Cryptopolitan

Meta released Muse Spark 1.3 on September 2 with two pricing tiers.
Developers who let Meta train on their prompts and outputs pay as little as one-twenty-first of the standard rate.
Muse Spark’s cached tokens drop 75x
A million input tokens run for $1.25, and a million output tokens run for $4.25 on the standard plan. The Contributor tier reduces those to 10 cents and 20 cents, a savings averaging about 95%.
Standard and Contributor tier pricing from Meta’s own Model API pricing page, viewed September 3, 2026.
The biggest discount is on cached input tokens, which allows developers to reuse context without paying for reprocessing. Those fall from $0.15 to $0.002 per million, a 75x cut that’s close to free for repeated workflows to run.
In return, developers surrender control of their data. All prompts, responses, and usage patterns on the Contributor tier are included in Meta’s training pipeline.
The standard plan lets users make 3,000 requests per minute, but Contributor access only lets users make 100 RPM. This is too few requests for production-scale functionality.
Meta paused internal employee tracking program
Meta Superintelligence Labs, the division behind the Muse Spark line, built these models to fill a specific hole.
Facebook, Instagram, and WhatsApp yield a torrent of conversational and visual data. However, they yield very little of the quality coding interactions that agentic tools need to improve.
Mario Zechner, the developer behind the open-source harness Pi, said the increase in coding agent capability from April to October 2025 was mainly because Claude Code stored user sessions by default and fed them into reinforcement learning training.
Meta has had a hard time sourcing the material itself. A program inside the company that tracked how employees used their computers came under heavy critique and was put on hold in June.
Arvind Narayanan, a computer science professor at Princeton, pointed out that big firms stick with token-billed enterprise plans even when consumer subscriptions like Claude Max and ChatGPT Pro cost 10 to 20 times less, because the enterprise plans keep their data out of training runs.
Narayanan suggested the Meta offer could make those companies more cautious in differentiating between proprietary data and data they’d happily swap for a discount.
“It lowers the barrier to entry for prototyping, testing integrations, and scaling experiments where training on your data is acceptable,” Meta’s pricing guide says about the Contributor tier.
Meta experimented with the two-tier structure in Muse Spark 1.2 in August 2026. Muse Spark 1.3 features a 1 million-token context window, which can accommodate ~750,000 words in a single session, and claims to necessitate fewer tool calls to complete complex tasks compared to its predecessors.
Anthropic brought down the cost of cached tokens for its new Fable and Mythos models, while at the end of July, OpenAI lowered prices for all of its latest models.
OpenAI gets training data from its free ChatGPT tier users, while Anthropic has stricter rules about data.
The smartest crypto minds already read our newsletter. Want in? Join them.

Apple (NASDAQ: AAPL) has been accused of holding third-party developers to stricter standards than Apple applied to itself. The accusation was made in a £2 billion ($2.7 billion) collective action filed Thursday at London’s Competition Appeal Tribunal.
UK app developers say Apple’s App Tracking Transparency rules are unfair because developers rely on advertising to fund a free app, and this rule puts a price on five years of consent prompts that quietly diverts ad money to Apple.
The double-consent complaint that forms the crux of the case
The complaint was made by ATT Collective Action Limited, and it relies on one argument, which is the fact that Apple forced third-party developers to comply with a two-step consent rule before they could track users across other apps and websites. Meanwhile, Apple’s advertising and data collection were exempted from such a requirement.
Third-party apps were required to obtain a user’s permission twice, while Apple’s in-house services were not saddled with such a restriction. Developers who rely on ad revenue eventually spend more to reach new users while experiencing weaker ad value.
App Tracking Transparency started in April 2021. It displays a prompt immediately someone opens an app, and asks if advertisers can track them. If the user refuses, the developer is prevented from using the person’s data for ad targeting.
Who is behind the claim, and what she says it’s about
Ann Pope is leading the charge against Apple. She is an erstwhile director for antitrust at the UK’s Competition and Markets Authority; this is the same regulator whose tribunal will decide the case, a notable twist to the lawsuit.
Pope believes the fight is not about privacy but rather about fairness. “Privacy is an important protection for consumers, but it should be applied fairly and in a way that ensures businesses of all sizes can compete on a level playing field,” she said.
She went on to add that Apple’s rule “resulted in very significant harm to businesses that depend on Apple as a gatekeeper.”
Apple has not issued a response and has remained consistent in its viewpoint, as it says its own apps are exempt from the tracking prompt because they don’t collect the data the prompt refers to.
This, in Apple’s view, makes the playing field even.
Fines and orders that piled up before London
The UK filing comes amid years of regulatory issues for Apple in Europe. Italy fined Apple €98.6 million over ATT in December 2025 while demanding changes. France fined Apple €150 million in April 2025 without demanding changes.
Germany has also moved against Apple in recent times, as its competition authority found that the ATT prompts were designed to favor Apple over competitors by producing more consent outcomes for Apple and demanded changes to the design.
Apple agreed to eight changes to its EU policies. Poland and Romania have also started reviews.
Where this sits among the tribunal’s Big Tech cases
The London tribunal has become famous as a venue for lawsuits against tech giants. It is currently deliberating over a separate $4.1 billion iCloud claim against Apple, with a hearing set for late 2028, a timeline that suggests the ATT case could take a while.
How the claimants arrived at the $2.7 billion figure is still not known, and it is not a guarantee that Apple will be fined that exact amount.
If you're reading this, you’re already ahead. Stay there with our newsletter.

Hargreaves Lansdown, the UK’s largest direct-to-consumer investment platform, has begun offering crypto exchange-traded notes to everyday clients.
The new offering gives millions of retail investors regulated exposure to Bitcoin and Ether, nearly a year after the ban on retail access was lifted.
How do crypto ETNs work?
Hargreaves Lansdown, which holds over a third of the UK market, now offers crypto ETNs to its customers. Crypto ETNs track the price of Bitcoin or Ethereum, but investors don’t actually own the coins, and there are no wallets or private keys to handle.
Instead, the notes are issued by a bank or financial firm that physically holds the cryptocurrencies in custody.
This system will allow millions of everyday investors to invest in Bitcoin and Ether through a regulated product, almost a year after the ban was lifted.
The Boring Money Market Monitor Q3 2025 report says that HL holds a 33.6% share of the UK direct-to-consumer market by assets under administration, more than double the 17.0% held by second-placed Interactive Investor. It also runs the country’s most-used self-invested personal pension.
HL is charging a 0.35% annual account fee to hold crypto ETNs in a Fund and Share Account or a SIPP, capped at £12.50 per month. Dealing charges run between £3.95 and £6.95 per trade.
The products are inside the firm’s Advanced Investing hub, and it bluntly warns that buyers should be prepared to lose everything they invest. Adding that, if the note’s issuer goes bankrupt, the money can vanish.
HL suggests such holdings make up only a small slice of an already diversified portfolio.
Revolut’s guidance also makes it clear that the Financial Services Compensation Scheme does not cover crypto ETNs, and that the products carry the full volatility of the assets they track.
Why were retail investors banned from buying crypto?
The ban keeping retail investors away from crypto ETNs was implemented in 2021. However, the Financial Conduct Authority (FCA) changed its approach in June 2025 and formally allowed retail access to certain crypto ETNs listed on recognized UK exchanges from 8 October 2025.
After the change, companies like 21Shares, Bitwise, WisdomTree, and BlackRock listed Bitcoin and Ether ETNs on the London Stock Exchange.
The catch, however, is in how these products will be taxed. HM Revenue & Customs confirmed that from 8 October 2025, crypto ETNs would initially qualify inside Stocks and Shares ISAs, the account most UK savers actually use.
But from April 6 this year, the HMRC reclassified the products as qualifying investments only within the Innovative Finance ISA (IFISA). Investors who already held the notes in a Stocks and Shares ISA did not have to sell.
Trade body CryptoUK has argued that the reclassification goes against the FCA’s work to open up access. It noted that around 15 million Britons subscribed to an ISA in the 2023–24 tax year, while the IFISA is used by well under 1% of ISA holders.
If you're reading this, you’re already ahead. Stay there with our newsletter.

