outperformed some traditional cyclical commodities but lagged precious metals by a wide margin.

Overall, the main commodity themes in August were the outperformance of precious metals, the persistence of a risk premium in energy, and a measured advance in copper. Institutional allocation decisions still needed to distinguish among the drivers of individual commodities: gold primarily reflected safe-haven demand and rate expectations, oil reflected geopolitical supply risk, and copper was more closely tied to global capital expenditure and real-economy demand.

2. Analysis of Key Market Segments

2.1 Tokenized-Equity Trading Volume Reaches a Record High as Alpha Incentives Amplify BNB Chain Liquidity

Tokenized equities continued to expand rapidly in August, while the market became more concentrated than it had been in July. According to RWA.xyz, the value of tokenized equities reached $2.53 billion, representing growth of 15.66% over the preceding 30 days. Monthly on-chain transfer volume exceeded $34 billion in August, while the number of holders reached 2.39 million, an increase of 149.21% month over month. Compared with monthly volume of more than $9 billion in July, August added nearly $20 billion, with trading activity expanding far more rapidly than net asset issuance. A small number of products accounted for most of this increase. Four bStocks—Invesco QQQ, SpaceX, SPY, and Nvidia—together with the Nvidia token on Robinhood Chain, represented the principal sources of incremental volume during the month.

August extended the redistribution of platform market share that began in July. Trading volume in June was concentrated primarily in xStocks and Ondo. In July, bStocks volume rose to $3.28 billion, tokenized-equity trading on Robinhood Chain exceeded $1 billion, and xStocks fell back toward the $1 billion range. In August, bStocks became an even more important contributor to tokenized-equity trading volume, with platform assets reaching $594 million, close to the $607 million held through xStocks. bStocks offered 68 products, while xStocks covered 715. Average assets per product were therefore approximately $8.73 million for bStocks and $850,000 for xStocks—a difference of roughly tenfold. This reflected bStocks’ more concentrated approach to asset organization, with market-making capital, user activity, and campaign resources focused on a small number of index ETFs, technology stocks, and closely followed private-company exposures.

The growth of bStocks was built on a coordinated distribution system spanning the exchange, Wallet, Alpha, and BNB Chain. Issued as BEP-20 tokens on BNB Chain, bStocks could be integrated with applications including PancakeSwap, Venus, Lista DAO, and Aster, while also entering the exchange’s spot, margin, and collateral systems. Users could move the same category of equity assets among exchange accounts, Wallet, and on-chain protocols. The platform, in turn, could draw on existing stablecoin liquidity, market-making networks, and account traffic to distribute the products. Compared with RWA issuers that must acquire users independently, bStocks directly leveraged the exchange ecosystem’s existing traders and capital base, shortening the time required for new products to establish liquidity.

Trading volume expanded significantly when this distribution advantage was combined with Alpha incentives in August. Beginning August 8, the exchange’s Wallet revised its bStocks reward structure: designated products generated four times the Alpha trading-volume credit from Monday through Friday, while other products received a standard one-times multiplier. From August 17 to September 1, Wallet, bStocks, BNB Chain Agent Studio, and CoinMarketCap also launched a trading competition with a prize pool of up to 100,000 USDC. By trading bStocks, users could simultaneously accumulate Alpha weightings, compete in campaign rankings, and pursue potential rewards. Returns therefore combined changes in equity prices, Alpha eligibility, and campaign incentives. The four-times multiplier reduced the opportunity cost of repeatedly turning over the same assets, causing designated products such as QQQ, SPY, Nvidia, and SpaceX to absorb substantial task-driven trading activity.

Invesco QQQ became the single largest asset by transfer volume in August, reflecting both its product characteristics and its compatibility with the incentive structure. QQQ provides exposure to large technology companies and offers better price continuity, market recognition, and market-making conditions than most individual stocks, making it suitable for frequent trading by users completing Alpha tasks. SPY provides broad-market index exposure, Nvidia enjoys considerable retail-trading interest, and SpaceX offers exposure that is difficult to obtain directly in traditional public markets. Together, the four products cover a technology index, the broad US market, a popular individual stock, and a private company. This mix aligned with user preferences and allowed the platform to concentrate market-making resources. High bStocks volume was therefore jointly shaped by platform distribution, asset selection, and incentive rules rather than by uniform growth across dozens of products.

