Cryptopolitan

Kazakhstan, one of the world’s biggest gold buyers, is preparing to sell some holdings as U.S. Treasury yields soar. Bonds now offer returns that bullion cannot: gold itself pays no interest.
The National Bank of Kazakhstan wants gold’s weight in its reserves to reduce from 78% to 75% by year-end. Reserves amounted to 386 tons after the sales were stopped in February 2025.
Foreign exchange reserves in September amounted to $66.4 billion, which makes access to money crucial for the officials.
Kazakhstan has a gold concentration problem
“We haven’t seen levels like this in a long time,” Deputy Governor Aliya Moldabekova said. “This makes the dollar an attractive option, thereby reducing the appeal of gold as an investment asset.”
Kazakhstan is the world’s eighth-largest gold producer, and its central bank can buy all domestic output. That has concentrated its investments.
“With such a share of gold, our holdings are starting to resemble a single-asset portfolio,” Moldabekova said.
Half the planned reduction will involve selling options; physical gold sales will cover the rest. Below $4,000 an ounce, falling valuations could reduce sales.
The bank was the world’s fourth-largest official gold buyer during the first eight months of 2026. Other buyers included Poland, China, and Uzbekistan, according to the World Gold Council.
The official demand for gold increased rapidly when Western nations froze huge Russian funds overseas in early 2022 after their diversification post-global financial crisis.
Gold reached its high in early January of about $5,600 an ounce as it was unsure of what President Trump would implement as policies. In the wake of the US-Iran war that started at the end of February, prices dropped by almost 20%.
The Treasury market is testing Washington, too
The 10-year Treasury yield hit a 24-year high last week; the 30-year yield reached 5.36%, a 19-year high. Treasury Secretary Scott Bessent has tried unusual measures to contain borrowing costs.
On July 31, the U.S. coordinated with Japan to strengthen the yen without Japanese Treasury sales. The New York Fed purchased yen using euros from the Exchange Stabilization Fund. Washington also encouraged Japan to consider FIMA repo loans, exchanging Treasury collateral for dollars.
The yen strengthened from 164 to 155 per dollar. Yet Japan later reported $99 billion spent on intervention from July 30 through August 25 and an $88 billion decline in foreign-securities reserves. That suggests securities sales, although valuations complicate the picture. Japan selling Treasuries could add pressure to American borrowing costs, exactly what Washington had hoped to prevent.
On August 19, the Treasury announced it would at least double long-term securities buybacks for liquidity support. Yields initially eased, then recovered. On September 10, it bought $5.2 billion in long-term debt, financed with short-term borrowing, barely 0.016% of the market. Rates remained elevated.
Meanwhile, technology stocks helped Wall Street shrug off rising oil prices and yields. The Dow and S&P 500 registered their second winning week in three; the Nasdaq posted its fourth consecutive weekly gain.
“Market forecasts indicate that the price of gold will continue to rise,” Aliya said. “Therefore, we would prefer not to plan too far ahead; rather, we would achieve the 75% level and then continue to monitor the situation from there.”
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

ブロックチェーン調査員のZachXBTは月曜日、数兆XRPを発行できてしまうバグの発見に対して、Rippleがセキュリティ企業Veria Labsに支払った25万ドルはあまりにも少なかったと述べた。
この一件を受け、Rippleの研究者への対応をめぐるZachXBTの長年の批判が再燃した。
RippleはVeria Labsにいくら支払ったのか？
ZachXBTは10月12日、Xへの投稿で、RippleがVeria Labsに授与した25万ドルを批判した。同氏は、Veria Labsと共同創業者のケイデン・リャオに対する報酬は「最低でも7桁になるべきだ」と書いた。

Microsoft is stepping up its push into custom AI chips with the launch of Azure Maia 300, its third-generation processor, as it looks to move most of its AI workloads to the Maia platform from 2027 and cut its reliance on Nvidia and AMD.
Microsoft unveiled the chip at the OCP Global Summit 2026 in San Jose, calling it a key part of the AI inference infrastructure behind its products and services.
Microsoft says Maia 300 can process 40% more tokens per megawatt than leading GPU systems, according to Yahoo Finance. The claim puts the chip in direct competition with Nvidia, whose GPUs dominate the market. In simple terms, it means Maia 300 can handle more AI workloads using the same amount of power.