The CEO of Saudi Arabia’s HUMAIN has said that they will be launching a large venture capital fund, and it could top a planned $10 billion vehicle.
Tareq Amin said this in an interview published Wednesday, as the Public Investment Fund (PIF)-owned company presses its bid to turn the kingdom into a hub for global AI compute.
It is said that HUMAIN is eyeing a $2.5 billion fund aimed specifically at data centers, a figure it noted sits at the smaller end of announced Gulf AI commitments.
No passive checks, and a domestic arm called HUMAIN Limitless
Amin said, “We don’t do passive investments, that will never happen.” The fund will back companies only if they agree to run some compute on Saudi data centers or move staff into the kingdom.
They are planning for offices in the United States, Saudi Arabia, France, and the United Kingdom, and the CEO said that the spread is meant to speed up the pace of deals.
Alongside the global vehicle, HUMAIN intends to put in place a separate arm, HUMAIN Limitless, to fund and support AI companies inside Saudi Arabia.
The xAI wager that emboldened the plan
HUMAIN’s appetite has grown on the back of one deal. Its CEO stated that the $3 billion investment they made in Elon Musk’s xAI in February was “a home run.” That stake was converted into SpaceX equity after xAI merged with the rocket company.
Saudi Arabia’s sovereign wealth fund, which controls HUMAIN together with Aramco, has since disclosed a $26 billion holding in SpaceX.
According to Amin, that earlier investment is proof that the strategy works.
A buildout financed on bankable offtake deals
Since HUMAIN was founded by the PIF, the AI company has moved pretty fast, as it flagged six gigawatts of AI data centers in the kingdom by 2034. HUMAIN also secured a 16-gigawatt power commitment from the Saudi Ministry of Energy and struck deals with Nvidia, Groq, Qualcomm, and xAI.
The spending is being funded partly through debt. Amin said global hyperscaler demand let HUMAIN borrow against offtake agreements with large technology firms, calling those contracts “bankable.”
A $3 billion partnership with Blackstone is helping cover data center deals. Amin stated that computing in Saudi Arabia is about 30% cheaper than elsewhere. He said this is why demand for it has been steady despite the conflict in the region.
Al Moammar Information Systems reportedly expanded a HUMAIN data center contract to 250 megawatts (MW). That deal is now valued at more than 8.76 billion riyals, or about $2.34 billion.
HUMAIN recently paired with DataVolt on the first 100 MW phase of a 360 MW facility in Neom’s Oxagon zone, slated for 2028. It is estimated that Saudi AI and cloud expansion could need up to $42 billion by 2030.
Chips secured, data centers hardened, a line drawn at Chinese models
Access to advanced silicon is very important in all these workings. The US approved its first exports of advanced AI chips to HUMAIN in November 2025.
Amin said a US-Saudi strategic AI partnership gives the company the chip capacity it needs. “I don’t see any obstacles or issues for us to obtain the chip capacity we need,” he said.
Based on the export controls imposed on China, Saudi Arabia, which is an ally of President Trump’s US administration, is ensuring that they stay compliant with the control. So HUMAIN will not let Chinese frontier labs train models on its compute.
It will only be offering inference access to Chinese open-source models, with Amin stating that the split keeps the company within the US framework because it is “not furnishing a training cluster to Chinese frontier labs.”
Security has climbed the agenda too. With AI infrastructure elsewhere in the Gulf hit during the US-Iran war, Amin said HUMAIN is hardening and distributing its sites, treating them, in his words, “like critical national infrastructure.”
The smartest crypto minds already read our newsletter. Want in? Join them.

Police have arrested more people in Poland as part of the probe into the collapse of a leading coin trading platform in the country and the region, Zondacrypto.
Among the recently detained is a prominent Polish stock market trader. The latest arrests come amid growing allegations of fraud, money laundering and political ties.
New suspects arrested in the Zonda case, Zaorski is in jail
Polish law enforcement has detained another three persons within the criminal investigation into the spectacular crash of Poland’s largest cryptocurrency exchange.
The arrests were made on Wednesday at the request of prosecutors who also ordered the seizure of luxury vehicles, valuables, and a significant amount of cash, local media revealed.
Officers from the provincial police headquarters in the Southern city of Katowice, capital of the Silesian Voivodeship, detained two individuals with the initials J.W. and A.P.
Representatives of the Central Bureau of Combating Cybercrime apprehended R.Z., the Prosecutor General’s Office announced, quoted by the business news portal Money.pl.
Famous stock market investor Rafał Zaorski was brought to the Katowice branch of the National Public Prosecutor’s Office in the afternoon of August 2, the PAP news agency detailed.
“I am a victim,” he told reporters while being escorted into the building, without commenting beyond the short statement, the Polskie Radio (PR) national broadcaster remarked in a report.
Zaorski has been previously linked to Sylwester Suszek, the still missing founder of BitBay, which was the predecessor of Zondacrypto.
Polish media identified the other two handcuffed as Jaromira W. and Anna P. The two women are also connected to the high-profile crypto case.
Jaromira is the former wife of Marian W., also known as “Maniek,” who is the main suspect in the disappearance of Suszek. Anna was a close associate of his, according to Onet.pl.
What happened with Zondacrypto?
Zonda was the biggest digital currency exchange in the Polish market before it collapsed earlier this year, amid liquidity issues resulting in halted withdrawals.
The crypto trading platform was launched as BitBay in Poland, in 2014, by Sylwester Suszek, who also served as its CEO. It was later sold, allegedly to a notorious Russian mafia group.
Later, the company rebranded to Zondacrypto and moved its headquarters to Estonia, from where it operated under a license obtained by BB Trade, an entity registered in the Baltic state.
It became one of the largest crypto trading venues in Eastern Europe under Przemysław Kral, who took over its management in early 2021.
Suszek disappeared shortly after and the new chief executive has been accusing him of never handing over the keys to a wallet with 4,500 BTC.
Kral, himself, went missing after denying the exchange was in trouble in mid-April of this year. He was thought to have fled to Israel, of which he is a citizen, and then Dubai.
According to Polish media reports, he eventually agreed to help authorities unravel the case, which looks like a scam that may have affected more than 30,000 Poles who have lost at least $95 million.
Marian W., the man who introduced Kral to Suszek, is thought to have been the one who was actually pulling the strings of the company. He is also reportedly hiding in the UAE.
Who is Rafał Zaorski and how he relates to Zonda?
Speaking at a press conference, Poland’s Minister of Justice Waldemar Żurek unveiled that the arrested individuals may possess information about the disappearance of Sylwester Suszek.
This applies in particular to Rafał Zaorski, who is believed to be connected to the missing crypto entrepreneur and his digital assets trading company.
Zaorski is a well-known Polish stock market speculator who gained his popularity by publically sharing the results of his investments, including multi-million-dollar profits and losses.
In April 2026, he was convicted of using insider information while trading shares of the Merlin Group, a Warsaw-based e-commerce operator and retailer.
The investigation into Zondacrypto’s collapse was opened on April 17 by the District Prosecutor’s Office in Katowice, which cited suspicions of large-scale fraud and money laundering.
However, the case has grown into a major political scandal as well, with Zonda’s management accused of sponsoring conservative political events and figures to ensure favorable crypto regulation.
The “Crypto-Asset Market Act” proposed by the centrist, liberal government of Prime Minister Donald Tusk has been vetoed twice by President Karol Nawrocki. It is returning to parliament this week.
Previous arrests in the Zondacrypto case include that of the head of the Polish Olympic Committee (PKOl), Radosław Piesiewicz, as reported by Cryptopolitan at the end of August.
He found himself in custody for accepting an expensive gift from the crypto firm’s boss, in the form of a €40,000 watch in exchange for providing favorable treatment in Zonda’s sponsorship deals.
The smartest crypto minds already read our newsletter. Want in? Join them.

