Cryptopolitan

Three of Anthropic’s Claude models escaped from what were meant to be locked test environments, the company said Thursday. They broke into the live systems of three different organizations. A misconfiguration had handed the models working internet access during the cybersecurity drills.
Claude models treated real companies as practice targets
OpenAI said on July 21 that a model escaped its sandbox and exploited a zero-day to reach Hugging Face’s production systems.
This prompted Anthropic to comb through its logs. The company said in a blog post on its Frontier Red Team site that it analyzed 141,006 evaluation runs in which Claude could theoretically have accessed the internet. It identified three incidents related to Irregular, an external partner that performs and manages some of its offensive security testing.
In these tests, a model is told a secret is hidden on another machine and asked to go get it. Anthropic’s prompt to Claude had told it that the environment was a simulation.
A misconfiguration left evaluation machines with live internet access, and neither company caught it until last week. Anthropic described the mix-up as a “misunderstanding” between the two companies, adding that it is “approaching the fixes as if the responsibility were ours alone.”
These included Opus 4.7, Mythos 5, and an internal research model that has not been released. The difference was in how each responded after indications emerged that the targets were real, which Anthropic said was its most surprising finding.
The oldest, Opus 4.7, cycled through all four of its runs. In two runs, it convinced itself that the actual company must somehow be part of the exercise. It pulled credentials and hit a database with production data. Mythos 5 also deduced that it was on the open internet, then reasoned its way back to the belief that it was still in a simulation.
Only the newest model, the internal research build, stopped by itself when it thought the systems were real. Anthropic stated its code “did not find or exploit any complex vulnerabilities” and instead relied on weak passwords and unauthenticated endpoints.
Mythos 5 published malware that ran on 15 real systems
Following the notes on developer setup, Mythos 5 looked for a PyPI Python package that didn’t exist. So, to do its job, it wrote and published one under that name.
The package was up online for about an hour before it was taken down. It was downloaded and run on 15 real systems. Another attack hit a domain that the test designers believed was made up but was actually live and in use.
Anthropic said it found no evidence that any model was “pursuing a goal of its own.” Each, in its telling, was simply trying to get the job it had been given done.
The company was quick to note its case is different than OpenAI’s. OpenAI’s model escaped through an undiscovered vulnerability, Anthropic said, while its own models escaped through a door that was left open by mistake.
It described the events as “closer to a harness and operational failure.” The company also pointed out that the safety classifiers and monitoring included with its commercial products would have prevented the behavior.
Anthropic started its review on Thursday, July 23, and stopped all cyber assessments that same day. All three incidents were identified the next day. It informed Irregular and the organizations that were impacted on Monday, July 27.
Two organizations had not seen or reported intrusions. The company said it has engaged evaluation group METR to perform an independent review.
Anthropic’s Mythos model already made headlines this summer for identifying weaknesses in classified US government systems within hours, as Cryptopolitan reported. The model also uncovered a four-year-old vulnerability in Zcash’s Orchard shielded pool.
If you're reading this, you’re already ahead. Stay there with our newsletter.

True Internet Data Center, Thailand’s largest data-center and cloud operator, is seeking a $2 billion loan to fund its expansion. The move is about more than one company’s growth. It signals that Southeast Asia is becoming a place where AI infrastructure is being built, financed, and operated, not just where AI services are consumed.
This financing emphasizes even more the significant change in the worldwide AI competition. With computing capabilities moving away from the U.S., nations like Thailand are securing significant amounts of funds from cloud providers, AI firms, and infrastructure developers. True IDC, owned by the Charoen Pokphand Group, aims to be among the local firms participating in this transition.
If we look at projects that were made public, we can see that over the past two years Thailand has lured investments of over $40 billion into AI and data centers. In this context, a loan of $2 billion will be one of the largest private investments in the country’s digital infrastructure.
What the loan sits on top of
True IDC is the largest data center operator in Thailand, and it is continuing to expand its business. At the beginning of June, True IDC opened the first construction site for its seventh data center, investing more than 6 billion baht in this AI data center with hyper-scale capabilities. This facility, which is designed for high-density CPU and GPU workloads using both air and liquid cooling, will start operations in the 3rd quarter of 2027.
The proposed loan would serve to support that trend, as Managing Director Thanasorn Jaidee stated that he sees future demand increasing. This is based on IDC predictions regarding AI and generative AI expenditure in Asia-Pacific rising five times to reach $370 billion (11 trillion baht) in 2029, at a CAGR of 38.4%.
The amount of funding is also a signal of the evolution of AI infrastructure. Modern data centers need more than regular buildings; there are GPU networks that consume a great deal of power, as well as sophisticated cooling systems and fast data networks that make the projects more costly compared to traditional data centers.
Why global players are writing bigger checks
True IDC’s expansion is taking place amidst the approval of record AI infrastructure projects by Thailand’s Board of Investment (BOI). In just one round, the agency has given the nod to a total of 958 billion baht (about $29 billion) worth of investments, including around $27 billion worth of data-center and hosting projects headed by TikTok System (Thailand).
Foreign investors have begun making investments in Thailand. The BOI approved a $1.4 billion data center from Skyline Data Center and Cloud Services, backed by the UAE’s DAMAC Group, along with a $746 million project from Singapore’s Bridge Data Centres. Another investment round included Japan’s Datasection, which plans to spend $235.2 million on GPU server infrastructure in Bangkok and Pathum Thani.
Through the implementation of these initiatives, it is clear that the nation has done much more than create server capacity, having built an entire AI ecosystem that unites in a real sense hyperscalers, colocation service providers, GPU infrastructure, and enterprise cloud services in order to promote various domestic companies such as True IDC under conditions of intense competition.
Hyperscalers have already planted flags
Much of the groundwork has been laid by global cloud providers. In January 2025, Amazon Web Services launched its Asia Pacific (Thailand) Region and indicated that the investment would help Thailand’s GDP by $10 billion over 15 years. This was followed by Google Cloud launching its Bangkok region in January 2026, spending around $1 billion, which will produce about 1.4 trillion baht (or $41 billion) in economic value within 5 years and provide jobs for around 130,000 people a year.
The demand is more than just in Thailand. In a report called the e-Conomy SEA 2025 that was jointly released by Bain & Company, Google, and Temasek, it projected that the Southeast Asian digital economy will hit more than $300 billion in 2025 for total gross merchandise value, and still keep growing at a pace of about 15% per year.
The bottleneck is power, not capital
Investment is available, but power continues to pose the greatest constraint. Thailand’s BOI has altered the guidelines for investing by substituting an earlier approval process on energy with a newly formed Sub-Committee for Energy Management. As per Baker McKenzie, the committee will judge the project based on the availability of electricity, water consumption involved in the process, and impact on the environment, with the new requirements being also applicable to the major expansion of already operating plants.
This reflects a problem concerning AI infrastructure all over the world. Raising funds is on the rise due to the investors who pursue AI growth, but getting a reliable energy source and grid capacity is becoming a determining factor. For companies like True IDC, succeeding would mean that they not only need to finance the expansion of AI infrastructure but also get necessary energy for the next-generation AI computers.


If you're reading this, you’re already ahead. Stay there with our newsletter.

Quantum Solutions sold 1,000 ETH for around $1.9 million, and lost its status as Japan’s largest corporate holder of Ether. Def consulting is now at the top spot. The money is earmarked for the company’s buildout of an AI data center.
Quantum swaps either for Nvidia GPUs
The disposal was recorded by the Tokyo-listed technology firm via its subsidiary GPT Pals Studio on July 30, according to a company filing. On average, the tokens sold for $1,903 each. It was worth around 311 million yen, or $1.903 million.
The proceeds will go toward the company’s AI Infrastructure Data Center division, including GPU hardware. Target purchases include Nvidia’s B300 and GB300 chips, the physical backbone of the AI business Quantum has been building since June.
With the sale completed, Quantum now holds about 4,764.8 ETH. Def consulting has 4,976 ETH. And that puts Def at the forefront of publicly traded Ether holders in Japan.
GPT Pals had already dumped 904 ETH for ~$1.6 million on June 16. In less than two months, Quantum has sold 1,904 ETH in two sales. That’s around 29% of the 6,668.8 ETH it held before it began selling, and the treasury is down 28.6% since mid-June.
Pledged collateral caps what Quantum can still sell
Quantum built most of its stack during the crypto run of late 2025, when Ether traded between $4,000 and $4,500. The remaining position is ~$9.1 million with ETH trading near $1,906 at the time of writing. The company expects to record a loss of about $100,000, or 17 million yen, on the latest sale. That charge hits earnings for the quarter of its fiscal year that ends in February 2027.
Quantum’s board has extended the runway for more selling. It raised the approved ceiling from 1,875 ETH to 4,375 ETH. That makes another 2,471 ETH clear for disposal through October 30 on paper.
Of the Ether still on its books, 3,050 ETH is pledged as collateral on a prior loan with a Singaporean lender. And there are 1,714.8 ETH in the trading account of GPT Pals. This leaves the freely sellable pile over 750 ETH short of the fresh authorization. Quantum would have to free up or swap out pledged collateral or buy more Ether to get to the full amount.
Future sales depend on Ether prices and how quickly the data center build-out burns through cash. BitMEX co-founder Arthur Hayes also cut his own exposure earlier this year, dumping ~6,000 ETH at a loss as the asset traded sideways. On-chain data at the time showed other large holders still accumulating their positions.

If you're reading this, you’re already ahead. Stay there with our newsletter.