Assets on bStocks totaled approximately $594 million, while cumulative trading volume approached $19 billion, creating a substantial gap between turnover and net asset value. The platform’s holder count and asset base did increase in August, but not enough to explain tens of billions of dollars in on-chain turnover on their own. Some users may have repeatedly bought and sold assets carrying the four-times Alpha multiplier, creating conditions for incentive-driven wash-like volume using limited capital. Market makers also needed to rebalance inventory among the exchange, Wallet, and liquidity pools.

The Nvidia token on Robinhood Chain followed a different growth path. After Robinhood Chain launched its mainnet in July, it connected equity tokens, wallets, an AMM, and brokerage accounts within a single infrastructure stack. By late August, only 49 of the chain’s 196 equity tokens generated more than $1,000 in daily volume. Nvidia, SpaceX, and SPY together accounted for $57.7 million of the chain’s $81 million in daily turnover, or 71%, with Nvidia alone contributing $29.2 million. This high concentration in well-known stocks indicated that Robinhood users continued to apply preferences developed in conventional brokerage accounts after entering on-chain markets. Capital flowed first toward companies with strong name recognition, sufficient price volatility, and frequent news catalysts.

By month-end, assets on Ondo, xStocks, and bStocks had reached $840 million, $607 million, and $594 million, respectively. Their combined assets of approximately $2.041 billion represented 80.7% of the tokenized-equity market. Asset share was becoming concentrated among three platforms, while trading volume was even more heavily focused on a small group of incentivized bStocks and popular equities on Robinhood Chain. Market growth in August shifted competition from issuance platforms toward distribution channels and incentives. A platform’s ability to control user access, stablecoin liquidity, market-making networks, and campaign rules began to influence on-chain volume rankings directly.

The nearly $30 billion in transfer volume recorded during August reflected the rapid expansion of tokenized-equity distribution, but also included a substantial increase in transaction frequency driven by Alpha campaigns. bStocks helped BNB Chain establish tokenized-equity liquidity rapidly, while Robinhood Chain created an independent trading center for brokerage users. The first-mover shares held by Ondo and xStocks were diluted further. The next stage of competition is likely to center on retaining traffic after incentives expire. Only platforms capable of converting campaign participants into long-term holders, translating high-frequency turnover into net asset inflows, and integrating equity tokens into collateral, margin, and asset-management use cases will be positioned to turn short-term activity into sustainable market share.

2.2 Meme Trading Front-End Activity Returns to a Cyclical High as GMGN Navigates the Cycle and Regains Market Leadership

Meme-market activity recovered significantly in August, making trading front ends some of the most direct beneficiaries of the renewed growth in on-chain activity. Weekly trading volume across Meme trading front ends briefly returned to nearly $4 billion, its highest level since the second half of February 2025. Daily volume climbed above $600 million, indicating that demand for Meme trading had rebounded rapidly after several months of cooling.

Unlike the previous market cycle, which was driven primarily by the Solana Meme ecosystem, incremental trading activity in this cycle began rotating among multiple chains, including Robinhood Chain, BNB Chain, and Solana. The wealth effects generated by new chains, new launchpads, and new assets became important drivers of the recovery in trading activity. At the same time, the life cycle of Meme assets shortened further and capital migrated more rapidly between ecosystems, increasing demand for token discovery, Smart Money tracking, fast execution, and cross-chain coverage. Meme trading front ends became important gateways for high-frequency speculative capital.

Viewed over a longer time horizon, the recovery in trading volume was accompanied by a significant reshuffling of the competitive landscape. During the 2024–2025 Meme cycle, BullX, Photon, and GMGN were among the market’s most prominent leading platforms. Their performance began to diverge materially after the start of 2026. BullX and Photon gradually moved away from the center of the volume competition, while a new generation of products, including Axiom and Fomo, quickly captured users and trading demand. Axiom expanded rapidly through its strong Solana trading experience, points program, and cashback incentives, at one stage becoming a primary gateway for Meme trading. Fomo grew through its mobile experience, social-trading features, and low-friction execution. By contrast, GMGN, which built its initial scale in the Chinese-speaking Meme market, did not retreat when the previous Meme boom faded. It maintained a relatively high level of trading activity despite repeated challenges from new platforms such as Axiom and Fomo, and returned to the center of the market when Meme activity recovered in August. Platforms that had risen rapidly through dependence on a single ecosystem and short-term traffic began to fall behind, while those able to continue acquiring users and follow shifts in trading activity secured more stable market share. Meme front ends consequently entered a new phase of consolidation and competitive elimination.