Nvidia (NASDAQ: NVDA) has agreed to buy Hugging Face, the open-source platform where more than 18 million developers share and test AI models, for $12,930,300,000.
The company’s CEO Jensen Huang announced the deal on its blog on September 3, 2026. Nvidia, which is the world’s largest AI chipmaker, will now control the most-used neutral hub for open models.
Why did Nvidia buy Hugging Face?
Reports about Nvidia’s deal with Hugging Face first surfaced on August 26 this year. The deal was priced at roughly $12.9 billion, and Cryptopolitan noted at the time that neither company directly confirmed it.
Nvidia CEO Jensen Huang’s recent blog post confirms that the company intensified its interest in Hugging Face after other buyers, allegedly including Microsoft, became interested.
Hugging Face, built by Clément Delangue, Julien Chaumond and Thomas Wolf, is the closest thing open-source AI has to a town square. Huang’s post put the numbers at more than 3 million models, 500,000 datasets and 1 million applications, hosted for over 200,000 companies.
Nvidia is already the single largest contributor of open models and data to the site, having posted more than 500 models and over 250 open datasets.
Huang has said the plan for the acquisition surrounds keeping Hugging Face open rather than folding it into Nvidia’s stack. He wrote that developers will still pick their own models, frameworks, clouds and chips, and “NVIDIA compute will not be required to build on or deploy through Hugging Face.”
The deal has raised concerns of neutrality, but in his blog post, Huang referenced an open letter he recently coauthored, arguing that open weights spread AI leadership across companies and countries.
Cryptopolitan previously stated that Hugging Face gets its value from supporting “models and hardware from across the industry,” which includes Nvidia’s rivals, AMD and Intel.
Why is Nvidia’s acquisition of Hugging Face concerning?
Owning the Hugging Face platform gives Nvidia a direct line to millions of developers, and that is an asset the company could use to its unfair advantage. The platform even turned down a $500 million Nvidia investment at a $7 billion valuation last year, unwilling to let one investor hold too much sway.
The company’s valuation shot up to its current $12.9 billion price following an incident in which an unreleased OpenAI model autonomously escaped testing and infiltrated Hugging Face’s platform, thrusting it into the spotlight.
Nvidia has the money to spare, reporting $96.2 billion in fiscal second-quarter revenue, up 106% from a year earlier. $89 billion of that money came from data centers. The company reportedly had $18 billion designated for equity investments through the rest of the fiscal year, on top of $47.9 billion already held in private companies.
Like Nvidia, other companies are attempting to take control of what analysts call the “AI middle layer.” For instance, Stripe recently confirmed its $7.5 billion acquisition of OpenRouter.
The smartest crypto minds already read our newsletter. Want in? Join them.

Echo Base, the special-situations firm that sent a $10 million pre-negotiated bankruptcy offer to the Bitmart exchange’s management, has followed up on its August moves with a call to customers who can’t access their crypto on the exchange to join its ad hoc committee.
The group that Echo Base is assembling is being pitched as a way for affected BitMart users to form a coordinated legal front after the exchange continues to fail to honor customer withdrawal attempts and ignore the formal rescue offer.
Echo Base presents itself as a privately funded platform that buys into and stabilizes distressed digital-asset companies. BitMart landed on the firm’s radar after it announced in July that it would shut down its crypto exchange on January 31, 2027, per Cryptopolitan’s earlier reporting.
What is Echo Base offering Bitmart customers?
First of all, the committee of affected Bitmart users that Echo Base is assembling will be represented by Young Conaway Stargatt & Taylor and Ashbury Legal, two retained law firms that will also weigh recovery options.
The committee claims that it now represents a “significant and growing” pool of users impacted by the Bitmart closure news. However, the press release did not put a dollar figure on how large the pool is.
Echo Base is moving forward on the basis that customer assets don’t become exchange property according to BitMart’s own User Agreement.
Another legal route on the table is whether qualifying creditors could force an involuntary insolvency case. The decision has not been made on moving forward with that option yet, per Echo Base.
Echo Base asked affected users to reach out to the bitmart@eb.global address.
Bitmart has not responded to Echo Base yet
Echo Base made the first move on BitMart on August 6 when it sent a written offer pledging as much as $10 million to bankroll a pre-negotiated bankruptcy that would cover professional and administrative costs through plan confirmation.
The offer also included debtor-in-possession financing and equity once the exchange emerged from restructuring, underwritten by Echo Base in its capacity as a claimholder.
In a second approach on August 8, an Echo Base affiliate sent a formal demand over a withdrawal request that BitMart had left unexecuted since July 24, roughly 31 hours before the exchange announced it was shutting down.
In that demand, Echo Base logged 15 attempts to reach BitMart, which did not turn up any legal or contractual explanation for the incomplete withdrawal attempts.
However, as of the September 2 call to the exchange’s users, Echo Base said BitMart had not replied to any of its approaches.
Echo Base and Bitmart aren’t working on the same timeline
Roshan Dharia, Echo Base’s chief executive, framed the standoff as a race against BitMart’s own timeline. “Administering a book this size takes years, and BitMart has publicly committed to a process it cannot staff past January,” he said in the statement.
He argued that without a court-supervised process, there is no automatic stay, meaning “a single claimant can stall the process for everyone.”
Dharia said the offer has sat open since August 6 and warned that the workable options shrink each week it goes unanswered. Echo Base added that it remains willing to negotiate a consensual wind-down with BitMart’s management and advisers, and reserved all legal rights in the meantime.
BitMart said on July 26 it would begin an orderly wind-down, halting new accounts, deposits and fresh orders the same day, with spot and futures trading due to end August 26.
Founder Sheldon Xia denied on August 8 that the exchange had run off with funds, telling users in a Chinese-language post that his team was still tallying assets, Cryptopolitan reported. He provided no figures, no dates and no proof-of-reserves report.
The numbers that are public do not reassure. CoinMarketCap data cited by Cryptopolitan showed BitMart self-reporting only about $5.36 million in reserves, most of it in its own BMX token, against daily trading volume near $272.6 million.
The exchange has promised a proof-of-reserves report since May 23 and has yet to publish one. It later floated a “potential restructuring and business resumption plan,” but the shutdown timeline still stands.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Unitree and AgiBot are among more than 900 Chinese companies coming to Berlin for IFA 2026, which opens Friday with the first walk on a catwalk by humanoid robots in the show’s history.
China built 84.7% of the world’s humanoids last year
The show takes place from September 4 to 8 at Messe Berlin under the banner “The Future Is Now.” There will be over 1,900 brands from 49 countries, and over 220,000 people are expected to visit across 27 halls.
“Robots on the Runway” is scheduled for September 5 at 12:45 p.m. on the Creator Stage. Bipedal machines will strike poses, dance, and show off sensor-based navigation and motor control.
IFA Berlin said its innovation zone, IFA Next, has grown to about 300 exhibitors this year.
Unitree, AgiBot, EngineAI, and DEEP Robotics will appear alongside Astrall Dynamics and AiMOGA.
A separate RoboCup showcase will pit robot soccer teams against each other, hosted by Humboldt University’s Team Berlin United.
Chinese participation has swept IFA for years, but 2026 is a record. Of the companies exhibiting, 932 are Chinese, almost half the total and about 200 more than a year ago, by one published count.
That includes appliance behemoths such as TCL, Hisense, and Xiaomi.
“When design, engineering, production and testing are all happening within the same ecosystem, ideas naturally move from prototype to product much faster,” Leif Lindner, CEO of IFA Management, said. He pointed out the manufacturing scale, heavy investment, and fully integrated supply chain.
According to Cryptopolitan, China built 84.7% of the ~17,000 humanoid robots shipped worldwide in 2025. Domestic sales are on track to double to 28,000 units this year.
TÜV Rheinland signed off on Chery’s Mornine for EU sale
Lindner mentioned NEURA Robotics, a German company that closed a funding round of up to $1.4 billion.
Right after the runway show finishes, it will give a main talk called “From Europe, for the World: Building the Ecosystem for Physical AI.”
AiMOGA’s Mornine, produced by a unit of Chinese carmaker Chery, has full EU CE certification for machinery safety, radio equipment, and cybersecurity, signed off by German testing house TÜV Rheinland. That makes it the first humanoid to obtain the go-ahead for both hardware and software under EU rules, and it is already working in car dealerships in more than 30 countries.
Lindner called the certification “the transition from a technology demonstration to a product that’s ready for commercial deployment in Europe.”
According to Cryptopolitan, Unitree has set an August 10 subscription date for its Shanghai IPO, at a valuation of almost 42 billion yuan, or about $5.7 billion. That would make the Hangzhou firm the first major humanoid producer to list on the Chinese mainland.
Unitree’s revenue was $250 million in 2025 and adjusted profit $87 million, with net profit up 674% year over year.
Its G1 robot has been chosen by Nvidia as the cornerstone for the Isaac GR00T research platform.
The smartest crypto minds already read our newsletter. Want in? Join them.