Microsoft says a rack equipped with 96 of the new chips can deliver up to seven times the throughput of one running the previous-generation Maia 200. Microsoft CEO Satya Nadella described the chip as “our biggest bet yet on our own silicon” in a post on X.
Today we are announcing our biggest bet yet on our own silicon.
Starting next calendar year, we will scale our Maia family to multi-gigawatt deployments, making it the primary inference system for our own product and services.
And we’re introducing Maia 300, which delivers 40% more tokens per watt compared to leading GPU-based systems.
We are already deploying Maia 200 across the US, Europe, and Australia, delivering up to 1.7x more tokens per watt across experiences in Teams and Copilot.
Our goal is to build on the full-stack efficiency gains we’ve already seen, while continuing to bring the latest innovations from across the ecosystem into our fleet.
— Satya Nadella (@satyanadella) October 12, 2026
Maia 200 already runs Copilot, launch spreading
Yahoo Finance reported that Microsoft’s Maia 200 chip runs Microsoft 365 Copilot and supports a range of AI models, including those from OpenAI, Microsoft’s own MAI family and other open-source developers. The tech giant says the chip can process 1.7 times more tokens per watt than rival GPU systems when running Teams, Copilot and Cowork.
Microsoft announced its first Maia chip in 2023, but delays pushed the launch of Maia 200 to January 2026. By this summer, it was running in just two US data centers, The Information reported. Microsoft is expanding Maia 200 across its data centers in the US, Europe and Australia, with GitHub Copilot and Microsoft Foundry also expected to run on the chip before the end of the year.
Billions in savings, and a scramble to catch up
GeekWire reported the shift could save Microsoft billions and ease its reliance on Nvidia, while more efficient chips could also also cut electricity draw at a moment when power supply is one of the tightest constraints on all forms of AI expansion.
Evercore ISI analysts estimate Maia will make up about 10% of Microsoft’s AI chip spending by early 2027, according to CNBC.
Microsoft is also entering the custom AI chip market well after several rivals have already made progress. Google is selling its Tensor Processing Units to external customers, while Amazon’s Trainium chips have attracted AI companies such as Anthropic. Meta has also shipped its own chip designs, and OpenAI unveiled its first custom chip, developed with Broadcom, in June. Maia, by contrast, stays completely in-house and cannot be rented by any third parties.
Microsoft’s push to develop its own chips could have implications for Nvidia, which gets more than half of its quarterly revenue from hyperscalers, according to Yahoo Finance. The two companies continue to work together, though, with Microsoft CEO Satya Nadella and Nvidia CEO Jensen Huang holding a joint appearance last week to unveil Nvidia-powered Windows PCs.
The tech giant has also said it will be among the first cloud providers to deploy Nvidia’s Vera Rubin and AMD’s Helios systems. Microsoft also plans to share further details at its Ignite conference in San Francisco next month.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Norway’s data protection regulator has hit Microsoft’s advertising arm Xandr with a 92 million kroner ($9.6 million) fine for sharing Grindr users’ sensitive data with advertisers without a legal basis. The penalty will be the largest the agency has ever imposed on a company.
Datatilsynet, the Norwegian authority behind the ruling, claimed the data involved information that could reveal users’ sexual orientation. The final decision closes a case that has been ongoing in Norway for over six years.
How Grindr data ended up with thousands of bidders
The case involves real-time bidding, a process that sells ad space in milliseconds as a user opens an app. When an individual launched the Grindr app in Norway, their personal data was sent through Xandr and potentially shared with thousands of companies and firms in the advertising industry, according to Datatilsynet. These companies then competed to deliver targeted ads to the user.
Because Grindr caters to gay, bisexual, trans and queer users, the regulator considered app usage alone enough to reveal a person’s sexual orientation. Under European data protection rules, that makes the information sensitive data that requires explicit consent to process.
The Norwegian authority found that Xandr had failed to obtain the required consent. It also ruled that the company shared responsibility for handling the data with Grindr, instead of just acting as an intermediary. Xandr’s direct financial benefit from the data sharing was another key factor in assessing the magnitude of any fines.
Blaming the app did not hold up
Xandr argued that it assumed Grindr had already collected valid consent from its users, an argument that was rejected by the regulator.