GMGN offers a representative example of this shift. As one of the earlier trading front ends to achieve scale in the Chinese-speaking Meme market, GMGN did not bind its product exclusively to a single ecosystem after Solana Meme activity cooled. Instead, it continued to expand its multi-chain coverage and rapidly absorbed trading flows as new Meme themes emerged. When activity on Robinhood Chain accelerated, GMGN was among the first platforms to capture demand from that ecosystem. As BNB Chain Meme activity subsequently increased, incremental trading on the platform expanded further into BNB Chain. Driven by these developments, GMGN’s weekly trading volume exceeded $1 billion in August, and the platform briefly became the largest Meme trading front end by volume. Its growth path differed materially from the previous cycle, when it depended primarily on Solana, with an increasing share of volume now generated by liquidity migrating among different ecosystems.

This shift also reflects a change in the competitive logic of Meme trading front ends. Meme traders are highly opportunity-driven. When a chain produces a new launchpad, low-market-cap assets, and a visible wealth effect, trading capital often migrates rapidly within a short period. A platform’s trading volume therefore increasingly depends on whether it can continue to follow Meme liquidity across chains. For trading front ends, the speed of new-chain integration, token-discovery efficiency, Smart Money data, wallet profiling, execution capabilities, and risk detection are collectively becoming barriers to user churn. A platform that depends heavily on a single ecosystem may acquire users rapidly during one market phase, but it can lose volume just as quickly when Meme activity in that ecosystem subsides. By contrast, a platform that allows users to track capital and asset opportunities across chains through a single interface has a better chance of converting cyclical Meme participants into frequent, persistent users. GMGN’s ability to capture incremental volume from Robinhood Chain and subsequently BNB Chain effectively validated its multi-chain strategy in an environment characterized by rapid cross-chain migration of Meme liquidity.

The recovery in trading activity also highlighted the strong monetization potential of Meme front ends. Conventional market-data tools and wallets typically rely on subscriptions, advertising, or downstream financial services to monetize users. Meme trading front ends, however, can charge a front-end fee on every transaction, creating a relatively linear relationship between revenue and volume. Using GMGN as an example, a static estimate based on a user trading fee of approximately 1% would imply potential weekly fees of around $10 million when weekly volume reaches $1 billion. Meme front ends therefore offer significantly higher monetization per user than conventional wallets and data tools. Platforms attract frequent traders through token discovery, data analysis, and execution, and then monetize those users directly through their trading activity. Once a platform becomes a user’s primary trading gateway, revenue can scale rapidly alongside market activity.

Another important development in August was the weakening relationship between Meme front ends and any single public blockchain. During the previous Meme cycle, infrastructure such as Pump.fun and Raydium concentrated substantial liquidity on Solana. Competition among BullX, Photon, Axiom, and other front ends could therefore largely be understood as a contest for access to Solana Meme users. As Robinhood Chain, BNB Chain, and other ecosystems began generating their own cyclical Meme booms, users increasingly needed platforms capable of discovering assets, identifying capital flows, and executing trades as soon as a new theme emerged. Trading front ends consequently began evolving into gateways for cross-chain speculative activity. Their value proposition also shifted from optimizing individual swaps toward aggregating high-frequency trading opportunities across different ecosystems.

2.3 Real x402 Transactions Return to Growth as High-Frequency Micropayments Begin Validating Automated Agent Billing

x402 activity recovered significantly in August, with daily real transaction counts at one point exceeding two million, a substantial increase from the previous low. Current x402 transactions occur primarily on Solana and are processed by facilitators such as PayAI and Figment. Demand is concentrated in infrastructure, public services, and AI-related services. Unlike conventional payment networks, x402 is designed primarily not for human consumers but for AI agents capable of autonomously discovering, accessing, and purchasing data, APIs, computing resources, and model services. Its transactions therefore display a characteristic high-frequency, low-value pattern. During August, daily real transaction value on x402 generally remained in the tens of thousands of dollars. Even when daily transaction counts reached the millions, the average payment was less than $0.015. Measured solely by payment value, the market remained at a very early stage. From the perspective of machines autonomously accessing and settling services, however, millions of daily transactions may indicate that agent micropayments are beginning to support real use cases at meaningful scale.