Attorney General Todd Blanche has said that the Justice Department is going after the cyber criminals who bombarded X users with unrequested password-reset emails this week.
The platform maintains that no accounts were compromised because the attack was disrupted in time.
Password reset attack on X
On September 2, 2026, Attorney General Todd Blanche wrote on X that “hundreds of thousands of X users” were affected by a coordinated attempt to hijack accounts through the password-recovery flow. He credited the company with stopping the attack in time and preventing user accounts from being compromised.
Blanche added that investigators are “working closely with X to track down the criminals.”
The attack first came to light on September 1, when users began posting about receiving waves of reset messages. Some inboxes reportedly collected about ten emails in a short window around 9:30 a.m. Eastern.
The messages appeared to genuinely come from the company as they were sent from the official email (info@x.com) and carried the six-digit code needed to finish a reset. Despite this, X engineer Mridul Singhai said the company found no sign of a breach. He also apologized for the volume of emails.
Singhai linked the attack to X Money, Elon Musk’s payment product that opened to the public in July, suggesting attackers “believe that, now that @XMoney is widely available, they can gain unauthorized access to accounts.”
Has X been hacked before?
X’s data has been loose for years following an incident in April 2025 in which a self-described data enthusiast using the handle “ThinkingOne” posted a 34GB file with 201,186,753 X records, including names, email addresses, usernames and follower counts.
The vulnerability exploited by ThinkingOne came from a 2022 bug-bounty report that let attackers search for users using their email or phone number.
Due to that history, there is speculation that X’s most recent attackers used a technique called credential stuffing, where automated tools test stolen username-password pairs against a login system.
Credential stuffing reportedly accounts for 31% of social media hacks, with more than 24 billion stolen pairs in circulation. Researchers who found one X-focused botnet watched it test 722,763 credentials in a 12-minute stretch.
U.S. authorities have been cracking down on cybercrime and have increased their focus on hacking. In late August, the DOJ and FBI announced court-authorized seizures of two hacking platforms, QScan and QTRouter, that a China state-sponsored group used against targets including NASA, the Federal Reserve and the U.S. Senate.
Days later, on September 2, authorities working with CrowdStrike and the Shadowserver Foundation dismantled Sality, a Russia-based botnet that had reportedly infected more than 11 million devices over a 23-year run.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

A senior Google executive has reportedly confirmed that the firm has found a workaround to the AI-led memory chip supply crunch and price hikes that have stalled the entire tech sector: recycling DDR4 memory from retired servers and using it in its newest AI machines.
Senior director of supply chain infrastructure, Nikhil Cherian, said that even with the DDR4 recycling pivot, the tech giant is still struggling to find enough chips for its AI systems. The only lasting solution to the extremely tight supply situation, as Cherian put it, is for memory makers to build more capacity and fast.
Firms are now reusing old memory chips
Google is not alone in seeking solutions to avoid AI accelerators that cost billions of dollars to build and sit idle while they haggle over per-module pricing for memory bandwidth that is hard to come by.
Meta laid out its own method of recycling DDR4 behind Compute Express Link, or CXL, a standard that pools older DDR4 and newer DDR5 in a machine, in a paper it published earlier this year.
Marvell’s associate vice president of product marketing, Khurram Malik, told EE Times that hyperscalers are finding ways to prolong the shelf life of DDR4 modules using the CXL controllers instead of them going obsolete when servers migrate to DDR5-only platforms.
Meta said it counted as much as a 25% efficiency in server count on some inference workloads across millions of servers. Even average latency dropped about 29% on distributed cache systems.
However, the workaround did not work for every memory chip. Meta said off-the-shelf CXL products bundled controllers with DRAM, which blocked reuse of existing DDR4 stockpiles. Those stacks also caused expansion memory to run roughly ten times slower on bandwidth and about 60% higher on latency than directly attached DRAM.
Meta’s answer to that problem was its in-house Vistara ASIC. That specific model is built to be reused, with power efficiency and low latency advantages already built in. On top of all that, the firm also pairs the ASIC with software that monitors workload and switches it off where unacceptable delay is detected.
Memory chip prices are through the roof
Memory chip prices continue to go up, with the arc getting steeper as the AI buildout continues to soak up everything manufacturers can put out. Counterpoint Research reported an 80% to 90% quarter-over-quarter price hike in Q1 of 2026.
Marvell’s estimation came in at 90% to 95% single-quarter increments on the price of conventional DRAM, with hyperscalers expected to spend roughly 30% on memory in 2026, up from about 8% in 2023 and 2024.
TrendForce expects another 13% to 18% quarterly rise in conventional DRAM contract prices in the third quarter of 2026.
No fast relief in sight
Supply is not catching up soon. As Cryptopolitan reported, China’s ChangXin Memory Technologies has hit a capacity ceiling near 240,000 wafers a month, held back by US export controls and yields Counterpoint estimates run 42% below Samsung and SK Hynix. SK Hynix has committed about $38 billion to two new fabs, but its Y2 DRAM plant is not expected to reach the cleanroom stage until mid-2029.
The pressure could get worse before more silicon arrives. Two unions representing about 10,000 Micron workers in Taiwan are weighing a strike vote in September over how bonuses are calculated, Cryptopolitan reported, and Micron is one of three firms that control roughly 94% of the DRAM market. A stoppage at its Taiwan fabs, which cannot be relocated quickly, would tighten a market SK Hynix chief Kwak Noh-Jung expects to stay short until the end of 2030.
The smartest crypto minds already read our newsletter. Want in? Join them.