“Responsibility for data protection cannot be delegated to others,” Datatilsynet Director General Line Coll said in the authority’s statement, as quoted by Tech Xplore. The director claimed each company must independently ensure and be able to demonstrate that it has a legal basis for using the data it procures.
Coll said the breaches were particularly serious, noting that the sensitive personal data of thousands of Norwegian users had been shared illegally. The data sharing took place between July 2018 and April 2020, before Microsoft acquired the business, then known as AppNexus, PPC Land reported. Xandr now has three weeks to challenge the ruling at an appeal.
The Norwegian decision follows a similar case in France earlier in 2026, where the Conseil d’Etat upheld a €40 million fine against retargeting company Criteo, rejecting its claim that partner websites were responsible for collecting consent. The two decisions make clear that companies remain accountable for data protection, and cannot rely on partners to collect user consent.
A complaint that started since 2020 in Norway
The complaint was filed in 2020 by the Norwegian Consumer Council and privacy group noyb over Grindr’s data-sharing practices. It was one of three complaints lodged that year, targeting both the app’s handling of user data and the adtech companies that received it.
Felix Mikolasch, a data protection lawyer at noyb, said the ruling could force greater accountability from adtech firms such as Xandr, whose operations are often largely unseen and unnoticed by the public, describing the decision as an important precedent for similar cases in a statement from the group.
Finn Myrstad, director of digital policy at the Norwegian Consumer Council, said the decision puts companies that rely on user tracking on notice. The latest ruling comes after Grindr faced a similar penalty in Norway. In 2025, the country’s courts upheld a 65 million kroner fine against the app for illegally sharing users’ sensitive personal information.
The fine targets a company that is valued at around $3.97 trillion, according to GuruFocus. However, noyb has also warned that the precedent could lose its intended effect if the European Commission’s proposed Digital Omnibus weakens how the GDPR treats online tracking IDs. This proposal is currently under consideration by the European Parliament and Council.
The smartest crypto minds already read our newsletter. Want in? Join them.

Tether has frozen $10 million of its USDT cryptocurrency after a theft drained up to $93 million in assets from Ledger hardware wallets sold by CryptoBilis, as the investigation into the money flow continues and Ledger claims the problem is isolated.
Confirmed information provided by Ledger so far
The French cryptocurrency wallet maker has yet to comment on the losses. Ledger has said that all cases the company has identified involve devices sold by CryptoBilis, a third-party vendor that operated in Malaysia, Indonesia and the Philippines. The company also stressed that there were no signs of compromise of Ledger’s systems, services and direct sales.
A discovery added an element of physical tampering to the case of software-related vulnerabilities. On October 10, the firm stated that one of the affected users owned a device that had an unauthorized physical implant installed, which has become the first evidence of physical compromise of Ledger wallets in this case.
The message for buyers from its side is very clear. If you have bought a CryptoBilis device within the last 90 days and did not configure it yet, leave it untouched. If you are using the device right now, transfer your coins to another device with a new recovery phrase.
How investigators tallied the losses
These figures are attributed to on-chain analysts rather than Ledger. Bitquery estimated that the attack involved $93.2 million siphoned off from 315 wallets on six separate blockchains, primarily USDT and TRON coins. The security platform YFarmX set a higher number, claiming the attack took away $94.5 million from 521 wallets on seven different blockchains. Both estimates were based on public chain data.
According to Bitquery’s analysis, most of the stolen money is attributed to TRON at around $70.5 million, with another $16.8 million in Bitcoin and $3.7 million in Ethereum, and much smaller amounts in BNB Chain, Polygon and Solana. The operation is claimed to have been planned for months: the addresses involved in the theft were created on September 25 and made dozens of trial transfers until early October.
The major transfer took place on October 9 when Bitquery claims the attackers rewrote the permissions for 30 wallets on TRON before transferring some 203.8 BTC from 111 Bitcoin wallets within a single block an hour later. This sort of operation can be carried out only if the attacker has the necessary keys.
This is important since the issuers of the stablecoins are able to blacklist coins used for illegal activities. Bitquery, which verified the Tether freeze, reported that there was still about $81 million of traceable tokens on-chain as of October 11, but some funds have passed through the mixer services of Tornado Cash and Zcash.
