Bitcoin miners are pouring large sums into AI and high-performance computing (HPC) ventures, but early financial results show that the shift is still far from economically catching up with the scale of the investment. According to BlocksBridge Consulting’s latest Miner Weekly update, miners and AI-adjacent data center operators have committed tens of billions to capital assets—much of it happening before meaningful revenue ramps up.

BlocksBridge reported that 15 publicly listed Bitcoin miners and AI data center companies collectively spent $30.7 billion on capital assets in their latest 2026 reporting periods. That figure is already 42.6% higher than the $21.53 billion they spent across all of 2025. For investors, the key question is whether current AI/HPC revenue growth can narrow the gap between upfront spending and cash returns fast enough to justify the pivot.

Key takeaways

• BlocksBridge Consulting says 15 public Bitcoin miners and AI data-center companies spent $30.7 billion on capital assets in their latest 2026 reporting periods—42.6% more than total 2025 capex.
• Nine comparable miners spent $5.11 billion on capital assets in the first half of 2026 while generating $341.2 million in directly reported AI and HPC revenue (about a 15-to-1 capex-to-revenue ratio).
• AI and HPC revenue from those nine miners rose to $205.8 million in the second quarter, up 52% quarter-on-quarter.
• BlocksBridge cautions that converting power and land advantages into AI-ready infrastructure requires expensive build-outs, including substations, buildings, cooling, networking—and sometimes GPUs.

Capex surges, revenue lags in the AI pivot

The strongest signal in BlocksBridge’s data is the imbalance between spending and monetization. While AI and data-center strategies are widely viewed as diversification pathways for miners facing cyclically tough mining economics, BlocksBridge’s numbers suggest the transition remains capital intensive.

BlocksBridge calculated capital spending by combining cash purchases with allocations to hardware, property, equipment and other productive assets—netting out proceeds and refunds from asset sales. The methodology matters because it points to a “build” phase rather than a purely expansionary one: companies are acquiring and deploying physical assets at speed, even as revenue capture is still ramping.

Drilling into Bitcoin miners specifically, BlocksBridge noted that nine comparable miners invested $5.11 billion in capital assets during the first half of 2026. Yet those firms generated only $341.2 million in directly reported AI and HPC revenue during the same window. The resulting ratio—roughly 15-to-1—illustrates how far the industry is from turning capital deployment into proportionate operating returns.

What is changing: faster AI/HPC revenue growth

Despite the gap, BlocksBridge reported signs of acceleration. In the second quarter, the same group of nine miners generated $205.8 million from AI and HPC businesses, representing a 52% quarter-on-quarter increase. BlocksBridge highlighted Core Scientific, TeraWulf and Bitdeer among the companies showing gains.

For readers watching diversification outcomes, the practical implication is that the pivot may be entering a more revenue-generating stage—at least for some participants. However, the magnitude of earlier spending underscores that even sharp quarter-to-quarter growth may still be insufficient to erase the balance-sheet effect of large capex programs in the near term.

What investors should watch next is whether accelerating revenue translates into improving margins and more consistent demand. BlocksBridge’s figures focus on “directly reported” AI and HPC revenue; the market will likely scrutinize whether additional segments scale without requiring equally steep follow-on investments.

Why the transition is expensive: power and land aren’t enough

BlocksBridge also framed why miners can’t simply repurpose existing infrastructure and expect AI profits quickly. In its analysis, the firm said that power contracts and available land may provide a starting advantage, but turning those inputs into AI-ready capacity involves additional, costly components.

According to BlocksBridge, the build-out can require substations, buildings, cooling systems, networking equipment, and in some business models, GPUs. This helps explain why capex-to-revenue ratios can remain elevated: building AI-capable data center and compute infrastructure is not just an incremental upgrade—it is a construction and integration project with multiple dependency layers.

At the same time, the source notes that it remains unclear whether any recovery in Bitcoin’s price will ease near-term pressure on miners that still operate sizable mining fleets. When cash flows from traditional mining are volatile, the timing of AI revenue maturation becomes even more important.