On July 29, 2026, Microsoft informed its shareholders that cloud computing and AI had actually helped the company during its fiscal fourth quarter. The next day, Wall Street reacted by giving Microsoft its biggest rise in a single day since 2008, a 15.5% gain that added roughly $480 billion in market value.
It is the year’s most direct answer to those still questioning whether investments in AI can produce any results reflected in the company’s financial reports. The answer arrived with a wrinkle most of the coverage skipped. Microsoft also cut its spending forecast.
The number the market rewarded
Microsoft put cloud and Artificial Intelligence front and center in announcing its earnings. The company’s July 29, 2026, investor announcement was titled “Microsoft Cloud and AI Strength Fuels Fourth Quarter Results,” which makes it clear what the management thinks is the key to the company’s growth.
The figures behind that title held up. Azure growth accelerated to 43% from 40% the prior quarter, ahead of the roughly 40% Wall Street had modeled, and Satya Nadella used the earnings call to disclose that Azure revenue crossed $100 billion for the full fiscal year for the first time, up 41%, per CNBC.
Quarterly revenue reached $90.01 billion, up 18% and above the $87.62 billion consensus, with adjusted earnings of $4.74 per share against the $4.24 analysts expected. Hood then guided 45% Azure growth for the September quarter, above the 41.4% consensus.
This announcement made a positive impact on investors. Movements like this usually take place due to investors believing that the underlying operations of the company surpassed expectations and that demand is stronger than previously anticipated. The Nasdaq Composite posted its best session since June, and the iShares Semiconductor ETF climbed 8%, erasing a 10% decline over the first three days of the week.
Hood cut the spending forecast, and that was the point
Capital expenditures including finance leases reached $41 billion for the quarter, up 69% year over year but under the $42 billion analysts had feared.
Hood then revised Microsoft’s calendar 2026 capex forecast down to approximately $175 billion from about $190 billion, per the Wall Street Journal via Quartz, achieved in part by lengthening the assumed useful life of the company’s office and data center properties to 25 years from 15. She said Microsoft expects to remain free cash flow positive in fiscal 2027.
The comparison the market made was with Alphabet, which reported in the week of July 20 and lifted its 2026 capital spending guidance to a range of $195 billion to $205 billion, above the $186.4 billion analysts expected. Those shares fell 6%.
Quarterly capital spending of $44.9 billion had outrun operating cash flow and pushed Alphabet’s free cash flow to negative $5.9 billion, its first negative quarter since the 2004 initial public offering. Same earnings season, opposite spending signals, opposite reactions.
Why crypto and AI watchers should care
The results of Microsoft also hold consequences for people who are not related to Microsoft itself. During the past two years, critics have argued that hyperscalers have been investing significantly in AI infrastructure without ascertaining the ability of their investments to bring reasonable returns.
This quarter contradicts this story, though not in the way the headlines suggest. Microsoft attributed its performance to cloud and artificial intelligence, showed accelerating revenue behind that claim, and trimmed the spending forecast at the same time.
The combination indicates that the market is starting to see AI as a profit-making endeavor rather than just a costly venture, and that it will pay for discipline alongside growth. Commercial remaining performance obligations, the contracted revenue Microsoft has yet to recognize, jumped 84% to $678 billion.
The results also set new expectations for the entire industry. Being among the biggest tech firms and having shown leadership in AI infrastructure, Microsoft becomes a point of reference for rivals in the industry.
As a result, cloud providers, AI chip manufacturers, data center operators, as well as any crypto initiatives related to AI and computing will be assessed for their significance against the indicators announced by Microsoft.
The caution the same day would not let go
The growth in the company’s earnings came against a conflicted economic environment. According to Marketwatch, although Microsoft’s results had a positive effect on the stock market, investors in the bond market continued to pay attention to the inflation situation. Investing.com reported that the economy grew at a low pace of 1.5% in the second quarter of the year, while inflation still stayed high.
This contrast deserves to be reiterated. High corporate earnings may spark enthusiasm, but do not erase wider macroeconomic dangers. On July 30, markets struggled with two competing truths: AI demand seemed to be higher than usual, whereas fears of a slowing economy and inflation were still present.
As Cryptopolitan earlier reported, Moody’s has calculated hundreds of billions of dollars in hyperscaler lease commitments that never register as debt on a balance sheet, which is the same category of asset Microsoft has now stretched across a longer depreciation schedule.
Microsoft demonstrated that AI infrastructure can generate substantial cloud revenue, while the rest of the hyperscaler industry is still under pressure to prove that record capital spending will translate into sustainable returns.
The smartest crypto minds already read our newsletter. Want in? Join them.

Blockchain analytics firm Chainalysis said in a report published July 30 that $20 billion was bet on the World Cup through on-chain prediction markets. Of that flow, the firm traced $5.4 million to sanctioned exchanges and other illicit sources.
Betting volume peaked above $300 million on final day
During the tournament, about 400,000 wallets bet on the blockchain. Chainalysis put daily volume at about $50 million during the qualifying stretch in early 2026 and more than $250 million once matches began on June 11. It pulled in more than $300 million the day Spain beat Argentina in the final.
Measured from January onwards, ~$5.7 billion of the total $20 billion was traded during the five weeks of the tournament itself. World Cup questions accounted for ~63% of all prediction market activity during that period.
“During the tournaments, FIFA World Cup markets made up approximately 63% of total prediction market volume,” the report said.
Traders in the US and China were the biggest contributors to the volume. Then came Canada, Thailand, and the United Kingdom. And the winners were the majority, with Chainalysis discovering that 55% of bettors were in profit. About 80% of them have already traded in prediction markets.
One market asked whether Cristiano Ronaldo would shed tears as his final international campaign drew to a close. The “yes” side was right, and that bet alone banked $49 million.
According to Arkham Intelligence data, a trader on-chain known as gud.hl staked $5.2 million on Argentina to win it all, primarily through Polymarket. Spain’s 1-0 victory in the final wiped out the position. It was most of the trader’s profit on the Hyperliquid platform.
Sanctioned wallets hit betting markets and skipped FIFA Collect
FIFA did its own on-chain experiment. The official FIFA Collect platform is built on Avalanche, and fans have used it to buy and sell digital moments from the tournament. These collectibles were also a way to the stadiums.
More than 100,000 match tickets were processed through the platform, with holders either redeeming or reselling assets for entry. Trades in the stablecoin marketplace amounted to $24 million, and FIFA earned at least $6 million in transaction fees.
FIFA Collect mandated ID verification. Chainalysis said less than 0.01% of its users held wallets associated with sanctioned entities. That is a much cleaner profile than the open prediction market.
Chainalysis pointed to illicit flows almost entirely on the betting side. The firm identified ~3,700 wallets associated with illicit actors, less than 1% of participants. Most of the $5.4 million was traced to the sanctioned exchange Huobi, now branded HTX.
The remaining funds were in scam-linked wallets and stolen funds. On a $20 billion base, even a sub-1% share is meaningful exposure for anyone screening the money, Chainalysis noted.
In June, Cryptopolitan covered TRM Labs’ flagging of live scam wallets connected to fake ticketing and rigged-match betting schemes. The FBI issued its own alert about more than 30 fake FIFA sites.
“The World Cup offers a microcosm of crypto’s expanding role in everyday life, and a preview of why the tools to follow the money will need to keep pace,” Chainalysis wrote.

If you're reading this, you’re already ahead. Stay there with our newsletter.

The European Union has initiated the bidding process for as many as seven AI “gigafactories” as part of its procurement program aimed at mobilizing more than 30 billion euros in public and private investments and enabling European firms to tap into the computing power they need to develop cutting-edge AI models.
Brussels wishes to have more of the required hardware for forthcoming AI technologies, instead of relying on the cloud offerings of U.S. companies.
This decision arrives at a time when computing capability has turned into one of the major impediments in this industry. Advanced AI models can only be trained using clusters containing tens of thousands of top-notch accelerators, and officials in Europe understand that the existing public infrastructure has come under heavy strain.
The initiative is part of a larger plan by the Commission called the AI Continent initiative, aimed at achieving a sovereign AI computing capacity.
Furthermore, the initiative shows a trend that has been identified by Stanford University’s 2026 AI Index, revealing that countries around the world are beginning to treat AI infrastructure as a strategic national asset within the same terms as semiconductors, energy, and data.
Why owning the hardware became the whole argument
In the words of a senior official at the European Commission, Europe’s problem can be stated in one sentence:
“The current infrastructure is saturated in demand.”
What this means is that the biggest challenge to AI in Europe today is not the research talent available but rather the lack of computing capacity.
This assertion reinforces Stanford’s conclusion that the development of frontier AI technology tends to favor entities that possess big budget to set up and run huge GPU farms.
According to the AI Gigafactories plan, every site will have a capacity of installing as many as 100,000 cutting-edge AI processors enhanced with ultra-modern networks, cloud software, reliable sources of electricity, and state-of-the-art cooling technology for the purposes of training and utilizing gigantic models.
Europe has already established 19 AI Factories connected to public supercomputers. The objective of these gigafactories is to provide an alternative to those facilities by creating bigger private-led computing plants.
The aim has been to ensure that the development of AI happens within the already established European regulatory and data-governance frameworks. The initiative also corresponds to the EU’s Coordinated Plan on Artificial Intelligence which has been aimed at putting together the investments of different member states since 2018 instead of pursuing fragmented national efforts.
How the €30 billion actually adds up
The funding scheme uses state money in order to attract more investment from the private sector.
In total, the funding is expected to include €10 billion from national and EU sources, which should leverage €20 billion from private investors, with public financing limited to just 35% of project expenses in general.
Some of the funding will still depend upon forthcoming negotiations regarding the EU budget. The European High-Performance Computing Joint Undertaking (EuroHPC JU) will manage the competitive selection process, maintaining the EU strategy of using public funding to attract private investment in critical technologies.
Two lots, and a lot of chips
The tender is divided into two categories. On the one hand, one can back up to four medium-sized gigafactories. On the other hand, the other can support as many as three large facilities. Depending on the project, the amount of funding available from both the EU and national governments can range from €1 billion to €2 billion.
In terms of smaller sites, they should be able to implement anywhere from 25,000 to 75,000 AI processors in an efficient way while the largest one can be able to exceed 100,000. Germany, Italy, Greece, Portugal and Spain are among those countries that are displaying interest in running the major facilities.
A cluster of this magnitude would allow an EU gigafactory to be ranked among the biggest AI training setups from some of the largest U.S. corporations running today, yet it would still fall short of the multi-thousand GPU setups planned by companies like xAI. The analogy shows Brussels’ goal of creating technologies that are able to train cutting-edge AI models, instead of regular research supercomputers.
The dependency Europe cannot design away
Establishing independent AI infrastructure alone cannot reduce Europe’s dependence on foreign manufacturers of semiconductors.
After the EU-U.S. trade agreement reached in July of 2026, the Commission has signed letters of intent with AMD, Nvidia, and Qualcomm to increase access to hardware in relation to the gigafactory projects.
However, officials are quite realistic about this development.
Future hardware upgrades are expected to involve a larger number of suppliers due to increased competition, but chips alone will not solve the problem. The International Energy Agency has also highlighted the fact that AI data centers are contributing towards the increasing global demand for electricity and hence there exist the need to focus on reliable sources of electricity, sufficient grid capacity and cooling systems along with advanced processors.
The clock is now running
Applications close on 12 November 2026. Awards based on evaluations conducted by EuroHPC are expected to be presented in early 2027. The selected projects are expected to begin operations approximately 18 months after signing a contract.
Henna Virkkunen, Executive Vice-President of the European Commission for Tech Sovereignty, Security, and Democracy, declared the launch “a milestone” with regard to the AI aspirations of the bloc.
The EU’s interest in the gigafactories is not limited to just increasing the cloud capacity of the economies, as there is a view that these facilities may provide the much-needed infrastructure for the faster deployment of AI in sectors such as manufacturing, healthcare, transportation, and public services.
However, the analysis by OECD shows that success in the adoption of AI across Europe is not uniform, which means that the €30 billion of additional computing power is only a small piece in the important battle for economic competitiveness.