The value of x402 lies precisely in the micropayment demand that conventional payment systems struggle to serve efficiently. Traditional card payments typically involve fixed transaction costs, making them uneconomical for API calls worth only a few cents or even less than one cent. In practice, an AI agent may sequentially call multiple external services, including models, search tools, databases, RPC endpoints, and computing resources, generating a large number of independent requests each day. x402 combines the HTTP 402 Payment Required status with stablecoin settlement. When an agent requests a paid resource, the server can return a price directly. The agent authorizes payment locally and resubmits the request, integrating payment and service access into a single interaction flow. For machines, this removes the need to register accounts, link credit cards, manage multiple API keys, or purchase subscription credits in advance. Instead, they can automatically pay per use based on actual consumption. Agentic Payments could become a new commercial layer for machines, enabling AI agents to discover, evaluate, and purchase data, APIs, compute, inference, and other tools autonomously.

The current composition of x402 transactions further demonstrates that the protocol is not a conventional crypto-payment system. Activity is currently concentrated in infrastructure and public services because these products naturally lend themselves to billing by API call, model invocation, or computing-resource usage. BlockRun, for example, integrates multiple AI models and tools behind a single gateway. Through an OpenAI-compatible interface, agents can access models from providers including OpenAI, Anthropic, Google, xAI, and DeepSeek, paying directly in USDC on a per-call basis without separately registering accounts or managing subscriptions with each model provider. Its catalog currently covers approximately 100 models, including 76 chat and reasoning models, and extends to image, video, music, and voice-generation services. Each invocation can be settled on-chain through x402, combining model selection, price confirmation, payment, and delivery of inference results into a workflow that an agent can execute autonomously.

The importance of this model lies in its ability to address a basic but persistent challenge in AI-agent commercialization: how machines can autonomously purchase external services. Traditional API business models are generally designed around individual or corporate accounts. Users must register, obtain API keys, fund account balances or link payment cards, and settle charges through monthly billing. When an agent needs to invoke dozens or even hundreds of services dynamically, this framework creates substantial account- and payment-management costs. x402 instead embeds pricing directly into the service request, allowing the agent to decide whether to pay based on the quote returned by the server. A paid BlockRun request, for example, first returns an HTTP 402 response and payment details. After the client authorizes payment in USDC, it resubmits the request. The server verifies the payment, executes the API call, and returns both the service output and proof of payment. Payment is therefore transformed from a separate financial process outside the product into an automated component of the agent’s tool invocation.

High-frequency micropayments are also beginning to form a preliminary commercial flywheel. PayAI began charging for its facilitator service in February 2026. The first 1,000 settlements per merchant each month are free, after which it charges $0.001 per transaction. At that rate, a facilitator processing one million billable settlements per day would generate approximately $1,000 in daily revenue. If transaction volume expanded to ten million per day, the same fee of one-tenth of a cent per transaction would generate approximately $10,000 in daily revenue. The x402 infrastructure business model therefore differs materially from that of conventional payment providers. Rather than taking a percentage of relatively large payment values, it relies on extremely high frequencies of machine-to-machine calls to generate meaningful revenue from very low per-transaction fees. As the number of agents and the frequency with which each agent calls external services increase, transaction count may become considerably more important than aggregate payment value.

BlockRun’s pricing model reflects the same logic. Its AI-model services are priced according to the actual token cost charged by upstream providers, with an additional fixed fee of $0.001 for each paid request rather than a substantial percentage markup on model usage. The fee is almost negligible for an individual call, but invocation volume itself can generate recurring revenue from agents that operate continuously and make large numbers of model calls. The model is also highly scalable. Beyond LLMs, similar billing mechanisms can be applied to images, video, real-time data, search, RPC services, and computing environments. Any service that can be standardized as a machine request can potentially be converted into a machine payment.