Broader market signals: miners still betting big as policy improves liquidity

While BlocksBridge’s report centers on AI/HPC economics, the surrounding market context matters because it influences how much funding and operational stress miners can absorb. The article points out that Bitcoin rose more than 13% over the week and returned above $72,000 following a US Treasury announcement that it would at least double the maximum size of its long-term bond buybacks to $4 billion per operation. The move was described as intended to improve liquidity in the Treasury market and was associated with lower yields and a boost to risk appetite.

Even with that supportive backdrop, the central takeaway from BlocksBridge remains: AI diversification is expensive upfront. For investors, this creates a tension—markets may improve financing conditions while the underlying monetization timeline lags behind construction.

Separately, the pivot to AI-linked compute and power has also appeared in investment products. CoinShares announced a strategic change to its industry tracking exchange-traded fund, rebranding it as the CoinShares Bitcoin Mining and Digital Power ETF (WGMI). CoinShares says the fund holds 29 companies spanning bitcoin miners, data center operators, AI semiconductors, power generation and HPC. As of the announcement, the ETF reported $222.4 million in assets under management, and CoinShares described the theme as “the businesses powering the digital economy,” according to its listing page.

For market participants, the launch and rebranding of a targeted ETF can be interpreted as demand from investors for exposure beyond pure mining. Still, such products ultimately depend on underlying company execution—especially whether AI/HPC revenue continues to grow fast enough to justify large capital programs.

Going forward, the most important uncertainty is whether rising AI and HPC revenues can outpace the continuing cost of expansion and integration. BlocksBridge’s quarter-on-quarter growth is encouraging, but investors should monitor whether that momentum persists, improves profitability, and reduces the still-wide spending-to-return gap highlighted in its capex-to-revenue calculations.

Binance has launched Agent OS, a new developer platform designed to let AI agents access market data, monitor user accounts, and execute crypto trades directly on the exchange. The announcement frames Agent OS as an infrastructure layer that can be connected to popular AI tools, with controls that aim to keep permissions and risk limits under the user’s authority.

According to Binance, Agent OS supports AI environments including ChatGPT, Claude Code, Codex, and Cursor. Users can authorize agents to view account information and place orders only within configured permissions and limits, and they can assign agents to dedicated subaccounts so trading activity and funds remain compartmentalized.

Key takeaways

• Agent OS gives AI agents access to Binance market data, the ability to monitor user accounts, and the option to execute trades.
• Binance’s model is authorization-based: users define which actions agents can take and impose trading limits.
• Agents can be tied to dedicated Binance subaccounts for clearer separation of funds and activity.
• Binance says it can observe trades executed via Agent OS but does not see the agent’s external data sources or internal decision-making.
• Agent OS also links agents to Binance’s payment and onchain tools for wallet and onchain-service interactions.

What Binance’s Agent OS is designed to do

Agent OS is positioned as a bridge between AI applications and exchange operations. Binance states that developers can connect agents to market information and to user account functionality, then grant those agents the ability to place trades through the exchange under a permissioned setup.

In practical terms, this matters because it reduces the friction of building agent-driven trading systems. Instead of relying solely on custom integrations, users can route trading actions through a platform that is already integrated with Binance’s account and execution infrastructure. At the same time, Binance emphasizes user control by allowing permissions to be configured and access to be revoked at any time.

Permissions, subaccounts, and the limits of what Binance can see

Binance’s announcement highlights a key operational safeguard: users can assign agents to dedicated subaccounts. That approach can help separate balances and trading activity for different strategies or different agent instances, which is particularly relevant when multiple automated systems operate under the same main account.

Binance also describes a visibility boundary. It says it can monitor the trades placed through Agent OS, but it cannot see an agent’s external information sources, interpretation, or decision-making logic—elements that occur within the user’s chosen AI application. That separation is important for privacy and for reducing the need to centralize all agent reasoning inside the exchange environment.