The smartest crypto minds already read our newsletter. Want in? Join them.

Amazon (NASDAQ: AMZN) reported $62.6 billion in second-quarter net income after a sharp rise in the value of its Anthropic holding added billions to the result. Profit reached $5.75 per diluted share, compared with $18.2 billion, or $1.68 per share, one year earlier.
The 2026 figure included $53.4 billion in pre-tax income outside normal operations, mainly from Anthropic. Amazon shares gained more than 10% after regular trading ended.
Quarterly revenue was up to $200.61 billion, exceeding the $196.47 billion anticipated by analysts polled by LSEG. Net sales increased by 20% compared to $167.7 billion posted in Q2 2025, according to Amazon.
With currency impacting revenue only by $100 million, sales growth was 20% without this effect. Wall Street projected earnings of $1.82 per share, but the sizable capital gain reported by the company makes the comparison less straightforward.
Amazon turns AI demand into faster AWS growth and higher operating profit
The sales figures for Amazon Web Services stood at $42.2 billion, exceeding the expected StreetAccount estimates of $40.54 billion. The revenue from AWS saw an increase of 37% compared to the previous year. Advertising was reported at $19.81 billion, which is greater than the StreetAccount estimate of $19.43 billion and increased by 26%.
Operating profit increased to $27.5 billion, against $19.2 billion before. AWS contributed $16.6 billion, against $10.2 billion previously. North America delivered $9.1 billion, against $7.5 billion, whereas international operations delivered $1.7 billion against $1.5 billion previously.
Amazon President and Chief Executive Officer Andy Jassy said:
“AWS is booming, growing 36.7% year-over-year in Q2, our fastest growth in 18 quarters, and our AI and Chips businesses each eclipsed run rates of more than $25 billion.”
Both operations grew by triple-digit percentages and passed a $25 billion annual revenue pace.
Amazon said Trainium secured multi-year, multi-gigawatt commitments from Anthropic and OpenAI. NEURA Robotics, Odyssey, TwelveLabs, Decart, Poolside, Karakuri, Metagenomi Therapeutics, NetoAI, and Splash Music also adopted or committed to the chips. Larger customers included Uber (NYSE: UBER) and Pinterest (NYSE: PINS).
Graviton5 became widely available during the quarter. Amazon said Graviton can deliver 30% to 40% better price-to-output results than similar systems. Graviton5 supplies up to 25% more computing power than Graviton4. About 98% of the top 1,000 EC2 customers use Graviton. Revenue commitments almost tripled from the previous quarter, and Graviton5 grew nearly twice as fast as its predecessor.
Bedrock added more than 10 managed foundation models. The new choices included OpenAI’s GPT-5.6, Anthropic’s Claude Opus 5, Alphabet’s (NASDAQ: GOOGL) Google DeepMind Gemma 4, and xAI’s Grok 4.3. Hundreds of thousands of customers now use Bedrock. The platform gained more customers during the past six months than it did in its first two years, while second-quarter customer spending exceeded all earlier quarters combined.
Bedrock AgentCore added Payments, Web Search, and Harness. Those tools let agents complete transactions, find current information without leaving AWS, and connect the systems needed to run large deployments.
Amazon Quick received agents that can run multi-step jobs from plain-language instructions. It also gained a personal activity feed and 16 integrations, including Adobe (NASDAQ: ADBE), Moody’s (NYSE: MCO), and Snowflake (NYSE: SNOW).
Quick works with Salesforce (NYSE: CRM), Slack, Microsoft Teams through Microsoft (NASDAQ: MSFT), Atlassian’s Jira (NASDAQ: TEAM), and ServiceNow (NYSE: NOW). It can arrange meetings, prepare and send email, update customer records, and create dashboards while keeping a company’s access rules.
Amazon raises AI spending as retail, logistics, media, and satellites expand
Amazon increased its planned 2026 capital spending to $220 billion. It had forecast $200 billion in February and kept that number in April. Jassy said higher memory prices added to the budget, while customer demand still exceeded available computing power.
“But even at that amount, we will still not have enough capacity to meet all the demand we have in 2026, and I believe this dynamic will also be true in 2027 too,” Jassy said. “In fact, the demand we already have for 2028 is striking.”
Operating cash flow for the latest 12 months rose 33% to $161.4 billion, from $121.1 billion for the period that ended June 30, 2025. Free cash flow went from an $18.2 billion inflow to a $7.6 billion outflow. Purchases of property and equipment, after sales and incentives, increased by $66.1 billion, mainly because of AI construction and hardware.
Amazon committed $1 billion to AWS Forward Deployed Engineering. The unit will place AI engineers directly with customers and build working agent systems in days rather than months. Early users include the Allen Institute, Cox Automotive, the NBA, the NFL, Ricoh (TYO: 7752), and Southwest Airlines (NYSE: LUV).
AWS Secret Cloud for Industry also became generally available. Northrop Grumman (NYSE: NOC) became the first company to run classified work on it. Amazon separately offered up to $1 billion in cloud credits to speed up cloud upgrades across the U.S. Intelligence Community.
Amazon said its global data centers use water more than seven times as efficiently as the wider industry. The company completed 75% of the work needed for its goal of becoming water positive across data center operations by 2030. Its direct operations in India reached that status before the 2027 deadline.
Amazon Business reached $60 billion in annualized gross sales. Its product range grew by almost 30%, and business buyers gained same-day delivery for fresh groceries in more than 2,300 U.S. cities and towns.
The next Proteus warehouse robot can carry goods weighing up to 1,300 pounds. Workers can give it instructions through normal spoken language. Amazon Pharmacy more than doubled its new customers during the first half, while same-day prescription deliveries grew almost fivefold. Automatic manufacturer discounts cut customer spending by nearly $250 million, over 400% more than one year earlier.
Ads Agent expanded into 11 additional countries. Advertisers using it recorded an 8% lower cost per impression and a 6% lower cost per customer acquisition than advertisers who did not use the tool.

The smartest crypto minds already read our newsletter. Want in? Join them.