The crypto market recovery in August happened on the back of $1B of stablecoins flowing into Binance. The short-term recovery was enough to lift BTC by around 22%, also creating the best month for crypto in 2026.
Stablecoin flows in August reached $1B net on Binance, becoming one of the major factors for the overall BTC recovery. Stablecoin flows are also closely watched for signals that the bearish trend is reversing in 2026.
Based on recent analysis by user @darkfrost, August’s liquidity recovery was still relatively small. Even the addition of $1B was only a signal, rather than a significant addition to liquidity on Binance.
Binance stablecoin inflows arrived in spikes, coinciding with some of the most active days of BTC expansion in August. | Source: Cryptoquant
The recent inflows have not yet compensated for the general loss of liquidity, where @darkfrost noted around $5.1B in stablecoin left Binance since the beginning of 2026.
As Cryptopolitan reported, stablecoins are gaining more mainstream acceptance. However, Binance still carries 71% of stablecoin flows, and the major effect on the market is linked to real demand from traders and a readiness to buy the dip.
Stablecoins reverse the general outflow trend
As of September 2026, stablecoins have not yet shown signs of rapid inflows, which would coincide with hype and an active bull market. Despite this, analysts noted the end of a prolonged period of outflows, turning liquidity to a neutral balance.
According to analyst Axel Adler Jr., exchanges see neutral stablecoin flows. For the past 30 days, stablecoin flows were slightly positive for the first time since May. For now, there are no signs of sustained stablecoin inflows, and liquidity is still allocated cautiously.
Despite the stablecoin inflows, the BTC stablecoin ratio shows the currently available liquidity may not easily boost BTC to a new price range. | Source: Cryptoquant
The question remains whether those small stablecoin inflows could push the BTC market to a new range, once again reclaiming the $80,000 level and potentially launching a new bull market.
Following the August rally, stablecoins may not be enough to boost BTC prices. Based on the stablecoin supply ratio, BTC will not be moved easily at the current level of liquidity.
Despite the overall high supply of stablecoins, the market still relies on specific liquidity allocation to sway prices. The recent behavior of stablecoins on the market is also pointing toward slightly cooling demand. The BTC stablecoin oscillator shows demand peaked on August 21, entering a cooldown period since then.
The sudden inflow of stablecoins shows the market is ready to react to liquidity allocation, but demand remains weak, with limited allocation.
Stablecoin volumes recover in August
Overall, on-chain stablecoin flows also recovered in August. For the past 30 days, stablecoin transactions rose by over 31%, according to Artemis data.
Average daily transactions increased by over 11% in the past month, suggesting a recovery in decentralized usage. The more active stablecoin supply followed a recent rise in DeFi lending, with an additional $7B added to the value of lending protocols.
The overall rise in ETH and BTC prices boosted collaterals, leading to increased stablecoin demand and on-chain transfers.
The stablecoin market sustained a supply of $304.6B in August. USDT remains the leading asset, while USDC increased its supply by 2.5%, mostly driven by new mints on Solana.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Following its worst monthly losing streak in more than ten years, Google is hoping to turn the page, kicking off September with a fresh wave of investor optimism. It started the month with fresh product launches and a major victory in federal court.
Specifically, the company just introduced Gemini 3.8 Flash, its third Flash release in just six weeks. Alongside it, the firm launched Gemini 3.8 Flash Cyber, a new security-focused model engineered for verified corporate and public-sector networks. Trusted defenders can now access the tool through the new Fairwind Program.
The firm also defeated a government antitrust play: a federal judge dismissed the Justice Department’s demand that the company divest its ad exchange. The decision represents the third time in recent history that U.S. antitrust authorities have pursued a structural breakup of a major tech firm and lost in court.
These simultaneous milestones pivot the firm into a much stronger position following a bruising summer slump. During that difficult period, Google battled high-profile staff exits, executed a sweeping internal reorganization of DeepMind, and fell behind in the AI arms race.
Gemini 3.8 flash surpasses the previous model in multi-step processing, Google says
Investors seem encouraged by Google’s steady rollout of new AI models, with the latest 3.8 Flash release adding to the momentum. Though Google’s stock isn’t in a full-blown rally yet, it only bounced 0.6% on Wednesday, paring a portion of Tuesday’s 1% slide.
Nonetheless, the slight equity recovery reflects a familiar market dynamic: companies that continue to deliver AI products tend to win back Wall Street’s confidence, even after a difficult period.
Google is especially enthusiastic about Gemini 3.8 Flash. It claims this is their most advanced programming and logic engine yet, offering significant upgrades over Gemini 3.7 Flash for multi-step processing and engineering tasks. Additionally, it highlighted that the model provides a high level of reliability essential for automated corporate workflows across highly specialized industries.
For its security variant, the Gemini 3.8 Flash Cyber, the firm also asserted that the system outperforms both the 3.5 Flash Cyber and much larger flagship models at autonomously uncovering software bugs.
For the past two years, Alphabet has been on the defensive against skepticism regarding its ability to match the cadence of OpenAI and Microsoft.
This pressure has intensified as both competitors aggressively deployed corporate AI software and consumer applications powered by their proprietary architectures.
With the new flash variations, the company may have quashed some of those doubts. However, Gil Luria, an analyst at D.A. Davidson, thinks the new developments are still not enough to beat the top companies.
“From a product perspective, this model seems to keep Google in the race, but probably won’t change the fact that they are a distant third in the enterprise market,” said Luria.
Judge Brinkema decided it would be better for Google to make behavioral changes
Meanwhile, Federal Judge Leonie Brinkema rejected the DOJ’s push to dismantle Google’s profitable ad exchange, ruling against the government’s assertion that a structural breakup was the only viable path to correct Google’s market dominance.
The DOJ had contended that Google was unfit to operate the online advertising exchange, citing Brinkema’s ruling that the company had illegally weakened competition. The company, however, countered that a forced sale would be extremely challenging and would subject customers to a long, disruptive transition.
Rather than ordering structural changes, Brinkema opted for behavioral remedies that could require Google to share data and make its technology compatible with competing services. The specifics will be determined later. A year earlier, Judge Amit Mehta reached a similar conclusion in a case over Google’s search engine.
Although he ruled that Google held an illegal monopoly, he stopped short of ordering the sale of Chrome and Android, as the DOJ had sought. He ordered Google to share more search data and barred the company from entering exclusive distribution agreements, such as its roughly $20 billion-a-year deal with Apple.
Google still faces pressure to prove its AI spending can deliver
While there has been a lot of product momentum and legal victory in recent weeks, Google still has one big problem to address: convince investors that its massive AI investments can translate into sustained revenue growth.
Alphabet has invested billions of dollars in data centers, chips, and other hardware to train and deploy ever-better AI models. Some of that spending has raised questions as to whether returns from AI will be realized quickly enough to justify the cost.
These Gemini releases could help address some of those concerns if businesses begin adopting the models at scale. Google has an advantage with its existing cloud, search, and advertising businesses, giving it multiple channels through which to monetize AI.
But competition is still intense; Microsoft-backed OpenAI and other large technology companies continue to release ever more advanced models.
For investors, the key question is no longer simply whether Google can produce competitive AI models. It is whether the company can turn that technological progress into stronger enterprise demand, higher cloud revenue, and long-term growth while controlling its enormous AI-related spending.
As of 2020, Ad Manager contributed 4.1% of Google’s overall revenue and 1.5% of operating profit based on Wedbush research and analysis of court filings.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

In a comment letter dated August 24, Ondo Finance has informed the SEC and CFTC that new regulations are not required to allow trade in perpetual contracts based on US stocks. The letter argues that the current framework of security futures accommodates the product. The document attempts to turn a matter of definition into a test of the readiness of the US regulators for incorporating the always-on derivatives market of crypto within their structure.
The stakes are already apparent abroad. Perpetual futures linked to stocks listed in the US have been traded out of reach of American investor protections and market surveillance. Analysis of equity-perpetual markets by Messari indicates that billions of dollars have already been transacted on newer platforms, and the inability of the US to regulate the market is the reason behind its development in overseas locations. One of the companies that is developing this market abroad is Ondo and it seeks the same solution to be made available in the US as well.
What is Ondo asking the agencies to do?
The comment, submitted jointly by Ondo Finance and broker-dealer affiliate Oasis Pro Markets, responds to the agencies’ June request for input under File No. S7-2026-21 on how “swap” and “security-based swap” should be defined.
The argument presented by Ondo is limited yet significant. According to Ondo, a cash-settled perpetual contract that is linked to a single stock can already be classified as a security future in accordance with the Commodity Futures Modernization Act of 2000. It establishes a dual regulatory framework by creating the SEC-CFTC joint regime and a notice-registration path, allowing an entity registered with either agency to easily register with the other without creating a separate entity.
The legal matter being examined comes down to expiration. Ondo claims that the statute defines a security future by the underlying asset as well as how it settles, not by the presence of a set maturity date. Viewed through this lens, a perpetual works like a traditional futures contract, but without an expiration date; funding payments and mark-to-market settlements replace the calendar. The larger point Ondo makes is that regulatory authorities do not need to wait for a one-size-fits-all approach before taking action, as the current framework allows them to review individual filings.
A live offshore platform makes the case
Ondo is not making inflatable statements. Its subsidiary, Ondo Global Panama Inc., is providing a platform for non-US customers that offers trading of US-listed stocks using stablecoin-settled perpetual futures contracts. Ondo indicated that the amount of trading on its platform over six weeks since its launch had exceeded $8 billion by August 14. Additionally, Onno Finance states that it has about $4 billion worth of tokenized assets in its ledger and is also the leader in the market of tokenized equities.
In a blog post on September 2, Ondo connected three comment letters concerning product definitions, portfolio margining and data reporting to the same idea: technology has evolved, and so should regulators’ approach towards meeting their policy objectives. The company pointed out that there is “something backwards about perps on U.S.-listed stocks trading entirely offshore” and pushed both regulatory bodies to prioritize making transactions happen in the U.S.
Why the timing matters now
Ondo is not the only one taking action in this area. The Blockchain Association, which claims to represent over 100 members, has also released an August 24 letter calling on regulators to adopt the current security-futures framework to be applied in equity perpetuals. It cautioned that, as time passes, it may be harder to attract liquidity back as user habits and market depth harden around foreign venues. Both letters highlight the SEC-CFTC harmonization effort started in March 2026.
The opportunity has increased since that time. According to Cryptopolitan, the CFTC greenlit Kalshi’s BTCPERP on May 29 and characterized it as futures contracts while saying that additional perpetual derivatives would be considered on a case-by-case basis. Michael Selig, the head of the CFTC, also claimed that the decision opened a way for crypto perpetuals to function legally in the US.
Since then, Kalshi has filed for perpetual contracts associated with a US stock market index and copper. On September 2, it was reported that the company is making preparations to file for perpetuals connected to West Texas Intermediate (WTI) crude oil. Former SEC and CFTC officials, in a public letter sponsored by Kalshi. It mentioned that poor regulations keep pushing trading offshore, referring to Kalshi’s estimates of perpetuals traded overseas in 2025 at over $90 trillion.
The road continues to be unclear. CME Group filed a lawsuit against the CFTC in June, claiming that perpetual contracts ought to be classified as swaps instead of futures. The resolution of that argument could determine the future of onshoring. The next sign will be whether the SEC and CFTC staff show openness to product-specific filings and how they work out their contradictory views on whether the instruments are classified as swaps or futures.
Security futures have a specific regulatory architecture involving both the SEC and CFTC, plus exchange, clearing, margin, surveillance, position-limit and trading-halt requirements. The CFTC confirms that single-stock futures were authorized by the Commodity Futures Modernization Act of 2000 and are subject to joint SEC-CFTC oversight. So, the debate isn’t about whether “Are stock perps legal?” but increasingly about whether individual-stock perps qualify as security futures.
Regulatory question Position Can a no-expiry contract be a future? Ondo/Hyperliquid: Yes Does funding substitute economically for expiration? Ondo: Yes Does that automatically make it a security feature? Disputed Could some stock perps instead be security-based swaps? FalconX: Yes Does classification solve exchange/clearing/margin issues? No Could existing U.S. rules accommodate them? Potentially, but implementation is unresolved