The current daily transaction value on x402, measured only in tens of thousands of dollars, may therefore understate the network’s actual progress. In consumer payments, GMV and aggregate payment value are typically important measures of network value. For Agentic Payments, however, more relevant indicators may include Real Transactions, the number of active service providers, the number of paid APIs, and invocation frequency per agent. If a $0.005 data request can autonomously complete discovery, pricing, payment, and delivery, its monetary contribution may be negligible, but it nevertheless demonstrates that a machine can execute a complete commercial transaction without human intervention. As AI agents evolve from chat and task execution toward persistent operation, a single task may require sequential calls to models, search engines, data providers, RPC endpoints, computing resources, and trading services. One human instruction could ultimately generate dozens or hundreds of independent machine-to-machine settlements, allowing transaction counts to grow much faster than aggregate payment value.

3. Industry Developments

3.1 Nvidia’s Earnings Again Validate Demand for AI Compute: An Industry Inflection Point or a Capital-Expenditure Cycle?

On August 26, 2026, Nvidia reported results for the second quarter of fiscal 2027, once again significantly exceeding market expectations. Quarterly revenue reached $96.22 billion, up 106% year over year and 18% quarter over quarter. Net income totaled $59.69 billion, compared with $26.42 billion in the same period a year earlier. Adjusted earnings per share came in at $2.22, above the market consensus of $2.09.

Nvidia’s outlook for future demand attracted even greater attention. The company forecast revenue of approximately $108 billion for the following quarter, plus or minus 2%, implying year-over-year growth of roughly 89%. Management also projected that revenue could continue growing by approximately 70% in fiscal 2028, materially above previous market expectations. Nvidia’s shares rose 8.7% following the earnings release, their largest one-day gain since April 2025, while the Nasdaq advanced approximately 1.6%.

3.1.1 Earnings Indicate That Demand for AI Compute Has Yet to Peak

Nvidia’s growth continued to be driven primarily by its data-center business. The segment generated quarterly revenue of $89 billion, up 117% year over year and accounting for approximately 92.5% of the company’s total revenue. By comparison, all other businesses—including gaming, personal computing, robotics, and edge computing—generated approximately $7.2 billion in revenue, an increase of 27% year over year.

These figures show that Nvidia is no longer a relatively diversified semiconductor company. It has become an infrastructure supplier heavily concentrated on the AI data-center construction cycle. Its revenue growth depends directly on whether cloud-computing companies, AI-model developers, sovereign AI initiatives, and large enterprises continue expanding computing capacity.

By customer segment, approximately $48.7 billion of data-center revenue came from large cloud-computing customers, while around $40.3 billion came from AI cloud providers, industrial customers, enterprises, and other clients. Traditional hyperscalers remained the largest source of demand, but AI-native companies, sovereign computing projects, and private enterprise deployments were also establishing a second growth engine.

Order and capital-expenditure data provided further evidence that demand remained strong. Amazon, Microsoft, Alphabet, and Meta spent approximately $166 billion in aggregate capital expenditure during the previous quarter, an increase of around 27% quarter over quarter. Nvidia management estimated that capital expenditure by large cloud-service providers could rise from approximately $800 billion in 2026 to around $1.3 trillion in 2027.

Nvidia and AWS also announced plans to deploy approximately two million additional Nvidia GPUs across AWS’s global infrastructure. This represented not only a major procurement commitment but also evidence that AI infrastructure was expanding beyond model training into inference, agents, robotics, and enterprise applications.

Another important signal was the simultaneous growth of revenue and profitability. Nvidia maintained a GAAP gross margin of nearly 75% during the quarter even as its revenue doubled. This indicated that the company had not yet needed to rely on substantial price reductions to sustain growth. Blackwell Ultra, networking equipment, the CUDA software ecosystem, and Nvidia’s integrated AI Factory solutions continued to command strong pricing power.

If AI compute represented only a temporary period of overinvestment, the first signs would typically include falling chip prices, rising inventories, and contracting gross margins. None of these signals was yet clearly visible. At least based on the quarter’s results, demand was still growing faster than Nvidia’s effective supply capacity.

3.1.2 The Greatest Risk Is Not Disappearing Demand, but Excessive Concentration

Despite the strength of the results, Nvidia’s long-term risks became clearer. First, more than 90% of the company’s revenue came from data centers, with a substantial share ultimately generated by a small number of major cloud providers. If Amazon, Microsoft, Google, and Meta reduced capital expenditure, Nvidia would find it difficult to offset the shortfall through its gaming or automotive businesses.