Coinbase (NASDAQ: COIN) exchange posted $1.2 billion in second-quarter revenue for the period ended June 30, below the $1.3 billion analysts expected in an LSEG (LSE: LSEG) survey. Its loss reached $1.36 per share, far worse than the 17-cent loss Wall Street had penciled in. The company’s stock was immediately sold off following this news, resulting in more than a 7% decline after hours.
The quarter was the company’s third straight miss on both sales and profit forecasts. Coinbase lost $359.5 million, compared with a $1.43 billion profit one year earlier. Earnings went from $5.14 per share last year to a $1.36 loss this time. Revenue also fell from $1.5 billion to $1.2 billion.
Coinbase builds subscription products while weak trading keeps hurting revenue
Bitcoin spent much of the second quarter stuck inside a narrow price band, as the market was healthier than it had been during the first quarter, but spot bitcoin ETFs faced a long run of withdrawals.
Coinbase then saw reduced action in the two largest areas of its operations, as the company earned $599 million from transaction revenue and earned another $555 million through subscriptions and services.
All of these figures fell below estimates and were lower compared to the previous year. The share of subscriptions remained larger compared to the total amount of revenues. For a number of years now, Coinbase has been seeking to reduce its dependency on trading commissions.
Revenue from stablecoins totaled $292 million, a decrease of $17 million compared to the same quarter in 2025. The expectation from StreetAccount was for $327.2 million. The disappointment resulted from management’s continued emphasis on services such as USDC, Coinbase One, Base, and others outside spot trading.
Coinbase CEO Brian Armstrong said Coinbase reached a new record for its share of crypto trading and argued that the company can operate through any market. In the earnings release, he said, “Coinbase is no longer a bet just on the price of bitcoin.” Brian added, “All of financial services are getting updated by crypto, whether that’s trading or payments or lending,” and called Coinbase the best-placed company to supply that infrastructure.
During the earnings call, Brian Jung of Jung Media asked why Coinbase appeared to be reconnecting with retail and crypto-native users after Brian appeared on Market Bubble and Cobie took control of the Base App.
Coinbase’s Brian said, “Yeah, so we have lots of different groups that like to use Coinbase, and build on top of the Base Chain is even probably a broader group. And so, you know, we try to make an effort to connect with all of them. It’s really a pretty diverse group of people that use Coinbase, right? There’s the largest, like, GSIB banks in the world are building on our infrastructure.”
Brian and Alesia detail how Coinbase plans to serve AI agents and grow USDC
Austin Hankwitz of Grit Capital asked whether AI agents would care about Coinbase’s reputation or simply choose the cheapest and fastest network. He noted that more than 90% of agent-driven stablecoin transaction volume settles on Base. Brian said price would matter, but not by itself. “AI agents are probably going to care about a similar set of things that humans would,” he said.
As Brian put it, Base provides settlements for less than a cent within less than a second. Additionally, Brian noted that automated clients will require other factors, such as security, liquidity, legality, reliability, and uptime. He compared that choice with selecting Amazon’s (NASDAQ: AMZN) AWS for cloud infrastructure. “Trust will continue to be important in that world,” he said, adding that Coinbase plans to welcome AI agents as customers.
Eric Pan of Ericnomics asked where Coinbase expects growth now that bitcoin-linked transaction revenue has fallen from more than half of company sales to 12%.
Brian said, “At any given time in trading, there’s always something that’s up and something that’s down. That’s part of the Everything Exchange strategy. You’ve got to have all the shelves stocked so you have the inventory when that thing trends that week. And then on the non-trading fee side with subscription and services, we’ve seen good growth of that over the past years as well.”
Coinbase wants enough products available so customers can trade whatever becomes popular at a given time. He said the wider menu should spread trading-fee income across more assets and products. Subscription and service fees, meanwhile, are meant to make company revenue easier to forecast.
Chief Financial Officer Alesia Haas said paid Coinbase One memberships reached a record during the quarter, even as crypto trading volume fell. “We saw an all-time high in paid Coinbase One subscribers this quarter,” she said. Alesia said those members tend to use more of the platform’s products, giving Coinbase another way to increase customer activity without depending only on trading.
Ken Worthing of JPMorgan Chase (NYSE: JPM) asked whether Coinbase’s deal with Hyperliquid gives large USDC holders too much of the stablecoin’s economics.
Alesia said institutions can hold USDC on Coinbase and earn rewards, while retail users can do the same through Coinbase One. She said Hyperliquid was treated like any other customer, though its role in perpetual futures and market-making made the relationship important.
Alesia said Coinbase was willing to share revenue because placing USDC deeper inside Hyperliquid could increase liquidity, usage, and adoption across the wider network.
Brian said the company would keep funding that push. He said USDC already ranks first for stablecoin transaction volume and first among regulated stablecoins, but remains second to Tether by market value when less-regulated products are included.

Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Apple (NASDAQ: AAPL) brought in $109.42 billion during its fiscal third quarter, yet the stock dropped more than 6% after regular trading ended. Traders focused on the company’s softer forecast for the current period, where sales are expected to rise by only 9% to 11%.
Apple blamed tight supplies for the weaker view. The reaction came even though quarterly revenue beat the $108.65 billion expected by analysts tracked by London Stock Exchange Group (LON: LSEG).
Profit came in ahead of analysts’ expectations as well. Apple posted $2.02 in earnings per share, higher than the expectation of $1.89. This is not a like-for-like comparison, however, since tariff reimbursements boosted EPS by 11 cents.
Apple’s net income reached $29.79 billion, up from $23.43 billion one year earlier. On a per-share basis, last year’s result was $1.57. Apple’s total revenue rose by more than 15% for the third quarter in a row, while cash holdings stood at $146.52 billion.
Apple gets most of its growth from iPhone and Mac sales
The iPhone business produced $54.25 billion, ahead of the $53.86 billion analysts had expected. Sales from the phone line climbed close to 22% from the same quarter last year. That increase came late in the iPhone 17 sales cycle, with another hardware release expected in September.
Mac revenue came in at $10.35 billion, well above the $8.74 billion forecast. That was an annual increase of almost 29%. Apple Chief Executive Officer Tim Cook tied the jump to demand for the MacBook Neo and the MacBook Pro.
The Neo is a lower-priced notebook that uses a chip based on Apple’s iPhone technology. It went on sale in March, shortly before the quarter began.
Apple hiked prices for many Macs during the reporting period. One such model was the MacBook Neo. Lack of memory chips and low manufacturing capacity for processors resulted in increased expenses and reduced availability of components. According to Tim, it was a “tough situation” that needed to be highlighted during the earnings call.
The iPad business had a weaker quarter. Revenue fell 6% from a year ago to $6.19 billion, missing the $6.92 billion estimate. Apple said last year’s comparison was unusually hard because a cheaper iPad had lifted sales during that period.
Wearables revenue reached $7.88 billion, slightly higher than the $7.82 billion analysts expected. Apple also reported that revenue from its wearables and home unit grew 6% to $4.9 billion, which came in above forecasts.
China remained Apple’s third-largest regional business. Revenue from the market rose 22% to $18.82 billion. Apple includes mainland China, Hong Kong, and Taiwan in that regional total.
Apple expands services while supply limits weaken its next-quarter forecast
The services unit generated $30.74 billion, below the $31.22 billion Wall Street estimate. Revenue from the division still increased 12% from the prior year. Apple said currency changes outside the United States reduced the final number.
The division received more money from advertising, the App Store, AppleCare plans, music, streaming video, online storage, and payment products. Tim said Apple now has 1.5 billion paid subscriptions. That figure covers iCloud accounts and subscriptions bought through the App Store, where Apple receives part of each payment.
Gross margin came to 50.1%, while analysts had expected 47.9%. The reported figure was helped by refunds tied to earlier United States tariffs. Those payments added about two percentage points and took the exact margin to a record 50.06%. Without the refund benefit, Apple’s margin would have landed close to the original Wall Street estimate.
The refund came after the U.S. Supreme Court struck down certain tariffs that were imposed by the previous government of President Donald Trump earlier this year. Under these emergency powers, Apple had been paying more import fees for the finished products as well as individual components that it brings to the country.
Apple is also preparing a new Siri system that will use technology from Alphabet’s Google (NASDAQ: GOOGL, GOOG). The voice assistant is expected to arrive with new iPhone hardware in September. Investors have questioned Apple’s position in artificial intelligence as competing companies release new AI products at a faster pace.
“Today, Apple is proud to report our strongest June quarter ever, with double-digit revenue growth across iPhone, Mac and Services, and in every geographic segment,” Tim said. He also pointed to the Siri AI update, new software, and child-protection tools introduced during WWDC26.
Apple Chief Financial Officer Kevan Parekh said, “We are very pleased with our record business performance during the quarter, which set new June quarter records for both EPS and operating cash flow.” Kevan added that Apple’s active-device count reached a new high across every main product group and every region.
The call was Tim’s final earnings appearance before John Ternus takes over as chief executive. John has worked at Apple for 25 years and currently runs the company’s hardware division. He spoke only briefly during the call.

The smartest crypto minds already read our newsletter. Want in? Join them.

Strategy Inc. (Nasdaq: MSTR) reported an $8.22 billion net loss for the second quarter, but Bitcoin remains the number investors watch.
As of July 26, Strategy managed 843,775 BTC, but executives rounded to 846,000 BTC while citing quarterly growth. The coins have a cost basis of $63.69 billion and a market worth of $54.77 billion at Bitcoin’s price of $64,915 on July 27.
The company’s average cost was about $75,476 per coin. Holdings grew 11% during the quarter, while Bitcoin Per Share rose 5%. Year-to-date BTC Yield reached 4.5%, BTC Gain totaled 29,997 coins, and BTC Dollar Gain reached $1.95 billion.
Strategy takes the Bitcoin loss while the software unit keeps producing revenue
Strategy posted an $8.33 billion operating loss, against $14.03 billion in operating income one year earlier. An $8.32 billion unrealized digital-asset loss caused nearly all of that result. The same assets produced a $14.05 billion unrealized gain during the second quarter of 2025.
The amount of loss attributable to common shareholders stood at $8.62 billion after deduction of $400.7 million in preferred dividends. In the previous year, the common shareholders had earned $9.97 billion in profits after paying $49.1 million in preferred dividends. The diluted loss per share was at $24.45, while diluted earnings were at $32.60 in the previous year.
Quarterly revenue climbed 6.9% to $122.4 million from $114.5 million. Gross profit increased to $81.6 million from $78.7 million, while gross margin narrowed to 66.6% from 68.8%.
Cash fell to $1.71 billion on June 30 from $2.21 billion on March 31. Strategy also held $736.1 million in short-term investments, versus none at the end of March.
President and Chief Executive Officer Phong Le said convertible debt dropped 18% to about $6.7 billion, while the dollar reserve rose 12% to $2.4 billion during the quarter.
“Our objective is for STRC to trade over time at $99 to $100,” Phong said. The company plans to make continuous buybacks below $100, with the magnitude determined by price and liquidity. Phong said that discounted purchases reduce future preferred dividend expenses, but ordinary market demand determines the trading level.
Chief Financial Officer Andrew Kang said the reserve later reached $3.75 billion, enough for more than 2.1 years of preferred dividends and interest. Strategy has made 18 consecutive months of dividend payments without a miss. Andrew said, “Our BTC Hurdle ARR of 10.8% represents our current effective cost of credit.” BTC ARR above that rate gives Net BTC Per Share a positive spread.
Founder and Executive Chairman Michael Saylor said weak Bitcoin sentiment has not changed the Digital Credit plan. Michael said Strategy wants STRC near par with steady demand, strong liquidity and low volatility. “Our plan is to return STRC to health,” he said.
Strategy sells securities, pays STRC dividends and buys shares below par
At-the-market sales raised $8.4118 billion during the second quarter. MSTR generated $2.9467 billion, while STRC generated $5.465 billion. STRK, STRF, STRD and STRE generated nothing. Between July 1 and July 26, MSTR brought in another $1.276 billion, including shares sold but unsettled by July 26.
Strategy also bought back $1.5 billion of its 0% convertible notes due 2029 during May for about $1.38 billion in cash. The negotiated deal carried an 8% discount to face value and cut outstanding convertible principal from $8.21 billion to $6.71 billion.
STRC paid $0.92 on January 31 at an 11% rate, based on the January 15 record date. It paid $0.94 on February 28 at 11.25%, based on February 15. Payments of $0.96 followed on March 31, April 30, May 31 and June 30 at 11.5%, using record dates of March 15, April 15, May 15 and June 15.
Semi-monthly payments then began. Strategy paid $0.48 on July 15 at 11.5% for holders recorded June 30. It paid $0.50 on July 31 at 12% for holders recorded July 15, and scheduled another $0.50 for August 15 for holders recorded July 31. The 12% rate will stay until STRC trades near its $100 stated value for a sustained period.
From July 20 through July 26, Strategy repurchased 288,930 STRC shares carrying $28.89 million in stated value for about $25 million. The average price was $86.53, or 13.47% below par. Roughly $975 million remains under the $1 billion Digital Credit buyback program. Strategy plans to buy more at wider discounts and less as STRC nears $100.
While there is an additional $1 billion MSTR repurchase program that has not yet been executed, Strategy indicates that the management can buy back the common stock if they feel that it is undervalued.
The board may authorize up to $1.25 billion in Bitcoin sales to build the dollar reserve. Strategy may also sell coins to pay preferred dividends, cover interest, refill the reserve, or finance MSTR and Digital Credit buybacks. It sold about $218.4 million of Bitcoin during 2026 to fund part of its preferred dividend payments.
The smartest crypto minds already read our newsletter. Want in? Join them.