Pushback over an authorized reseller
However, Ledger’s characterization of the situation as a CryptoBilis issue has not gone over smoothly. The reseller had an official badge of Ledger on their website, hence making Ledger responsible for this situation since, “It’s not a CryptoBilis incident, it’s a Ledger incident.” X user MyCryptoParadise told the company to “take ownership of this mess.”
There is yet another side to the story from the history of the reseller. According to Arravind Prabu, one of the co-founders, CryptoBilis was acquired in March 2026, and the entire team transferred all system access during the acquisition, as quoted in Learning Crypto.
The only new investor in the company was registered in China, as per the company’s reported records. So far, there have been no links made between the thefts and the ownership change and the investigation found no tampering with new investors.
CryptoBilis has suspended sales and deliveries of all brands that it sells and shut down its brick-and-mortar stores in all three countries.
The implant Mark Karpelès pulled apart
Perhaps the most definitive description of the hardware can be given by the founder of Mt. Gox, Mark Karpelès. Karpelès disassembled a Ledger Nano X purchased in Malaysia and, as reported by Learning Crypto, discovered additional components that were hiding under the screen: a cellular modem, an eSIM and a tiny chip that would enable it to intercept what was displayed on the screen.
When setting up the wallet, the recovery words are displayed on that screen. A component that can intercept the data from the display would allow the words to be sent using the mobile network and, as a result, the wallet to be recreated somewhere else and the money to be stolen. It still needs to be proven whether all of the devices contained those components.
It should be noted that this case is distinct from another theft in March, which happened in April when the Ledger Live application in Apple’s App Store was used by phishers to steal at least $9.5 million.
If you're reading this, you’re already ahead. Stay there with our newsletter.

Michael Saylor’s Strategy (NASDAQ: MSTR) spent roughly nine times the $21.5 million on buying 250 BTC during the period between October 5 and 11 on repurchasing its own STRC preferred shares, which it directed $192.4 million into.
Strategy confirmed the nine-to-one STRC-to-BTC purchase in its 8-K filing with the Securities and Exchange Commission (SEC) on October 12.
What did Strategy buy this week?
Notably, the 250 BTC purchase, at an average price of $86,194 per coin, is the latest leg in Strategy’s record stack, bringing it to 848,250 BTC. The Virginia-based firm has spent roughly $64 billion to build its portfolio, paying an average of $75,444 per token.
The bigger headline was the amount Strategy committed to buying back its Variable Rate Series A Perpetual Stretch Preferred Stock (STRC). At an average buyback price of about $99.69 per share, the firm spent $192.4 million to repurchase 1,932,145 STRC shares.
Strategy raised $21 million of the $21.5 million it spent on Bitcoin from selling about 125,000 MSTR common shares via its at-the-market offering program. The remaining $500,000 came from the USD Cash.
Strategy did not sell any preferred stock during the week from October 5 to 11.
Strategy now has $354.8 million left for buybacks under the digital credit securities repurchase program that it expanded to $2 billion in September. That program is different from the MSTR common stock repurchase program, which still holds $1 billion in capacity.
Why Strategy keeps buying STRC
Strategy laid out the logic for buying back its own STRC whenever the preferred share trades below its $100 par value in a July 27 press release announcing the program, which began with a $25 million repurchase.
The company’s stated goal is for STRC to trade near $100 with high liquidity and low volatility.
Strategy’s president Phong Le explained: “At prices below $100 per share, STRC repurchases represent an attractive allocation of capital because they can reduce future preferred dividend requirements at a discount.”
In Strategy’s second-quarter results released July 30, the firm claimed a 13% discount on the $25 million it spent to acquire STRC, which was worth $28.9 million at face value.
The cash position behind the dividends
As of October 11, Saylor confirmed that Strategy reported a USD Reserve of $4.88 billion and a separate USD Cash balance of $640.5 million, together about $5.5 billion. The former is earmarked for preferred dividends and debt interest, while the latter funds Bitcoin purchases and other uses.
In a separate post, he put Strategy’s USD duration at 3.5 years and STRC’s Bitcoin credit at 52 basis points, based on assumptions of 10% annual Bitcoin returns, 40% volatility and a Bitcoin price of $82,252.
The buyback pace continues a pattern from earlier weeks. Strategy repurchased $151.7 million of STRC and bought 1,665 BTC in the period ending September 27, as Cryptopolitan reported.