How this fits into the broader “agents” push by exchanges

Agent OS arrives amid a broader trend: crypto trading venues are moving from basic automation toward infrastructure that supports more autonomous AI-driven behavior.

Earlier in the year, Coinbase launched “Coinbase for Agents” in June. That tool also targets AI models such as ChatGPT and Claude, enabling connections to user accounts so models can execute trades and strategies, alongside support for agent-driven payments through Coinbase’s x402 protocol.

Different exchanges are taking different stances on autonomy. In July, Kraken unveiled an AI-powered investing assistant that monitors markets and recommends trades based on users’ goals and risk preferences, but requires user approval before executing trades.

Other players are extending the concept beyond trading. OKX launched a beta marketplace where AI agents can find work, transact using stablecoin payments, and hire other agents for tasks, backed by an onchain reputation system.

Taken together, the sector is converging on a common idea—agents should be able to interact with financial rails—but it’s still diverging on the degree of autonomy and how much responsibility belongs to the user versus the system.

From trading to payments and onchain interaction

Beyond order placement, Binance says Agent OS can connect agents to its payment and onchain tools. The stated goal is to allow agents to make payments and interact with wallets and other onchain services.

This broader scope is a notable shift from “agent as a trading bot” toward “agent as an onchain operator.” If agents can perform payments and wallet interactions in addition to trading, they can potentially be used for a wider range of workflows—such as managing funds across strategies, executing routine onchain actions, or coordinating multi-step operations that blend exchange and onchain activity.

However, the same expansion also raises the stakes for governance and risk controls. Binance’s emphasis on permissions, subaccounts, and revocation becomes even more important when an agent can potentially do more than place orders.

Why industry leaders see agents as a major onchain driver

Binance is not operating in a vacuum. The announcement echoes comments from other crypto executives who have argued that AI agents could take on a meaningful portion of onchain activity. Coinbase CEO Brian Armstrong and Circle CEO Jeremy Allaire have both pointed to the potential for agents to become active participants in onchain ecosystems.

Binance co-founder Changpeng Zhao has also described cryptocurrency as a “native currency” for AI agents, reinforcing the idea that exchanges and payment infrastructure could become the operational backbone for agent-driven finance.

Agent OS can be viewed as a concrete attempt to operationalize that vision—turning “agents will use crypto” into “agents can securely interact with exchange systems.” The key question for users and developers will be how quickly these platforms converge on shared standards for authorization, auditing, and safety.

For now, investors, traders, and builders should watch how Agent OS performs in real deployments—especially around permission granularity, subaccount segregation, and what types of agent workflows users actually adopt. The most important unknown is how these exchange-based agent systems will balance autonomy with practical safeguards as AI-driven onchain activity scales.

Coinbase CEO Brian Armstrong has said Bitcoin could be entering its next bull cycle as the asset trades above $72,000 ahead of a key U.S. Senate vote on crypto legislation.

Armstrong sees a possible Bitcoin bull cycle

In an Aug. 20 interview with CNBC, Armstrong said the crypto market “is likely at the starting point of the next bull market.”

Armstrong based his view partly on the length of the recent downturn and Bitcoin’s position in its market cycle. He also identified the Senate’s scheduled action on the Digital Asset Market Clarity Act and Bitcoin’s historical performance during the final three months of the year as possible catalysts.

While discussing the coming congressional vote, Armstrong described Sep. 15 as “what’s most important next” for the crypto industry. Passage of market-structure legislation could give digital asset companies clearer rules in the United States, according to the Coinbase executive.

October, November and December have also produced some of Bitcoin’s strongest historical returns, CoinGlass data shows. Traders commonly call October “Uptober” because of that record, although past monthly returns do not establish how the asset will perform in 2026.

Bitcoin broke its October winning streak in 2025, when the asset posted its first loss for the month since 2018, according to Reuters. The decline showed that seasonal patterns can fail when macroeconomic pressure, weak risk appetite, or heavy derivatives positioning outweigh historical trends.