Perceptron announced a successful funding round of $6.5M from leading Web3 investors. The funding will go toward building AI infrastructure and sourcing AI training data from a Web3 community.
Perceptron closed a strategic funding round, gathering some of the leading Web3 investors and ecosystem partners. The backers for the latest funding include Sigma Capital, Selini Capital, and other trading firms and Web3 investors.
“Our mission is to create the world’s first decentralized AI data mesh where user interactions directly fuel AI growth. We’ve already shown that mission can become a reality, as evidenced by our ability to scale to hundreds of thousands of nodes organically. Now, with this funding, we are launching our data-questing platform, which will allow AI companies to commission specific, high-value datasets directly from our community,” said Peter Anthony, co-founder and CEO of Perceptron.
The latest funding is the biggest and widest for Perceptron. The startup was incubated in the Colloseum ecosystem. In June 2025, Blockmesh merged with Perceptron Network to create the current project.
As Cryptopolitan reported, Perceptron has already worked on its decentralized data network, creating a resource for AI training.
Perceptron to launch data-questing platform
The strategic funding round will support the launch of Perceptron’s data-questing platform, expand tools for contributors and build an incentive and rewards program. The goal for Perceptron is to scale its decentralized network to 5 million nodes.
Perceptron’s goal is to make the sourcing of real-world data fast and accessible, creating a resource for AI companies. Currently, AI trading relies on centralized scraping infrastructure.
On the Perceptron ecosystem, both individuals and AI companies can generate complete, verified datasets, and even order new datasets to be delivered within days.
The decentralized data infrastructure sits between centralized scraping tools, and closed data partnerships. With Perceptron, contributors own their data and share in the earnings, so they can monetize their contribution.
Perceptron offers a new use case for decentralized Web3 networks, but takes the tools a step further. Instead of only organic data, Perceptron enables AI companies to commission data from the community. This means no incentive to merely spam data, but tools to produce valuable, tailored datasets.
“Perceptron has demonstrated an impressive ability to mobilize a decentralized workforce and build a globally distributed network. Their model allows them to tap into niche expertise – from doctors and lawyers to native speakers – on demand, from anywhere in the world,” said Nathan Gurr, Investment Analyst of P2 Ventures.
Early transaction activity is already showing demand for the network. In the first phase of data-questing, live data agents reached mode than 200,000 users from Telegram and Discord communities. The platform reports 300K daily active users, and 807,000 active nodes. The figures show an active audience that may want to monetize its data, but previously lacked convenient infrastructure.
Perceptron prepares for airdrop
Perceptron Network is currently in the pre-airdrop stage, which incentivizes on-chain usage.
The Perceptron team warned that the widely circulated NFT screenshot is a social media fake. The network users could indeed acquire NFT before the Perceptron Network rebrand, a legacy of the Blockmesh project.
Instead, point allocations will be based on contributions and general data activity. The airdrop will depend on points earned, Internet contribution, network participation, and other activities.
As with other potential airdrops, scams and fake links may appear as the TGE date approaches.
PERC tokens will be airdropped to high-trust network members, to reflect real contributions.
The smartest crypto minds already read our newsletter. Want in? Join them.

Binance Research’s new report titled “Half-year 2026: Macro & Bitcoin” has shed more light on how the cryptocurrency performed in the first half of the year. The report showed that Bitcoin declined by 32% year-to-date, and it is currently down by over 50% from its all-time high of $126,080, which it hit in October 2025.
The decline was the world’s largest cryptocurrency by market capitalization’s third straight quarterly loss. Per the report, the loss coincided with a repricing across global markets that is being referred to as “re-anchoring.”
This re-anchoring refers to all the following actions:
A shift away from central-bank support
Consumer spending
Rich valuations toward a tighter Federal Reserve
AI-driven capital expenditure cycle
Earnings-led returns
The report pointed out that Bitcoin absorbed the sharpest hit of any major asset class in that transition, even as its underlying market structure held up better than in prior cycles.
Why is the Fed now Bitcoin’s biggest headwind?
Per Binance Research, the shift in how markets price Federal Reserve policy was one of the biggest drivers of Bitcoin’s slide.
The implied spread between expected and actual Fed funds rates moved from approximately −230 basis points (bp) in August 2024, when deep cuts were priced in, to around +33 bp by mid-2026. Markets are now reportedly assigning roughly 80% odds of a hike by December.
The new Fed Chair Kevin Warsh unsettled markets in his debut press conference by focusing on inflation rather than employment, and this caused short-term Treasury yields to go up. That repricing has moved in near-lockstep, inversely, with Bitcoin’s price over the past year.
According to Binance Research, AI hardware investment was responsible for about 40% of the first quarter GDP growth. The report pointed out that it was the first time since 2009 that it surpassed consumers’ contribution.
Meanwhile, US equities kept climbing through the same period; the S&P 500 rose 18.5% over twelve months.
However, earnings have been driving the advance, as forward price-to-earnings multiples actually compressed from 22x to about 20x.
Binance Research reads current Fed pricing as “too hawkish.”
Japan is also responsible for another pressure point, even though it is relatively less discussed.
The Bank of Japan’s balance sheet has contracted by ¥125.3 trillion, or 16.4%, from its 2024 peak. It is reportedly the largest such contraction in its history. The yen still touched a 40-year low near 162 yen per dollar in June, even after a rate hike to 1%, a level that came despite record central-bank intervention.
Binance Research pointed out that Japan is constructive and not complacent as it enters into the second half of the year. It also stated that rate pricing is hawkish and wrote, “the AI capex cycle points to a slowdown rather than a stall, and earnings have replaced multiples as the engine of returns.”
Has Bitcoin entered the final stage of its correction?
Onchain data suggest the market is deep into capitulation territory. By the end of June, around 10.83 million BTC were held at an unrealized loss compared with 9.22 million BTC still in profit. It is the first time losses have outnumbered profits this cycle.
It was also stated in the report, “Combined with a 50%+ drawdown and 275 days since October 2025 highs, this places BTC within a plausible, though unconfirmed, historical bottoming window into Q4 2026.”
Despite its current state, Bitcoin is still the dominant player in crypto, as it still accounted for 57 to 60% of the market throughout H1. The report stated that this is because it remains the preferred fallback exposure during the sell-off.
In times when the dominance dipped, the report says that the outflows were either going into stablecoins or out of the market.
Altcoins were not in season as there was no sustained rotation into them. Cryptopolitan reported that selling reached new peak levels in June.
How did Bitcoin hold up against other assets?
Bitcoin’s position as a portfolio diversifier was put to the test in the first half of 2026, and it did not perform great.
BTC underperformed every major asset class in H1, falling around 32% while US equity indices closed the half near record highs and gold ended down about 7%.
Bitcoin sold off ahead of equities during bouts of macro stress but then failed to participate when stocks staged their AI-led recovery.
Binance Research says that this is due to Bitcoin’s ETF-era structure, where it trades continuously. This allows it to reprice to shifting rate expectations faster than traditional markets can.
The report concluded that Bitcoin did not offer the perks of a stable hedge in the first half. It behaved more like a liquidity-sensitive macro asset whose diversification value changes ground depending on the market regime.
Why did Bitcoin’s usual buyers turn into sellers?
The demand channels that powered Bitcoin’s prior rallies reversed in H1. US spot Bitcoin ETFs recorded their first-ever year-to-date net outflow; a record $4.5 billion was pulled out in June alone, over three-quarters of it from BlackRock’s IBIT.
Corporate treasury buying, meanwhile, became almost entirely dependent on Strategy, whose enterprise valuation fell below the value of its own Bitcoin holdings for the first time, a threshold that made further share issuance dilutive rather than accretive.
The company disposed of 32 BTC in May, its first sale since 2022, followed by 1,363 BTC in the last days of June, both moves aimed at supporting its reserve and distribution obligations rather than signaling a change in conviction.
Public miners added to the pressure, selling at a record pace as hash price hit an all-time low. That stress has widened the gap between pure-play miners and operators pivoting toward AI and high-performance computing contracts, a shift that is reshaping how miners allocate power, capital, and balance-sheet capacity and one that could reduce their reliance on Bitcoin sales over time, even as it diverts resources away from mining itself.
The report also notes that quantum-computing risk to Bitcoin’s cryptography moved from theoretical research toward concrete migration planning in H1, with draft protocol proposals now circulating, a longer-term diligence item for institutional holders.