Strategy’s common stockholders vote on October 28 on a proposal to switch STRC and its other preferred securities to daily dividends. If approved, STRC would see its first daily record date on November 1 and first payment on November 2, with the annualized rate unchanged, according to the company’s investor materials.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Intel (NASDAQ: INTC) CEO Lip Bu Tan specifically commended DeepSeek, Kimi and GLM during a speech at a panel hosted by Singapore lender DBS, where the chief executive of the Washington-backed chipmaker praised Chinese frontier AI models as “good enough.”
Tan’s Monday comments echoed earlier sentiments shared by Jensen Huang, the CEO at Nvidia (NASDAQ: NVDA).
The US government, which owns a roughly 10% stake in Intel, already accused the same three Chinese AI labs (DeepSeek, Kimi and GLM) that Tan commended in a September IP theft release.
Intel’s CEO backs “good enough” and cheaper to run Chinese AI models
Lip Bu Tan told the DBS audience that he is “a strong believer of open source.” He also credited them for being more cost-efficient and that the market would decide which models suit which jobs best, with adoption being “very fast” so far.
Developers can download and adapt open-weight models, including those from the Chinese labs that Tan named, at a fraction of the price it takes to run similar-performing closed models from American labs.
Tan also forecast a future where quantum computing could reach the mainstream within three to five years.
Washington is investigating the Chinese labs Tan praised
Tan’s endorsement came about five weeks after Cryptopolitan reported that a September joint advisory from the FBI, National Security Agency and Cybersecurity and Infrastructure Security Agency accused DeepSeek, Moonshot AI (Kimi) and Z.ai (GLM) along with three other Chinese firms of running industrial-scale distillation campaigns from US models since late 2024.
Distillation, the process of training a smaller model on a larger one’s outputs, is a standard research method. However, Washington has often pushed back at the scale and intent of these Chinese campaigns as crossing a line.
Notably, Tan is not the only US-based executive to praise models from Chinese AI labs. Nvidia’s Jensen Huang dismissed fears in July that Chinese AI could run US firms “off the road.” The CEO argued that US companies should “absolutely” be allowed to use Chinese open-source models, which he described as “excellent” at the time.
During the same month, Nvidia joined Microsoft, Meta, Hugging Face and Mistral as signatories of an open letter asking policymakers not to rush to “premature restrictions” on open-weight models.
While the letter never named China, it coincided with reports about ban and sanction threats from the Trump administration and Treasury Secretary Scott Bessent.
The US government has a stake in Intel
Intel entered into a deal with the Trump administration, under which the US government agreed to buy 433.3 million Intel shares at $20.47 each for an $8.9 billion investment, representing a 9.9% stake, according to Intel’s filing. Counting earlier CHIPS Act grants, total government backing reaches $11.1 billion.
Another wrinkle could be added to the ownership structure based on Trump’s comments suggesting that the government “might” take Intel-style stakes in OpenAI and Anthropic, the two labs loudest about Chinese distillation.
That would leave Washington owning both the chipmaker praising Chinese models and the labs accusing those models of theft.
If you're reading this, you’re already ahead. Stay there with our newsletter.

AI infrastructure spending is helping the global economy maintain growth of around 3%, while creating a financial windfall for Asian hardware suppliers, IMF Managing Director Kristalina Georgieva said in Bangkok, where finance officials gathered for the Fund’s annual meetings.
An investment shock cushioning the energy supply problem
Georgieva described the AI boom as both a demand and investment shock, Reuters reported. She said the spending has helped keep output from falling further, even as disruptions to energy supplies continue to weigh on economic activity.
“So that shift into higher gear is what is keeping the world economy growing at around 3%, despite the fact we are being dragged down by the energy supply shock,” she told the IMF and World Bank gathering in the Thai capital.
The 3% figure matches the projection the Fund issued in July, and an updated World Economic Outlook is scheduled for release on Tuesday, October 13. The update will give the first read on whether the AI capital cycle has bettered the Fund’s numbers in any way.
Where Asia’s hardware makers fit
Countries supplying the hardware needed to expand AI infrastructure stand to gain the most from the spending, and Thailand is among those best positioned to benefit, according to IMF Managing Director Kristalina Georgieva.