Armstrong’s cycle argument also follows Bitcoin’s April 2024 halving, which reduced the reward paid to miners from 6.25 BTC to 3.125 BTC per block. Earlier halvings preceded major rallies, but each cycle developed under different liquidity, regulatory, and economic conditions.

Bitcoin price has returned above $72,000

Bitcoin was trading near $72,660 on Aug. 20 after rising about 6.2% during the latest session. The asset reached an intraday high of roughly $72,868, extending a rally that lifted it from below $65,000 earlier in the week.

As previously covered by crypto.news, Bitcoin gained 11.4% within 24 hours and approached $72,000 after U.S. Treasury buyback changes, ETF inflows, and forced short liquidations supported the move.

More than $1 billion in bearish crypto positions were liquidated within one hour during the initial breakout, according to CoinGlass data cited in the report. When Bitcoin moved through resistance between roughly $65,000 and $67,000, traders holding leveraged short positions had to close their bets, adding forced buying to the rally.

US spot Bitcoin ETFs provided another source of demand. SoSoValue recorded $517 million in net inflows on Aug. 19, the products’ strongest daily intake since May. The total exceeded the roughly $172 million collected across all of July.

Earlier in August, the funds had already recorded four consecutive positive sessions. Farside Investors data showed that they attracted $137.6 million on Aug. 6, bringing inflows over the four-session period to about $763.6 million, according to an earlier report on Bitcoin ETF demand.

The latest price increase has also taken Bitcoin back above the $69,000 to $70,000 area that had acted as resistance. Technical data cited in a recent Bitcoin chart review placed the next resistance around $72,500, while a daily close below $69,000 could weaken the breakout.

CLARITY Act faces a procedural vote on September 15

The U.S. Senate’s official cloture record shows that Majority Leader John Thune filed a motion on Aug. 8 to proceed with H.R. 3633, the Digital Asset Market Clarity Act.

The motion is expected to face a procedural vote on Sep. 15 after senators return to Washington on Sep. 14. A successful vote would allow the Senate to begin formally considering the legislation, but it would not amount to final passage.

Senators could still debate and amend the proposal before holding a separate approval vote. If the Senate passes a version that differs from the House measure, lawmakers from both chambers would need to reconcile the texts before sending the legislation to President Donald Trump.

The House approved its version by a 294–134 vote in July 2025, with 78 Democrats joining Republicans. In May 2026, the Senate Banking Committee advanced its portion of the legislation by a 15–9 vote after Democratic Sens. Ruben Gallego and Angela Alsobrooks supported it.

Reaching the Senate floor will require 60 votes. Republicans control 53 seats, leaving the bill dependent on Democratic or independent support as lawmakers continue negotiating several disputed sections.

Reuters reported that unresolved issues include political ethics restrictions, stablecoin rewards, anti-money-laundering requirements, decentralized finance, and the treatment of tokenized securities. The proposal would also divide regulatory duties between the Securities and Exchange Commission and the Commodity Futures Trading Commission.

Under the Senate draft, digital commodity exchanges, brokers, and dealers would be treated as financial institutions under the Bank Secrecy Act. Covered companies would have to follow customer identification, due diligence, and anti-money-laundering rules.

The legislation would allow certain crypto businesses to raise as much as $50 million annually and $200 million in total without completing full SEC registration. It would also create tests for deciding whether a decentralized finance platform is sufficiently decentralized or should face requirements similar to those of an intermediary.

Armstrong maintains a bullish long-term Bitcoin forecast

During a separate Fox Business interview on Aug. 20, Armstrong said Bitcoin could reach between $300,000 and $400,000 by 2030.

“I think over the next couple of years — if I say 2030 — I think it’s very likely we’ll see $300,000 and $400,000 Bitcoin, and we’ll see how it goes.”

The forecast would require Bitcoin to rise more than fourfold from its price near $72,660 to reach the lower end of Armstrong’s range. A move to $400,000 would represent an increase of about 450%.