AI investment by cloud providers also cannot remain permanently detached from economic returns. If revenue from AI services, enterprise conversion rates, and inference usage fail to keep pace with server investment, capital-expenditure growth will eventually slow. The market would then reassess how many orders currently viewed as “long-term structural demand” actually represented advance purchases or duplicative infrastructure build-outs.

Second, the supply chain has become a constraint on growth. In addition to advanced wafer fabrication, Nvidia depends on HBM, advanced packaging, networking equipment, electricity, and data-center cooling systems. To secure supply, the company has made large procurement commitments for memory and other critical components. Related memory-purchase commitments have reportedly risen to approximately $160 billion. While these arrangements support delivery capacity, they also increase inventory and contractual risks if future demand falls below expectations.

Third, competition is expanding from individual GPUs to complete computing systems. AMD, Broadcom, and cloud providers developing their own ASICs are all seeking to reduce their dependence on Nvidia. Nvidia’s current advantage extends beyond chip performance to the switching costs created by CUDA, networking, software libraries, and its developer ecosystem. The key issue to monitor is therefore not merely whether competitors can launch faster chips, but whether customers begin migrating more workloads to internally developed platforms.

China represents an additional source of uncertainty. Nvidia did not include revenue from Chinese data-center computing in its guidance for the following quarter, reflecting the potential effect of export restrictions and geopolitical policy on its addressable market. Demand from Chinese customers has not disappeared, but there remains substantial policy uncertainty over whether Nvidia can legally deliver compliant products.

The earnings report therefore demonstrated that demand for AI compute remained strong, but it could not by itself prove that the current intensity of investment could continue indefinitely. Key indicators to monitor include cloud-provider capital expenditure, the growth rate of Nvidia’s data-center revenue, gross margins, supply commitments, customer concentration, and the rate at which AI-service revenue grows relative to infrastructure investment.

As long as end-market AI revenue can gradually cover the enormous capital expenditure involved, Nvidia will remain in a structural growth cycle. If chip procurement continues to increase while customers’ commercialization efforts lag materially behind, however, the current boom could evolve into a capital-expenditure cycle of unprecedented scale.

3.2.1 The US Treasury Expands Bond Buybacks: How a Liquidity Shock Ignited a Crypto Short Squeeze

In mid-August 2026, the crypto market experienced an unusually rapid rally. After trading within a range for an extended period, BTC broke above $67,000, while ETH gained as much as approximately 18% in one week. During the most volatile four-hour period, around $1.4 billion in crypto short positions were forcibly liquidated. As the rally continued, total short liquidations for the week exceeded $4 billion.

The advance was not driven independently by a single crypto-project upgrade or regulatory announcement. The initial catalyst that altered market pricing was the US Treasury’s announcement of a significant expansion in its long-dated Treasury buyback program. Treasury yields and the US dollar subsequently fell in the short term, supportive policy developments improved risk appetite, and heavily concentrated short positioning ultimately amplified a conventional rally into a violent short squeeze.

3.2.2 Why Treasury Buybacks Can Affect BTC and ETH

A US Treasury bond buyback involves the Treasury purchasing previously issued government debt in the secondary market and adjusting the debt structure through refinancing or new issuance. Its primary objectives typically include improving the liquidity of older securities, reducing market frictions, and easing pressures caused by an oversupply of bonds at specific maturities.

On August 19, the Treasury announced that it would at least double the scale of its long-dated bond buybacks. Markets quickly interpreted the move as an official response to liquidity pressures in the Treasury market. Long-dated Treasury prices rose, yields declined, and the US dollar weakened in tandem. As the relative appeal of holding cash and bonds diminished, risk assets gained room for repricing.

Changes in interest rates affect crypto assets through discount rates and financing costs. When long-term yields decline, the relative valuations of equities, gold, and cryptocurrencies—which do not rely on fixed interest payments—generally receive support. A weaker dollar also makes dollar-denominated BTC and ETH more attractive to non-US investors.

The relationship between Treasury yields and crypto prices is not consistently one-directional, however. As the chart shows, the 10-year Treasury yield fell significantly on the day of the announcement but subsequently rebounded, while BTC and ETH remained at higher levels. This suggests that the bond buyback acted more as a trigger than as the sole explanation for the entire rally.