The growth boost is expected to differ across Asia. Advanced economies in the region could see growth rise by 0.5 to 1 percentage point, while emerging markets could gain 0.2 to 0.8 percentage points, according to Traders Union.
The United States, China and India continue to lead AI development. Still, the rising demand for data centers, components and related equipment is extending the investment boom beyond AI leaders to manufacturers across the Asian region.
The meetings are expected to draw around 18,000 people, including officials from the IMF’s 191 member countries. US Treasury Secretary Scott Bessent and Spanish Finance Minister Arcadi España are reportedly set to miss the meetings to attend to domestic matters.
IMF gives labor warning alongside growth story
Georgieva paired her optimism with a warning, calling AI “a slow-moving tsunami” that could eventually augment, downgrade or eliminate 40% of jobs. She also pointed to an early study showing that 10% of US jobs already require AI skills.
Jobs involving routine, repetitive tasks are becoming less common, while hiring is growing in some lower-paid sectors, including construction and data centers. The shift from labor to capital also presents a challenge for governments, which may need to overhaul their tax systems as more economic value accrues to capital and not just workers.
Singaporean President Tharman Shanmugaratnam also warned that entry-level job opportunities, especially for recent graduates, are already shrinking, and countries that are slow to adapt risk making the gap between richer and poorer economies much wider.
World Bank bets on small-scale tools
World Bank President Ajay Banga said poorer economies could benefit from a more practical approach to AI, focusing on smaller applications that require less computing power instead of attempts to compete in the race to build LLMs.
The World Bank is supporting the use of these tools in sectors such as agriculture, healthcare and education, Reuters reported.
The aim is to let developing countries put AI to practical use without building out costly infrastructure first. It remains to be seen if that would keep them in the game, as countries with weaker technology bases risk falling behind since capital flows to the suppliers and producers already positioned to take advantage of the AI boom.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Taiwan Semiconductor Manufacturing Co. (TSMC) is expected to post its sixth consecutive record quarterly profit when it reports results on October 15. Analysts forecast net income of about NT$746.3 billion ($23.4 billion), with demand for AI chips continuing to rise amid the general AI boom.
Analysts expect TSMC’s third-quarter net profit to rise about 65% year over year, according to a Reuters report citing an LSEG SmartEstimate based on forecasts from 16 analysts.
A profit figure above NT$706.56 billion would set a new quarterly record and mark the company’s 11th consecutive quarter of profit growth. The earnings call is set to drop at 0600 GMT, which is 2 p.m. in Taipei.
As the world’s largest contract chipmaker, TSMC supplies chips to major technology companies, including Nvidia and Apple. Valued at about $2.1 trillion, TSMC is Asia’s most valuable listed company, with a market cap at almost twice that of Samsung Electronics.
Revenue already cleared the bar
The sales numbers for the quarter are already well known and discussed, with TSMC reporting third-quarter revenue of about NT$1.49 trillion ($46.71 billion) on October 8, a 50% rise from one year earlier. This was higher than the NT$1.46 trillion SmartEstimate and came in above even the upper limit of the company’s own July guidance set at around $44.6 billion to $45.8 billion.
September carried most of the weight, with a monthly revenue of NT$511.86 billion, up 54.6% from last year. Across the first nine months of 2026, sales hit NT$3.90 trillion, equaling a 41.1% gain. TSMC has since raised its full-year forecast, projecting revenue growth of slightly more than 40% in US dollar terms.
Where the margin pinch comes from
TSMC’s net margin is expected to fall to around 50% in the third quarter, compared with 55.62% in the previous quarter, based on profit and revenue estimates cited in a Yahoo Finance earnings preview.
TSMC’s second-quarter margin was also lifted by a one-off gain of NT$63.20 billion from selling an 8.1% stake in Vanguard International Semiconductor. That boost is unlikely to recur, leaving the company without the same contribution in the current quarter.
TSMC’s transition to 2nm production is also expected to weigh on the company’s income margins. The company has said the manufacturing change could cut second-half gross margin by 3 to 4 percentage points, as early production faces low yields and high depreciation costs before output can reach previously attained levels of scale.
TSMC also raised its capital spending plans in July, setting a 2026 budget of $60 billion to $64 billion, up from its previous range of $52 billion to $56 billion. The new target is also significantly higher than the $40.54 billion the company spent in 2025.