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

Robinhood Chain is seeing parabolic growth over the past week. It is currently the leading fee-generating chain in crypto and only second to Solana when it comes to chain revenue. On September 2, the network collected $4.45 million in chain fees and kept $4.01 million of that as revenue, according to data from DefiLlama. This makes it the fourth consecutive day where we’ve seen fees and revenue grow at a rapid rate. Just six days ago, chain fees were at $200,211 and revenue stood at $179,815. That’s around a 22x increase in under a week.
Source: DefiLlama
What Chain Fees and Chain Revenue Actually Measure
Chain fees refers to the total amount of gas users paid to transact on the network. Whatever Robinhood keeps after covering its own costs, which includes posting transaction data back to Ethereum and the 10% of net protocol revenue it owes Arbitrum for running an Orbit chain, is the Chain revenue. 8% of the protocol revenue goes to the Arbitrum DAO while the remaining 2% to the Developer Guild. The cut is charged on profit and not on gross throughput.
The gap between the two lines is thin. On August 27 it came to $20,396, which implies Ethereum costs of roughly $400 that day. The Arbitrum payment was effectively the entire spread. On September 1, the Arbitrum ecosystem’s take from Robinhood Chain was around $370,000. Arbitrum One itself did under $15,000 in fees the same day.
Where the Fee Growth is Coming From
Pons is a token launchpad on the Robinhood Chain that launched 12 days after its mainnet went live on July 1. Fast forward to the end of August and the platform took 63.9% of every dollar in launchpad fees across crypto.
Pons V2 launched August 3. It charges a launch fee, then collects swap fees twice over, once on the bonding curve and again in the Uniswap v4 pool that graduating tokens land in. 80% of the V1 protocol cut goes toward buying and burning PONS. The team said on August 29 that 29% of the original supply had been retired. The token is up over 300% on the week and printed an all-time high near $0.50 alongside the fee record.
Source: CoinGecko
Uniswap Is Getting Paid on Every Graduation
Every Pons graduation ends up in a Uniswap pool. The v4 and v3 deployments on Robinhood Chain pulled $2.68 million and $1.45 million in fees over 24 hours and together account for 81% of the chain’s DEX volume. Robinhood Chain now handles 51% of all Uniswap v4 volume across every network Uniswap runs on.
Tokenized Assets Are the Smallest Number on the Chain
TVL sits at $783.09 million, with $868.28 million in stablecoins and $2.707 billion bridged in. 24 hour DEX volume is $1.405 billion, with another $304.64 million in perps. RWA active market cap is $193.25 million. A network built to host tokenized equities is currently operating as a memecoin venue.
The Gas Subsidy Runs Out on September 29
Chain revenue is not Robinhood revenue. CFO Shiv Verma told analysts on the Q2 call that the company earns a few basis points per transaction with roughly half shared with Arbitrum, without disclosing a rate or any reconciliation.
All of this also happened while gas was being subsidized. Robinhood cut the gas sponsorship threshold in its Wallet from $5 to $0.50, and the program runs until September 29, the end of the original 90-day window. Whether these numbers hold once users start paying their own gas is the question worth watching.
The smartest crypto minds already read our newsletter. Want in? Join them.

On September 1, Anza initiated Transaction V1 on the testnet of Solana, allowing for the final rehearsal of this upgrade prior to its launch on September 9.
The change is intended to allow cryptographic and DeFi workloads that had initially to be fragmented in different transactions to now be processed in one single atomic transaction, as stated by Solana Compass.
Jobs that used to be split now fit in one call
The most notable alteration is the increase in space. The maximum size of a Solana serialized transaction has increased from 1,232 bytes to 4,096 bytes which is an increase of 3.3 times and took effect at the testnet epoch 1025.
According to the website dedicated to Solana Foundation’s large transaction update, such large transaction size will make it possible to implement zero-knowledge proofs, large multisig transactions along with various signature schemes which were unachievable within one transaction.
In turn, fewer transactions have to be processed, less money is required to pay for signatures, and there is no need to wait for a chain of confirmations. BLS signature schemes, as well as confidential transactions and multisig setups will benefit from this upgrade.
There are two main proposals that are essential for the improvement. SIMD-0296 increases the transaction size from 1,232 bytes to 4,096 bytes, while SIMD-0385 defines the V1 message format and moves transaction configuration into the message itself.
Less than a day after the upgrade was activated Anza CEO Brennan Watt made a post on X mentioning a large transaction on the testnet explorer with the message: “ATTENTION: large transactions have hit the testnet.”
Why one transaction beats a bundle
The importance of atomicity comes from the fact that at times, developers bypassed Solana’s old size limit with Jito bundles. In fact, SIMD-0296 explains that workaround and also mentions that bundles lack atomicity at protocol level as in the case with one native transaction.
According to Jito’s documentation, bundles can include up to five transactions processed sequentially on an all-or-nothing basis. But bundles operate within Jito’s block-engine and compete among validator tips. V1 transaction, on the other hand, refers to a single native Solana transaction, which carries the transaction-level atomicity guarantee from Solana.
For DeFi developers, this can help simplify helping in establishing swap routes, proofs, and other kinds of multistep transactions. If the operation is entirely covered by V1 transaction, it either succeeds or returns it back as a single-native transaction instead of relying on a bundle of transactions.
What DeFi gets beyond extra bytes
V1 modifies the transaction structure as well. Whereas Legacy and V0 processes priority fees and resource requests using ComputeBudget instructions, SIMD-0385 takes these configurations into V1’s transaction configurations, thereby simplifying the job for the validators in the process of identifying resource needs without actively searching through the instructions list.
V1 also eliminates Address Lookup Tables, or ALTs, utilized by V0 for the purpose of shortening account addresses. According to the Foundation’s analysis of the V1’s ALT tradeoff, V1 instead performs address placement inline. This makes validator input easier compared to the method used in heavy account applications in V0, which consume more bytes when converted.
The practical benefit for routers and DEX aggregators is that there is now more room for complex atomic routes. However, this improvement does not come without limitation. Indeed, Solana’s maximum capacity of 64 different accounts remains unchanged, and thus applications that require heavy use of accounts can still suffer from this limitation, even if byte-limit capacity is no longer an issue.
The upgrade is being introduced as Solana is making a leap into on-chain finance. The report about Solana for Q2 2026 provided by Galaxy showcases increasing activity of stablecoins, tokenized equities and real-world assets.
According to Galaxy, in the future, the potential of Solana will largely depend on the ability to use assets in lending, collateralization, margin, and yield, instead of just issuance or trading. Based on DeFiLlama data on Solana, the platform seems to have a strong DeFi presence.
Moreover, the timing of the upgrade matches the Agave 4.2 upgrade phase. Informatively, Cryptopolitan stated that the plan for gradual implementation of cost reduction was expected to lead to savings of about 90%, along with decrease of time needed for processing from 500 to 200 milliseconds, leading to benefits for stablecoin issuers, wallet providers and payment systems.
RPC operators have a week to update
V1 is opt-in, so Legacy and V0 transactions keep working. The issue is whether infrastructure reading V1 traffic is ready.
Anza has obliged RPC providers to move to Agave v4.2.2 or v4.3.0-beta.3. The Helius Agave 4.2 migration checklist also stresses that RPC consumers who don’t state V1 compliance can experience call failures such as getBlock when a V1 transaction is used.
The process began with local testing of V1 on August 24, shortly after which the mainnet date was confirmed as September 9 on August 29. V1 enabled the Testnet on September 1 at epoch 1025, with a scheduled activation on the mainnet on September 9.
This latest testnet run serves as the last preparatory step for RPC providers, indexers, wallets, SDKs and analytical platforms before the mainnet launch.