3iQ, the Toronto digital asset manager now owned by Japan’s Coincheck, has announced that it will manage a portion of the Bitcoin reserves behind Bhutan’s Gelephu Mindfulness City (GMC).
3iQ will be the special economic zone’s first named manager for the sovereign crypto it plans to build a financial hub on.
Why is 3iQ managing Bhutan’s Bitcoin?
A partnership was announced between 3iQ, which is now owned by Japan’s Coincheck Group (NASDAQ: CNCK), and Bhutan’s Gelephu Mindfulness City (GMC). 3iQ, GMC’s first named asset manager, will be in charge of handling a portion of its Bitcoin holdings.
3iQ has also agreed to train Bhutan’s local staff, transfer expertise, and maintain a permanent office in Gelephu, helping the region develop into a digital offshore financial center.
Pascal St-Jean, 3iQ’s Director and Chief Executive, said the company would put Bhutan’s capital to work “responsibly, transparently and for the long term.”
Jigdrel Singay, a board director at GMC, called the deal “an important first step” in building the city’s digital asset hub. He also revealed that 3iQ’s commitment to building local capabilities was part of the reason for its selection.
The Bitcoin 3iQ will manage comes from a promise the King of Bhutan, King Jigme Khesar Namgyel Wangchuck, made in December 2025 during his National Day address. Cryptopolitan reported that he said up to 10,000 BTC would be used to support the long-term development of GMC.
Those reserves were mined using Bhutan’s hydropower and are watched over by Druk Holding and Investments (DHI), the country’s sovereign investment arm.
3iQ, which was started in Toronto in 2012 and became Canada’s first regulated digital asset fund manager, had Coincheck Group take over about 97% of the firm from Monex Group in a stock deal valuing 3iQ near $112 million. This Bhutan mandate is one of 3iQ’s first sovereign clients under its new ownership.
How much Bitcoin does Bhutan have left?
Cryptopolitan reported in May that data from Arkham Intelligence shows that Bhutan has moved more than $237 million out of its reserves since January, with sovereign holdings sliding from about 13,390 BTC in October 2024 to a much lower estimate of 4,973 BTC, making it the seventh-largest government holder of Bitcoin.
Now in July, Bhutan reportedly holds 5,600 BTC valued at approximately $381 million.
Blockchain analysts believe the proceeds may be funding domestic priorities, including infrastructure projects like the Gelephu Mindfulness City initiative. 3iQ’s new role to manage a portion of the reserves shows that at least some of the Bitcoin within GMC is not going to be sold off.
Notably, GMC launched a fast-track licensing route in May for firms already regulated in hubs like Singapore, Hong Kong, and Abu Dhabi, in an attempt to attract other operators. GMC also offers 0% corporate tax for some sectors and banking through DK Bank.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Aave’s founder Stani Kulechov named Scroll, zkSync and Aptos among the smaller deployments that the project plans to fully retire in part of a sweeping proposal to deprecate another 50 low-adoption asset reserves in a governance push announced on Thursday.
According to the governance forum filing drafted by risk provider LlamaRisk, the two-part plan is supposed to remove about $98.1 million in supplied assets and $15.6 million in outstanding debt from these so-called low-adoption asset reserves.
That total is a combination of:
Individual reserve removals across 11 Aave V3 deployments, covering 50 reserves and another 21 matured Pendle Principal Tokens. Altogether, those come up to $85.3 million in supply and $11.5 million in debt.
Another 25 reserves could be taken offline via six whole-market shutdowns worth $12.8 million in supply and $4.1 million in debt.Contribution of each deployment to the supplied value on Aave. Source: LlamaRisk
Which chains will Aave deprecate whole markets on?
Kulechov confirmed that the governance filing is proposing an “orderly winding down” of entire markets on six chains (Sonic, Scroll, zkSync, Metis, Soneium and Aptos) instead of doing a reserve-by-reserve review.
Chain Supplied Balance Sonic $7.6 million Scroll $2.2 million Aptos $1.7 million zkSync $0.8 million Metis $0.3 million Soneium $0.2 million

DefiLlama data backs up LlamaRisk’s rationale for pushing to completely take the selected networks off Aave. The $1.81 million in total value locked on Scroll and roughly $873,000 on Aptos barely show up on the balance sheet for Aave, which holds $14.47 billion across all chains.
An important bit of context to note is that these proposals are pending a governance vote . Holders still have to approve the wind-downs before any freezes and rate changes take effect.
Why is Aave taking down markets?
Regardless of size, the operating cost on every listed reserve stays at a standard rate. Those costs often include oracle costs, risk parameter monitoring, and maintenance of liquidation paths,
According to LlamaRisk, the cuts are simply portfolio maintenance moves to deployments that no longer made fiscal sense, rather than any one big blowup or response to a catastrophic event.
In short, if a reserve’s activity cannot cover the overhead to maintain it, it’s time to wind it down. And according to LlamaRisk, Aave is now paying more to support the markets on Sonic, Scroll, zkSync, Metis, Soneium and Aptos than the revenue they generate.
The review runs against Aave’s proposed Risk Framework, a four-layer standard that LlamaRisk introduced to govern listings, quarterly due-diligence refreshes, and deprecation decisions across V3, V4 and Aave Horizon.
Aave floated the standard in June after the roughly $292 million KelpDAO bridge exploit, which routed stolen rsETH into Aave markets as collateral and left the protocol facing potential bad debt.
A companion proposal targets shaky price feeds
Running in parallel is a second LlamaRisk proposal aimed at long-tail reserves flagged for Chainlink price-feed risk. Chainlink has tagged the feeds behind these assets as high or very high operational risk because the underlying tokens lost liquidity, leaving too little trading to price them reliably.
That set spans 10 deployments with $6.76 million in supply and $4.29 million in debt. Because live positions remain open, the feeds cannot simply be switched off.
Instead, each reserve gets frozen, its caps cut to one, and its live oracle swapped for a fixed-price adapter, so remaining loans can unwind without a manipulable feed. SCR on Scroll and USDm on Celo appear in both proposals, LlamaRisk noted.
The trims sit against a growing V4
The housekeeping arrives while Aave’s newer plumbing expands. On July 29, the protocol said V4 deposits had nearly doubled over the prior month, and that deposit and borrow caps had been raised for the eleventh time, per its account on X.
The two moves fit Kulechov’s stated “revenue-led protocol strategy,” under which reserves that no longer earn their keep are pruned, and capital is pushed toward institutional and real-world-asset markets.
Aave TVL and revenue. Source: DefiLlama
Aave holds around $116.6 million in annualized revenue and $14.47 billion in TVL, according to DefiLlama, still the largest lending protocol by both measures. AAVE traded at $95.52 on Thursday, down 2.75% over 24 hours.
If you're reading this, you’re already ahead. Stay there with our newsletter.

The founder and chief executive of the Russian Bitcoin mining giant BitRiver, Igor Runets, has been charged with large-scale fraud and remanded in custody.
The crypto entrepreneur, who was detained earlier this year on tax evasion charges and initially placed under house arrest, has been moved to a detention facility.
BitRiver CEO Runets is facing fraud charges in jail
The Russian judiciary is tightening the detention regime for Igor Runets after bringing new charges against him, local media reported.
The prominent crypto figure has been transferred from his home, where he had been held for almost six months, to a pretrial detention facility.
The measure has been approved by the Zamoskvoretsky District Court of Moscow on July 22, the business news outlet RBC revealed Wednesday.
Runets, who was first apprehended for evading taxation, has now been charged with large-scale fraud related to the failed supply of crypto mining hardware.
The equipment was ordered by companies from the En+ Holding, the major Russian energy and metals producer which relies heavily on hydroelectric power generation.
The financial damages inflicted have been estimated at almost 1 billion Russian rubles (more than $12.5 million), the report further unveiled, quoting two knowledgeable sources.
The head of the Russian mining behemoth BitRiver will spend at least two months in pretrial detention while the investigation into the case and his role continues, Bits.media added.
Igor Runets has been charged under Part 4 of Article 159 of the Russian Federation’s Criminal Code, “large-scale fraud committed by an organized group,” the news outlet also detailed.
The crypto executive was moved to a dedicated detention facility on the request of security officials who cited the size of the damage caused and expressed concerns he might exert pressure on witnesses.
The district court in the Russian capital agreed and granted the motion, while his defense team and legal representatives of BitRiver are yet to comment on the decision.
Further developments will now depend on the results of upcoming examinations of the equipment and the testimonies of employees of En+, founded and controlled by Russian oligarch Oleg Deripaska.
Story of Russia’s largest miner turns into criminal case
The Odessa-born Russian crypto entrepreneur Igor Runets established BitRiver in 2017 as a large operator of data processing centers and eco-friendly coin minting enterprise.
Headquartered in Zug, Switzerland, the company’s main operations were based in energy-rich Russian regions such as Siberia, which also offer appropriate climatic conditions for mining.
Things were going well until the now 40-year-old businessman was detained in Moscow at the end of January 2026 and charged with multiple counts of tax fraud, as reported by Cryptopolitan.
Runets was placed under house arrest immediately after facing allegations of concealing assets to evade taxation and was supposed to remain at his residence for the duration of these proceedings.
His detention coincided with the launch of formal bankruptcy monitoring against entities in the BitRiver holding, the leading player in Russia’s crypto mining industry, booming despite bans and restrictions.
These proceedings began amid mounting debt and legal disputes. The petition against BitRiver’s main owner and manager, Fox Group, was filed by Infrastructure of Siberia, an affiliate of Deripaska’s En+.
Fox has accumulated an estimated debt of approximately $9.2 million and has been unable to secure sufficient funds to pay it off. As a result, BitRiver is now at risk of change of ownership as well.
Ukraine war-related sanctions imposed by the U.S. Treasury Department in 2022 added to these troubles and BitRiver reportedly began slashing budgets and scaling back operations in late 2024.
A lawsuit filed by Infrastructure of Siberia in 2025 alleged that BitRiver received payment for mining equipment from the En+ firm, which was its key electricity supplier, but failed to deliver it.
In February 2026, Russia’s Federal Tax Service (FNS) filed for the bankruptcy of a BitRiver subsidiary responsible for a massive 100 MW mining project in the country’s Far East.
The unsuccessful undertaking in the Republic of Buryatia, which imposed seasonal restrictions on mining, is believed to have contributed significantly to the mining giant’s downfall.
If you're reading this, you’re already ahead. Stay there with our newsletter.