Coinbase shares also advanced alongside the crypto market. COIN traded near $171.34 on August 20, up about 7%, after reaching an intraday high of $174.75. The U.S.-listed exchange had a market capitalization of approximately $45.2 billion at the time.

Coinbase reported a $359 million net loss for the second quarter, while Bitcoin generated 12% of the company’s revenue, down from more than half historically. Subscription and services revenue reached $555 million, compared with $6 million per quarter in 2020, according to the company’s reported figures.

A group of Bitcoin wallets dormant since 2014 has transferred 1,214.42 BTC worth about $86 million as the cryptocurrency traded near a weekly high of $72,400.

Dormant Bitcoin wallets move $94 million

Bitcoin.com reported on Aug. 20, citing btcparser.com data, that 28 long-inactive wallets moved a combined 1,314.41 BTC between Aug. 19 and Aug. 20.

The Bitcoin was worth approximately $94.03 million at the time of the transactions. Wallets created in 2014 accounted for 1,214.42 BTC, or 92.4% of the total, with the coins valued at about $86 million.

Rather than moving through one large transaction, much of the activity appeared in matching amounts. Twenty-one transfers each carried 50 BTC from addresses created in November or December 2014, according to the report.

Several transactions were recorded in the same Bitcoin blocks, including block 963203. Their timing, matching sizes, and common address format suggest that the wallets may have been controlled by one or a small number of holders, although the available blockchain data does not establish their ownership.

Alongside the 2014 coins, three wallets created in 2016 transferred a combined 79.99 BTC. Two addresses dating to 2017 moved another 20 BTC during the same 24-hour period.

The report said another wallet, first seen on Dec. 26, 2014, separately transferred 150 BTC worth approximately $10.73 million. Its coins also went to a newly created address with no public entity label.

Bitcoin moves from legacy to newer addresses

Most of the 2014 transactions sent funds from Pay-to-Public-Key-Hash, or P2PKH, addresses to Pay-to-Witness-Public-Key-Hash addresses known as P2WPKH.

P2PKH is a legacy Bitcoin address format commonly identified by addresses beginning with “1.” P2WPKH addresses use Segregated Witness and generally begin with “bc1q,” offering smaller transaction sizes and lower fees than older address types.

Moving coins between the two formats can occur when a holder reorganizes self-custodied funds or adopts a newer wallet setup. Blockchain records show where the Bitcoin went, but they do not reveal why the owner moved it or whether the receiving addresses remain under the same control.

Arkham Intelligence’s public labels had not associated the destination addresses with centralized exchanges when the report was published. Without an exchange link, the transfers alone do not prove that the holders intend to sell the Bitcoin.

Blockchair’s privacy tool gave several of the 50 BTC transfers a score of 22 out of 100, according to Bitcoin.com. The tool identified about four privacy concerns, including the repeated use of the same address among transaction inputs.

Consolidating multiple holdings can make future spending easier, but placing several inputs into one transaction may also reveal links between addresses. Such links can help blockchain analysts group addresses that may belong to the same entity, even when the owner’s identity remains unknown.

Early holdings have risen at least 16,600%

Bitcoin traded between approximately $310 and $427 during November and December 2014, when the main group of wallets first received or became associated with the coins.

Using the upper end of that range and the value at the time of the latest transfers, Bitcoin.com estimated that the holdings had appreciated by at least 16,645%. The percentage represents the increase in Bitcoin’s market price rather than a confirmed realized return because blockchain data does not show that the coins were sold.

At $427 per Bitcoin, acquiring 1,214.42 BTC would have required roughly $518,000 before fees. The same amount was worth about $86 million when the wallets became active again.

Bitcoin’s price later fell to between $152 and $170 in January 2015, leaving holders from late 2014 facing steep paper losses before the asset recovered over the following decade. The wallet activity therefore covers coins held through several Bitcoin market cycles.

Dormant-wallet transfers have appeared several times during 2026. As crypto.news reported in May, a wallet inactive since November 2013 moved 500 BTC worth about $40 million to a new address with no known exchange connection.