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

Investment in AI has become an anchor for the world’s biggest economy, a trend which has now been noted by the Fed in its own evaluation of local economic activity.
According to the Beige Book published on September 2, the Fed reported that U.S. economic activity showed only slight uptick in activity until the end of August, while consistently emphasizing the importance of data centers and AI for the overall demand when other segments of the economy have been slowing down.
This is of great significance beyond Washington. The various hyperscalers that are financing the expansion of data centers, such as Amazon and Microsoft, are setting the trend in the world concerning capacity for AI.
If the money, spent on building data centers is helping boost manufacturing and construction in the U.S. economy, the investors have one more reason to regard the AI surge as an economic phenomenon rather than just a tech-sector trend.
What the Beige Book actually flagged
According to the report, which Reuters said includes qualitative readings from all the twelve Federal Reserve regional banks, there has been only a minor increase in employment and moderate increase in prices, based on the available information up to August 24. Businesses expressed cautiousness and cited higher energy costs, uncertainty pertaining to policies, and international conflicts.
Given that background, AI-related activity was particularly prominent. The Fed reported both positive and negative consequences of AI on labor demand, while the manufacturing sector benefited from the orders associated with data centers and defense. The non-residential construction was also increasingly focused on data center-related activities.
One contact quoted in the Chicago Fed’s district report captured the divide:
Without data centers, construction would be in a recession.
This observation also highlights the extent to which sections of the construction industry have become reliant on data-center expenditure, which would raise the stakes should investment from hyperscalers begin to fall off.
The numbers behind the global buildout
The enormity of the investment became more evident later in the same week. According to PwC’s Global Data Centre Outlook issued on September 2, global AI infrastructure capital expenditures (capex) will reach a total of $31.6 trillion by 2050. In the forecast, annual capex in data centers increase from about $800 billion in 2026 to $1.8 trillion in 2050.
According to projections, almost half of that amount, or $15.1 trillion, will be invested in the U.S. while the region of Asia Pacific is expected to seize $8.2 trillion from the overall investments, with China and India being the leaders in this regard. PwC highlights that the lack of affordable, reliable, and low-carbon electricity are the biggest constraints on where that investment can go.
The same issues are being observed by Wall Street. As stated in the report of Goldman Sachs Research published on the 19th of August, the increase in spending by hyperscalers, better performance of servers and models, and opposition to data center establishment in the U.S. question the potential of the ongoing cycle.
Why the timing cuts both ways
The Beige Book arrives ahead of the Fed’s September 15-16 meeting, with the rate outlook still unsettled. As of September 2, Reuters reported that markets were pricing roughly a 65% probability of a rate hike, compared with 35% for another hold.
The shift followed Fed Chair Kevin Warsh’s August 28 Jackson Hole remarks. Warsh did not commit to a hike, but said policymakers would have “work to do” if underlying inflation failed to move clearly and quickly enough toward the 2% target.
As Warsh said in his Jackson Hole speech:
The Fed’s predominant focus right now should be on prices.
That is where the AI boom becomes double-edged. Data-center spending is supporting construction and manufacturing, but it is also increasing competition for electricity, materials, skilled workers and capital. The Beige Book reported elevated input-price pressure in construction and manufacturing, especially for energy, transportation, metals and petrochemicals.
The scale is already visible in investment-bank forecasts. Morgan Stanley’s March 2026 outlook projects about $2.9 trillion in global data-center construction costs through 2028 and estimates AI-related investment could account for roughly 25% of U.S. GDP growth in 2026.
In its previous coverage, Cryptopolitan reported that capital expenditure for hyperscalers, according to another Morgan Stanley report, could go from $805 billion in 2026 to $1.1 trillion in 2027.
If the Fed continues to tighten amid this investment tsunami, financing would become more costly just as businesses, utilities, and government scramble to acquire the computing, power, and infrastructure they require. This makes the September rate decision important not just for American growth, but also for how fast the global growth of AI can continue.