Binance.US is gearing up to take on prediction market leaders Kalshi and Polymarket. The American platform is seeking a Commodity Futures Trading Commission (CFTC) license to launch its own prediction market.
During a panel at Rare Evo in Las Vegas, the platform’s CEO, Stephen Gregory, confirmed that they plan to apply for a designated contract market (DCM) license next month.
A DCM classification should open the door for retail clients to trade futures, options, and event contracts legally. Crypto America Host Eleanor Terrett even noted that the planned application is part of the company’s “comeback strategy,” which is built around lower trading fees and diversification beyond spot trading into perpetuals and prediction markets.
Binance.US’s strategy for the future comes after years of regulatory setbacks that have drastically diminished its U.S. market presence. Since Gregory took over as CEO this year and the exchange has turned to the business as a whole, rebuilding its trading fees from lower to increased revenue, product expansion and the organization’s compliance controls.
Since relaunching its growth strategy, Binance.US has introduced near-zero trading fees, announced plans to expand into derivatives, and set a goal to capture even more of the U.S. crypto exchange market. The addition of prediction markets would broaden the product landscape and generate more revenue than just trading on crypto.
Binance.US wants to recover its consumer base and trading volumes
Prediction markets have become one of the fastest-growing segments in financial technology, allowing users to trade contracts tied to the outcomes of real-world events. The sector has attracted growing institutional and retail interest following regulatory developments that have expanded the availability of event contracts in the U.S.
The industry has seen rapid growth following legal victories that expanded the availability of event contracts in the U.S.. This has encouraged more financial firms to explore the sector.
Binance.US is the latest exchange wanting to join the prediction market space. If it secures CFTC approval, it will join established players such as Kalshi and Polymarket. Gemini also recently obtained its license, and Coinbase partnered with Kalshi to offer event contracts in the U.S.
Since 2023, the US entity’s trading volume has cratered. A direct casualty of the broader global platform Binance Holdings’ massive legal battle and settlement with federal agencies. Back then, Binance Holdings admitted to violating major banking acts and thus had to pay about $4.3 billion in penalties. The platform’s former chief, Changpeng Zhao, even spent four months in prison, though he secured a formal pardon from President Donald Trump.
Currently, Zhao still holds a majority stake in Binance.US. In the past few months after his release, he’s been actively championing growth avenues that could revitalize the platform’s domestic footprint. Nevertheless, he insists that his role in the company is solely technical.
In one of his statements, he commented, “We want to do much more business in the US. We want to bring a superior product into the US. We want to make the superior product offering much more accessible to the US consumer.”
Around the same time, a platform spokesperson, speaking on his comments, stated that their core mission is to become the leading U.S. venue for buying, trading, and earning crypto assets, as they focus on new products and changing to meet user needs.
That said, entering the prediction market space could possibly give the platform the boost it needs.
Robinhood may be considering a deal with Crypto.com
Reportedly, Robinhood Markets is also exploring a deal with Crypto.com to list the exchange’s prediction market contracts on its app. Still, there is no guarantee that the negotiations will result in an agreement, according to the Wall Street Journal. If made, the partnership will allow retail traders to access Crypto.com’s binary markets directly from their primary Robinhood accounts.
At the moment, Robinhood sources its event contracts from Kalshi, Interactive Brokers’ ForecastEx, and Rothera. However, the company’s share of Kalshi’s trading volume has been steadily declining since Rothera’s addition. Overall, trading activity on the platform had exceeded 16 billion event contracts this year, compared with more than 12 billion throughout 2025, a notable surge. Even Kalshi’s CEO, Tarek Mansour, acknowledged the company’s potential.
This June, he said he considered Robinhood a key rival, despite their partnership. Nonetheless, Kalshi still heavily dominates the U.S. market. The firm saw World Cup betting handles reach an incredible $27 billion, completely eclipsing the $1 billion wagered on the Super Bowl. Combined trading metrics showed that sports categories pulled in 80% of customer order flow across both Kalshi and Polymarket during the cup.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Situational Awareness, a hedge fund specializing in artificial intelligence that was established by Leopold Aschenbrenner, a previous employee of OpenAI, is seeking to get additional funding from investors as semiconductor and AI stocks plummeted. The timing is important because this hedge fund has the biggest disclosed bet against the companies causing the drop in the stock prices of semiconductor companies.
Therefore, the act of raising money makes it more than a normal venture. It also gives us a glimpse into the perspective of one of AI’s prominent investors towards the future development of the sector. Financial Times reported that instead of investing heavily in shares of chip-producing companies, the company has developed an investment strategy that allows it to gain profits even when chip manufacturers are losing money.
A short book aimed at the AI chip trade
Situational Awareness LP revealed a U.S. equity portfolio valued at $5.52 billion in its Form 13F filed with the U.S. Securities and Exchange Commission by March 31, 2026. The report showed that the fund was mainly linked to put options valued at $8.7 billion, attributed to companies in the semiconductor and AI sector. This includes $2.04 billion in VanEck Semiconductor ETF and $1.57 billion in Nvidia stocks. Investors appeared to have put more than $1 billion in put options of companies such as Oracle, Broadcom, and Advanced Micro Devices.
However, the long portfolio tells a different tale. Rather than taking a position in chip makers, the fund opted to invest in firms connected to AI infrastructure. Bloom Energy was the biggest reported position in the long portfolio with a value of $879 million, while the second biggest position was with SanDisk and CoreWeave. The fund also invested in several bitcoin miners including IREN, Core Scientific, Riot, and CleanSpark.
The fund has increased some of these investments in mining companies in the quarter which implies that the fund believes that power generation and data center infrastructure will benefit from the spread of AI in the future, instead of just relying on the prospects of semiconductor companies.
What a single session cost the fund
An unidentified X account, @LeopoldTracker_, posted an unverified estimate on July 28 about Aschenbrenner’s portfolio being down approximately $600 million in one trading session due primarily to the rapid falls of Bloom Energy and SanDisk. Cryptopolitan could not independently confirm this number.
Leopold Aschenbrenner’s fund has lost roughly $600,000,000 in a single session today
But he’s still up over 1,000% on most of these positions from his entry and his puts are printing:
• Bloom Energy, $BE, -$210,000,000 (-14.9%)
• SanDisk, $SNDK, -$180,000,000 (-11.0%)
• Core… pic.twitter.com/apL862fyLH
— Leopold Stock Tracker (@LeopoldTracker_) July 28, 2026
The estimate comes with several qualifications. Filings made under Form 13F only include some U.S. based equity assets and options for the quarter until the end of the quarter. Other financial commodities are not included, such as cash, derivatives, short positions, foreign securities, and various private investments, making them an incomplete picture of a hedge fund’s overall exposure or day-to-day gains and losses.
According to the same tracker, the fact that the fund’s sizable put positions have appreciated due to a drop in the value of chip companies may potentially offset the losses related to its long portfolio. Even if it is not possible to determine whether this is the fund’s true results from public disclosures alone, it shows that the fund’s hedged approach is entirely different from the strategy simply betting on AI stocks only.
OpenAI alum’s billion-dollar bet
Aschenbrenner first received extensive coverage in June 2024 when he released his 165-page essay, Situational Awareness: The Decade Ahead. In that essay, he claimed that artificial general intelligence would likely occur in 2027 and that this would lead to an “intelligence explosion” resulting in increasing expenditure on GPUs, data centers, and electricity as well as heightening the competition between the United States and China in terms of technology.
He launched Situational Awareness in September 2024 around that investment thesis. The firm’s website describes it as a global equity investor focused on AI as “the dominant driver of global market returns over the next decade.”
The proposal was well-received by investors, and by October 2025, Fortune stated that the fund had over $1.5 billion in assets under management, which is a remarkable success to achieve for a first-time fund manager in his twenties who worked for OpenAI and Sam Bankman-Fried’s charity. Supporters of the venture called him an early adopter in the field of AI infrastructure with all the critics highlighting the fact that the strategy may be too reliant on general interest in the sector.
Why the raise is a market signal
The positioning of the fund showcases its opinion which is becoming critical as investments in AI keep on changing. While Aschenbrenner holds publicly optimistic views regarding AI’s future, he has also heavily short-sold several semiconductor businesses that aided the surge while championing power suppliers, infrastructure companies, and bitcoin miners.
If he manages to bring in new investors at this point in the semiconductor downturn, this apparently demonstrates the conviction that there is more to come with the correction in chip stocks despite the investment in AI infrastructure remaining on an upswing.
In a wider context, the portfolio points to a challenge investors could be increasingly up against: belief in the rise of AI does not imply that each part of AI’s chain will be performing well at the same time. These shifts could mean that investment would go from buying GPUs to investing in power generation, networking and data center capacity, leading to the expansion of the leadership group in the AI sector beyond chip-making companies. That would reward investors who can identify which parts of the AI ecosystem are positioned to capture value at each stage of the industry’s buildout.
The smartest crypto minds already read our newsletter. Want in? Join them.