Policy expectations provided a second source of support. During the same period, Trump urged Congress to advance crypto market-structure legislation, while the CFTC signaled that it could use its existing authority to relax certain crypto rules. The simultaneous improvement in macro liquidity and regulatory expectations made traders less willing to maintain their previous bearish positions.

Capital flows provided further confirmation of the shift in sentiment. During the week ended August 21, US spot BTC ETFs recorded approximately $1.9 billion in net inflows, while ETH ETFs attracted around $697 million, bringing combined inflows to approximately $2.6 billion—the strongest performance since October 2025. Turnover across the two ETF categories exceeded $29 billion, more than doubling from the preceding week.

From Rally to Short Squeeze: How Leverage Amplified the Move

Before the event, BTC had traded in a range of approximately $62,000–$67,000 for several weeks. Because macro conditions remained tight and crypto had previously underperformed US equities, many traders expected the upper end of the range to continue limiting prices. Short positions, put options, and bearish trend strategies gradually became concentrated.

When BTC broke above $67,000, short positions first began hitting stop-loss levels. When margin on perpetual-futures positions becomes insufficient, exchanges must purchase the underlying asset in the market to close those shorts. This forced buying pushed prices higher, triggering another wave of liquidations at higher levels and creating a positive feedback loop of “rally–liquidation–buying–further rally.”

Market data showed that approximately $1.4 billion in short positions was liquidated during the initial four-hour period. BTC rose by more than 20% within a week and briefly exceeded $77,000. ETH was more sensitive to changes in liquidity and leverage, and its short-term gains significantly outpaced those of BTC.

Liquidation volume, however, cannot be treated as equivalent to new capital. Forced purchases by short sellers merely close existing positions rather than represent new positions initiated by long-term investors. Assessing the durability of the rally therefore requires distinguishing among three sources of capital: spot net inflows through channels such as ETFs, futures short covering, and highly leveraged long positions established in response to the breakout.

If the rally depends primarily on short liquidations, price momentum could weaken rapidly once the available short positions have been exhausted. Conversely, continued growth in ETF inflows, stablecoin supply, and on-chain activity would indicate that the short-term squeeze was being converted into more durable spot demand.

The central significance of the event was not that Treasury bond buybacks necessarily cause cryptocurrencies to rise. Rather, it demonstrated how liquidity policy, regulatory expectations, and leverage structures can reinforce one another over a short period. Treasury buybacks reduced market concerns about long-term interest rates, policy statements altered risk appetite, and excessively concentrated short positions amplified the repricing into a nonlinear market move.

Going forward, investors should closely monitor the 10-year Treasury yield, the US Dollar Index, net ETF subscriptions, open interest in perpetual futures, and funding rates. If prices rise while open interest recovers rapidly and funding rates increase significantly, the risk may shift from crowded shorts to crowded longs. The rally will have a more durable foundation only if spot capital continues to enter the market.

Data Source:

• Dune, https://dune.com/gateresearch/prediction-markets-overview
• Paymentscan, https://paymentscan.xyz/
• Gate, gate.com/announcements/article/51452
• CoinGecko, https://www.coingecko.com/research/publications/rwa-report-2026
• Investing, https://www.investing.com/indices/us-30-futures-chart?cid=1175152
• Coinmarketcap, https://coinmarketcap.com/etf/
• Tradingview, https://www.tradingview.com/markets/indices/quotes-major/
• Nvidia, https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Announces-Financial-Results-for-Second-Quarter-Fiscal-2027/default.aspx
• Cointelegraph, https://cointelegraph.com/news/solana-validators-approve-proposal-to-accelerate-sol-disinflation?utm_campaign=rss_partner_inbound&utm_medium=rss&utm_source=rss_feed
• Solana, https://forum.solana.com/t/simd-0550-proposal-to-double-disinflation/4874

Gate Research is a comprehensive blockchain and cryptocurrency research platform that provides deep content for readers, including technical analysis, market insights, industry research, trend forecasting, and macroeconomic policy analysis.

Disclaimer

Investing in cryptocurrency markets involves high risk. Users are advised to conduct their own research and fully understand the nature of the assets and products before making any investment decisions. Gate is not responsible for any losses or damages arising from such decisions.