CryptoQuant CEO Ki Young Ju described the May transaction as “classic OTC prep, not dump pressure,” citing its low fee and non-exchange destination. No comparable analyst assessment has been provided for the latest 1,214 BTC movement.

Later that month, a separate whale transferred 2,650 BTC worth approximately $203 million to FalconX and Cumberland. Onchain Lens, citing Arkham data, said the wallet retained nearly 6,000 BTC valued at about $462 million after the transactions.

Unlike the latest movements to unlabeled addresses, the May transfers reached named crypto trading firms. Even then, the transactions did not confirm that the holder had sold the coins, since trading companies can also handle custody and over-the-counter deals.

Dormant wallets also face a US ownership dispute

Long-inactive Bitcoin addresses have also entered a legal dispute in New York, where a plaintiff using the name Noah Doe has sought control of 39,069 wallets under the state’s lost-property law.

In July, a listed wallet transferred 30 BTC worth about $1.88 million after nearly 15 years without an outgoing transaction. Galaxy Research said other addresses named in the lawsuit had also begun moving funds.

The plaintiffs have argued that the listed wallets qualify as abandoned property under Article 7-B of New York’s Personal Property Law. A defendant claiming control of one address asked the court to dismiss the case, arguing that a Bitcoin address is a data string rather than a legal entity that can be sued.

No public information links the 2014 wallets behind the latest $86 million movement to that lawsuit. Their activation nevertheless shows why inactivity alone cannot establish that a wallet has been abandoned or that its owner has lost access to the private keys.

For US taxpayers, the tax result depends on whether the Bitcoin merely moved between addresses controlled by the same person or changed ownership. The Internal Revenue Service states that transferring cryptocurrency between wallets, accounts, or addresses owned by the same taxpayer is not a taxable event.

A sale or exchange would receive different treatment. The IRS requires taxpayers to calculate capital gains or losses using the difference between the amount received and the adjusted cost basis, while records used to identify particular Bitcoin units should include their acquisition date, basis, disposal date, and fair market value at disposal.

Binance has launched Agent OS, a developer platform that allows AI agents to access market data, monitor user accounts and execute crypto trades on the exchange.

According to a company announcement, the platform supports AI tools including ChatGPT, Claude Code, Codex, and Cursor, allowing users to authorize agents to view account information and place trades within configured permissions and limits.

Users can assign agents to dedicated subaccounts to separate funds and trading activity, configure their permissions and revoke access at any time. Binance said it can monitor trades placed through Agent OS but cannot see an agent’s external information sources, interpretation or decision-making, which occur within the user’s chosen AI application.

Agent OS also connects agents to Binance’s payment and onchain tools, allowing them to make payments and interact with wallets and other onchain services.

Related: Bitcoin miners pour billions into AI as capex outpaces revenue 15-to-1

Crypto exchanges move deeper into AI agents

Binance joins a growing group of crypto exchanges opening their trading infrastructure to AI agents.

Coinbase launched Coinbase for Agents in June, allowing AI models including ChatGPT and Claude to connect to user accounts and autonomously execute crypto trades and strategies. The tool also supports agent-driven payments through Coinbase’s x402 protocol.

Source: Coinbase

Other exchanges have taken different approaches to AI autonomy. In July, Kraken unveiled an AI-powered investing assistant that monitors markets and recommends trades based on users’ financial goals and risk preferences, but requires user approval before executing a trade.

The push has extended beyond trading. OKX launched a beta marketplace where AI agents can find work, transact autonomously and hire other agents for tasks, using stablecoin payments and an onchain reputation system.

Several crypto industry leaders, including Coinbase CEO Brian Armstrong and Circle CEO Jeremy Allaire, have argued that AI agents could soon account for a significant share of onchain activity. Binance co-founder Changpeng Zhao has expressed a similar view, describing cryptocurrency as the “native currency” of AI agents.

Magazine: 200,000 fake AI ‘victims’ deployed to scam bait online fraudsters