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

Instead of attempting to pursue the money, agents of the US Federal Bureau of Investigation (FBI) broke into the fundraising mechanism itself. Via their own informants, the bureau’s Albuquerque Field Office tracked down the servers, domains, and rotating crypto addresses through which donations were solicited for Hamas, thus being able to take control of parts of that mechanism.
According to the announcement from the US Justice Department (DOJ), the operation resulted in the seizure of nearly $560,000 in cryptocurrency that was meant to go to the military wing of Hamas, the Al Qassam Brigades.
This case shows that public blockchains can also be used against those who are trying to fund illegal activities. The same transparency that makes it possible for donors to find an address enables investigators to track transactions, recreate the flow of funds, and determine where the money has gone.
How agents got inside the donation pipeline
According to the DOJ, the fundraising scheme used an encrypted group chat to direct supporters to a website and provide rotating crypto addresses. Changing addresses may complicate tracing, but it did not stop investigators.
The FBI received assistance from several human resources in mapping the operation from inside. The investigators then carried out a number of cryptocurrency seizures through warrants granted on March 25, June 25, and October 10, 2025. Separate warrants issued in 2026 named domains and servers used to make donations to Hamas and manage the main website AlQassam.ps.
The contents of an unsealed seizure affidavit show how far the takeover went. The affidavit identifies a server at IP address 45.134.141.68 belonging to DataCamp Limited and requires the information provider to block the server and connect it to law-enforcement-designated name servers. The seizure has been approved by Magistrate Judge Matthew J. Sharbaugh.
Once investigators were able to take control of the infrastructure, they could intercept incoming donations rather than waiting until the money moved to try to recover it.
What the takeover exposed about donors
The funds raised was only part of what investigators found out. By taking possession of the fundraising mechanism, the FBI could also get the personal data of people who managed to communicate with Hamas regarding donations, says the Justice Department.
A review of the court documents made by Decrypt mentions Bitcoin, Ethereum, Wrapped Ethereum, Tether, and Tron used in 18 addresses controlled by Tether and in three accounts at Binance.
Tether has a significant role because it can block certain addresses. The seizure warrant issued in June 2025 ordered Tether to burn USDT from certain addresses and issue a similar amount of money to an account controlled by law enforcement agencies.
It fits into a general trend. FATF is warning about the use of stablecoins in illegal activities, while Chainalysis states that stablecoins made up 84% of illegal crypto transactions in 2025.
“Your networks are not secure, your crypto is vulnerable, and we will not stop until your ability to wage war is defeated,” U.S. Attorney Jeanine Ferris Pirro said.
According to Brett Leatherman, the Assistant Director of the FBI’s Cyber Division, the agency “will continue to use its authorities to intercept illicit funds and prevent terrorist organizations from exploiting digital networks.”
Why the ledger keeps beating the fundraiser
There had already been a previous seizure linked to the same funding system. In March 2025, authorities in the U.S. confiscated about $201,400 in USDT, out of which $89,900 was in crypto accounts and $111,500 in different other accounts. The fundraising addresses transferred a total sum of over $1.5 million since October 2024. The event was covered by Cryptopolitan at the time.
The next reports of the authorities prove that the fundraising network did not cease to function. According to the affidavit from August 2026, the blockchain analysis showed that the money from the fundraising network reached an amount of about $3 million by November 2025. The FBI human sources still received the instructions to make donations in February 2026.
That distinction is essential because it indicates that the value of $201,400 seized in March is included in the sum of $560,000, which was confiscated in total under the three warrants of the year 2025, while the figure of $3 million is solely about the money received by the broader network and not seized yet.
Elliptic argues that the traceability of cryptocurrency makes it a bad means of fundraising for terrorist organizations, but urges not to consider every transaction in a related wallet as a sum actually raised by any terrorist organization.
What infrastructure provider was identified in the warrant?
The FBI’s branch in Albuquerque still moves on with its investigation with the help from both its Counterterrorism and Cyber units, as well as the Justice Department’s National Security Division and prosecutors from the District of Columbia.
The donor information collected from the seized infrastructure may become the most important find. The following campaign may change direction from targeting wallets and servers to individuals trying to finance terrorists.
The warrant identified DataCamp Limited, headquartered at 9 Coldbath Square, London, United Kingdom, as the electronic service infrastructure provider controlling the server at IP address 45.134.141.68. The FBI affidavit says DataCamp leases and controls the server, which was used to support the AlQassam.ps cryptocurrency fundraising operation.
The warrant did not accuse DataCamp itself of participating in Hamas’ fundraising. It directed the company to help execute the seizure by redirecting the domain to FBI-controlled name servers and preventing further modification or transfer of the server.
Key data point: The FBI’s Aug. 5, 2026 WHOIS lookup placed the server’s IP address in São Paulo, Brazil, while identifying DataCamp Limited as its owner/provider.


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

Broadcom’s custom AI silicon brought in $16.7 billion in its fiscal third quarter.
CEO Hock Tan told investors the run is not slowing.
Semiconductor solutions climbed 127% to $20.8 billion
Broadcom’s revenue for the quarter ended August 2 was $29.6 billion, up 86% from a year earlier, the company said in an earnings release.
The semiconductor solutions segment grew 127% to $20.8 billion. Infrastructure software grew 29% to $8.75 billion.
AI semiconductor sales were $16.7 billion, a 221% increase from the same period in 2025 and up 54% from the prior quarter, Tan said. He called demand for the company’s custom accelerators and networking gear “very strong.”
Broadcom generated $13.7 billion in free cash flow, or about 46% of revenue. Non-GAAP operating margin was 66%.
Tan said in the earnings release he expects AI semiconductor revenue to climb to $21.7 billion in the fourth quarter. That would be a 236% rise from the same period a year ago.
Hock Tan’s statement in Broadcom’s third quarter fiscal 2026 earnings release, published September 2, 2026.
CFO Amie Thuener said the company sees fourth-quarter revenue guidance of about $34.8 billion, up 93% from a year ago. The company foresees that its non-GAAP operating margin will remain at 66%.
Google’s custom silicon deal runs through 2031
The board declared a quarterly dividend of $0.65 per share. Broadcom’s fiscal fourth quarter ends November 1, and the company cautioned that actual results may vary materially from the outlook.
That demand is from a short list of named clients who are building their own chips to reduce their dependence on Nvidia GPUs.
In April, Broadcom signed a long-term agreement with Alphabet’s Google to supply custom AI chips through 2031, prolonging a TPU partnership that has been in place for about 10 years, Cryptopolitan reported.
The same stretch landed a deal to supply Anthropic with 3.5 gigawatts of Google-powered computing capacity starting in 2027. That commitment could approach 5 gigawatts over time.
In June, ChatGPT’s maker unveiled “Jalapeño,” an inference chip it designed and built with Broadcom. It’s the first part of a larger hardware platform, both companies said.
OpenAI president Greg Brockman framed the move as a way to “serve more intelligence with greater efficiency.” The two companies set a target of eventually powering about 10 gigawatts of capacity.
Broadcom’s play is centered on the ASIC, an application-specific chip that’s less flexible than a general-purpose GPU but cheaper and tuned for specific AI jobs. That tradeoff has found a market in inference, the work of running trained models for end users.
Nvidia’s lead in inference is not quite as absolute as it is in training. None of Broadcom’s customers is ditching Nvidia.
OpenAI’s Sam Altman has publicly said Nvidia makes “the best AI chips in the world.” Anthropic runs Claude on AWS Trainium, Google TPUs, and Nvidia GPUs to distribute its risk.
The smartest crypto minds already read our newsletter. Want in? Join them.

The Wyoming Stable Token Commission on Wednesday adopted Chainlink Proof of Reserve to post verified backing data for its Frontier Stable Token onchain in near real time.
The move pushes FRNT, the state-issued stablecoin, past the monthly disclosure minimum in the federal GENIUS Act.
The Network Firm examines the backing under AICPA standards
The Network Firm, an outside examiner, checks FRNT’s reserves and token supply under AICPA standards. The Chainlink oracle network then posts this verified data onchain and refreshes it near real time.
Wyoming frames the combination as a way to close the gaps of information that spring up between formal reports.
FRNT launched in January as the first stable token issued by a US state. It is backed by US dollars and short-term US Treasuries, and the yield generated by those reserves goes to Wyoming’s School Foundation Program rather than to token holders.
“By adopting Chainlink Proof of Reserve as Wyoming’s exclusive onchain asset verification infrastructure, we’re providing transparent, verifiable confirmation that the Frontier Stable Token is fully backed by high-quality reserve assets,” Anthony Apollo, executive director of the commission, said.
The federal stablecoin law, the GENIUS Act, sets a baseline. Issuers have to disclose reserve composition and outstanding supply once a month.
Wyoming was already over that bar, publishing reserve attestations for FRNT every day. The new integration takes it beyond a daily snapshot.
The commission expects the onchain feed to reflect changes in FRNT’s backing as they happen between reporting cycles, rather than having anyone await the next scheduled disclosure.
Secure Mint would stop an infinite-mint attack before it starts
Wyoming is also lining up a second Chainlink feature, Proof of Reserve Secure Mint, which the commission says it’s still adopting. The mechanism precludes the creation of new FRNT unless the verified reserves are at or above the total supply of the token.
The point here is to avert infinite-mint attacks, where an attacker mints tokens with no backing, and to give holders cryptographic proof that any newly minted FRNT is backed by real assets. That wires the reserve check into the minting process itself.
This comes after Wyoming’s decision in August to migrate FRNT from LayerZero to Chainlink’s Cross-Chain Interoperability Protocol as the token’s sole cross-chain rail. Wyoming did this following a security review, joining a migration that has moved about $15 billion in onchain value from LayerZero to CCIP.
Chainlink recently became the source for the price of Coinbase’s B20 tokenized equities. It also joined a group of banks in the euro-and-won settlement effort Project Pangea and signed on with the Depository Trust and Clearing Corporation.
Earlier, Cryptopolitan reported that DTCC has selected Chainlink’s Runtime Environment to run a collateral appchain anticipated to launch in Q4 2026. LINK, the token for Chainlink, traded Wednesday near $11.07, up about 35% over the prior month.
If you're reading this, you’re already ahead. Stay there with our newsletter.