gumi Inc., a Japanese gaming and blockchain company, has joined with a financial group, SBI, to launch SBI Crypto Fund I. It will begin operating on August 1, 2026.
The fund is aiming for 3 billion yen, which makes up $18.3 million. It will employ methods such as staking, hedging, and portfolio rebalancing to manage the money.
SBI Financial Services holds 51% of the stake in the fund, while Gumi’s unit gC Labs owns the remaining 49%. Moreover, Daiwa Securities Group and Yamada Securities Group are among the investors. The fund only puts money into Bitcoin and other major altcoins that trade on recognized exchanges.
Gumi has been active in digital assets since at least 2018. As of April 30, 2026, its total crypto holdings were worth around 14 billion yen, or about $86 million. The company wants to become the largest XRP treasury company in Japan.
SBI already owns about 34% of gumi through a capital and business partnership formed in 2022. Gumi also created a unit called “Neo Crypto” to bring all its digital asset work under one roof. The division is being built to gain operating experience and get ready for the possible approval of crypto ETFs in Japan.
Gumi bets on Japan’s crypto reforms
Japan’s parliament approved new laws that reclassify digital currencies as financial instruments rather than just payment tools. The legislation amends the Financial Instruments and Exchange Act and the Payment Services Act. The changes are set to take effect in 2027, as reported by Cryptopolitan previously.
The new rules open the door for future spot bitcoin ETFs, though no such products have been approved. The Financial Services Agency said it will now work on building a proper framework for crypto ETFs.
【お知らせ】
当社子会社の株式会社gC Labsを通じて、SBIファイナンシャルサービシーズ株式会社と共同で組成した暗号資産運用ファンド「SBI Crypto Fund Ⅰ」の運営を2026年8月1日より開始することを決定いたしましたので、お知らせいたします。https://t.co/dXHOQgIJy8
— gumi公式 (@gumi_pr) July 28, 2026
The law also brings stiffer penalties. Anyone running an unregistered crypto operation can now face up to 10 years in prison, compared to three years before. The maximum fine has gone up from 3 million yen, about $18,500, to 10 million yen. Stricter insider-trading rules are included, and both crypto issuers and exchanges will have to share more information with investors and the public.
Lawmakers also backed a plan to cut the crypto tax rate from as high as 55% down to 20%, though that change will not take effect until 2028. Under the new system, 15% of the tax goes to the national government and 5% goes to local governments.
Businesses and users drive account growth
SBI VC Trade, the crypto exchange arm of Tokyo-based SBI Holdings, said its registered accounts recently passed 2 million. That is about double the 1 million accounts it had in 2025. The company says more businesses are moving money into Bitcoin and XRP to reduce their exposure to a weak yen. Some of those companies also give out Bitcoin or XRP as part of shareholder benefit programs.
The account total covers both its VCTRADE and BITPOINT platforms, which came together after SBI VC Trade merged with BitPoint Japan in April 2026. The two brands are expected to fully combine around the end of December, which the company said should lower costs and bring service levels in line across both.
Stablecoins have been another factor behind the growth. SBI VC Trade listed USDC in March 2025, calling it Japan’s first dollar-backed stablecoin listing. In June 2026, it added Ripple’s RLUSD alongside JPYSC, a yen-pegged token it described as the country’s first trust-based yen stablecoin. It also began offering loans against stablecoin holdings.
Japan’s crypto market is still smaller than those in the United States and South Korea, partly because of its strict rules. But rising account numbers and growing corporate interest in digital asset strategies suggest the market is gaining real momentum.
The smartest crypto minds already read our newsletter. Want in? Join them.

An unreleased Anthropic AI model has broken a post-quantum signature scheme in about 60 hours, reopening a question Bitcoin developers have yet to answer: when should the network begin migrating to quantum-resistant cryptography?
The finding poses no immediate threat to Bitcoin. HAWK-256 was never used by the network, and no existing wallets are affected. But it changes an important assumption behind Bitcoin’s post-quantum roadmap: AI is making it faster and cheaper to test, and potentially break, advanced cryptographic systems.
What Mythos did to HAWK, and what it cost
On July 28, Anthropic announced that its Claude Mythos Preview model, which is being made available to a select group of approved users, discovered an attack on HAWK-256, the last lattice-based signature candidate still in NIST’s third round of post-quantum cryptography evaluation, which has not previously been documented.
The AI revealed a concealed symmetry in the design of HAWK, which experts overlooked, causing the algorithm’s effectiveness in terms of safety to decline by fifty percent as well as bringing down the anticipated amount of work needed from around 264 to 238 operations. Based on the research blog by Anthropic, the attack took approximately 60 hours and required about 100,000 dollars worth of computation, and it was carried out by a researcher without formal training in cryptography.
The reaction from the cryptography community was quick. Sophie Schmieg from Google said, “Basically with this paper, HAWK is dead.” Matthew Green from Johns Hopkins University pointed out that while the research had no groundbreaking mathematics, it was a new combination of existing ones. Ars Technica reported that the creators of HAWK withdrew the algorithm from the public the next day.
Why every Bitcoin wallet is untouched
The result does not affect Bitcoin.
Bitcoin employs ECDSA on the secp256k1 elliptic curve, which has nothing to do with HAWK. The algorithm has not been implemented in Bitcoin, any blockchain, or commercial software. Its failure happened during the review process intended to reveal flaws before use.
The anticipated migration of Bitcoin has already circumvented HAWK. The Bitcoin Improvement Proposal repository received BIP-360 in February 2026, which introduces a new output type known as P2QRH that uses ML-DSA and SLH-DSA, two algorithms finalized by NIST in 2024 after many years of public scrutiny. Moreover, BIP-361 was submitted in April 2026 by Jameson Lopp and others, a proposal to phase out ECDSA in Bitcoin. HAWK was not included in BIP-360 or BIP-361.
The test that decides when migration starts
As discussed by TFTC, the core issue is whether AI-assisted cryptanalysis can eventually decrypt the algorithms that Bitcoin hopes to use, namely ML-DSA or SLH-DSA. If either of these algorithms proved vulnerable before the network completed its migration, then the strategy of waiting for mature standards may no longer be appropriate.
The economics are also changing. Anthropic has indicated that meaningful cryptanalysis can now be performed in about 60 hours for about $100,000, making this type of testing much more repeatable. Intelligence agencies like the NSA or GCHQ almost certainly are doing similar work behind the scenes so that public findings are likely to lag behind those done in private.
Researchers have drawn parallels between this case study and AGU, a different post-quantum contender, that prevailed through years of investigation, only to be ultimately attacked within 1 hour on a laptop in 2022. AI appears to be accelerating that pattern.
Why the timeline keeps compressing
Cryptopolitan has reported in the past that AI-supported cryptanalysis is reducing anticipations as to when quantum attacks might become efficient. Forrester’s 2026 quantum security report released makes the same claim and warns that the arrival of new algorithms compresses security timeframes “overnight” by minimizing computing resources necessary for solving complex encryption challenges. The company forecasts worrying security threats associated with Q-Day by 2030.
Google has similarly expedited its efforts, setting the goal of retiring RSA and elliptic-curve cryptography from its systems by 2029 and expressing the desire for other companies to follow suit.
For Bitcoin, the situation is critical, as estimated by analysts cited by TheStreet, who determine that there are approximately 6.7 million Bitcoins worth nearly $600 billion sitting in wallets that could potentially come under threat from a sufficiently advanced quantum computer. Additionally, a study called Quantum Horizon states that around 2.3 million Bitcoins would be lost forever if quantum attacks happen, while it estimates a one-in-six chance of a quantum computer capable of breaking cryptography being operational by 2035. Hence, the authors argue that Bitcoin’s main problem is not hardware anymore, but rather if the network could pull off a successful transition before that technology becomes available.
What cryptography does Bitcoin use?
Bitcoin uses several different cryptographic primitives, each serving a different purpose. Importantly, Bitcoin does not encrypt transactions. Instead, it relies on cryptography for authentication, integrity, and consensus. Here’s a breakdown:
Function Cryptography Used Purpose Quantum Vulnerable? Digital signatures ECDSA (legacy) and Schnorr signatures (Taproot) over the secp256k1 elliptic curve Prove ownership of bitcoins Yes (Shor’s algorithm) Hashing (Proof-of-Work) SHA-256 Mining and block hashing Much less affected (Grover’s algorithm provides only a quadratic speedup) Address generation RIPEMD-160 + SHA-256 Create Bitcoin addresses from public keys Not directly vulnerable like signatures Merkle trees SHA-256 Verify transaction inclusion in blocks Much less affected

The concern is that Bitcoin’s signature schemes (ECDSA and Schnorr) are vulnerable to a future quantum computer running Shor’s algorithm. Developers and researchers are therefore exploring how Bitcoin could transition to post-quantum signature schemes, such as hash-based or lattice-based alternatives, through future protocol upgrades. Any such migration would require broad agreement across the Bitcoin ecosystem due to its impact on wallets, transactions, and consensus rules.
The HAWK attack did not affect Bitcoin directly, because Bitcoin does not use HAWK. However, the result illustrates how AI can accelerate cryptanalysis and reduce the time needed to evaluate or break emerging cryptographic schemes. Combined with recent advances in quantum algorithms, these developments reinforce the urgency behind Bitcoin developers’ ongoing efforts to prepare for a post-quantum transition.

The smartest crypto minds already read our newsletter. Want in? Join them.