Morgan Stanley analyst Charlie Chan expects TSMC to raise its full-year revenue growth forecast to the low-to-mid 40% range, according to an October 6 report. He pointed to strong demand for the company’s 3nm chips as a key factor behind the expected revision.
Texas investment question that remains unanswered
TSMC is putting $265 billion into factories in Arizona, including an extra $100 billion, which was announced in July. Reuters reported in September that TSMC was looking at an investment in Texas, and Tesla CEO Elon Musk has said his companies would build and operate the Terafab AI chip complex, with TSMC potentially subleasing a part of the site.
“Texas is a good location for a 2030s or 2040s expansion grand plan,” Morningstar analyst Felix Lee said, suggesting the proposal “may be a dangling carrot to preempt potential US trade and other policies.”
TSMC shares have seen a 63% increase in Taipei this year, slightly behind the general market’s 69% gain. With another record quarter largely expected, investors may require a little more than an earnings beat to take the stock to higher prices. Attention could instead turn to management’s fourth-quarter margin outlook, which may have a greater influence on the shares than just the headline profit figure.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Brad Smith, Microsoft’s president, has told the BBC that AI is currently working with fewer safety rules than commodities like milk or washing machines, and this should push governments, with the UK out front, to write binding controls, including kill switches and mandatory reporting when AI agents misbehave.
Smith made the comparison in a BBC interview published over the weekend, where he described AI as both a “tool and a weapon” that needs urgent oversight. “Right now, your washing machine is subject to more safety rules than AI,” he said, according to the BBC. “Every carton of milk that you see in the grocery store is subject to safety rules and inspection.”
Smith is calling for a legal structure that would make AI companies more accountable when things go wrong from their products or end. Under his proposal, firms would have to report model failures, alert anyone whose systems are breached by a rogue AI agent, and submit powerful models to independent scrutiny before making them public.
A split with Trump over self-policing
Smith’s push for legally enforceable AI safeguards puts him at odds with the White House. Trump has argued that the US should let its AI industry regulate itself, warning that tighter rules could leave the country behind China. President Trump reiterated that position at a summit with technology executives last month.
The summit also saw executives from OpenAI, Anthropic, Nvidia, Meta, Google and Elon Musk’s X sign a one-page agreement that Cryptopolitan described as “morally binding.” The signatories pledged to use independent auditors and put safeguards in place to prevent AI models from going rogue. Microsoft, however, did not sign this accord, according to The Telegraph.
Smith welcomed the agreement and said Microsoft would be willing to sign it. But in an interview with the BBC, he cautioned that voluntary commitments alone would not go far enough. He called the agreement “a good step” and “a necessary step,” but added that he did not thik it was sufficient any longer.
He warned that voluntary commitments may do little to curb the biggest risks, since companies that already take safety seriously are more likely to follow them, while those that pose the greatest danger can choose to ignore them.
Microsoft’s Smith wants Britain to lead
Microsoft’s President Smith also named the UK as the country best placed to set global standards and not the US or even Brussels. He pointed to its AI Security Institute, launched in 2024 under then prime minister Rishi Sunak as the AI Safety Institute. The body has secured voluntary agreements allowing it to test frontier models before they ship.
According to Smith, the UK is in a stronger position than the EU when dealing with Washington, as US policymakers view London differently from Brussels, he told the BBC, adding that the EU’s risk-based AI Act has made US policymakers wary.
The UK government is also preparing to tighten its oversight of AI, with AI minister Kanishka Narayan expected to call on developers of frontier models to do more to address safety risks in an upcoming speech, as legislation to strengthen government oversight continues to take shape.
Separately, The Telegraph also reported that the White House had restored the UK institute’s access to the latest AI models for testing. The access had previously been withdrawn after firms including Anthropic and OpenAI were asked to stop sharing their models.
Kill switches and rogue agents
Smith’s “emergency brakes” idea echoes Microsoft CEO Satya Nadella, who called for AI models to carry a mechanism that lets a human pause or shut down a system mid-task if it runs out of control by any chance.
“We must build contained systems whose behaviour we can observe, limits we can test, and actions we can always contain,” Nadella had previously written, according to the Guardian.
The warnings follow real incidents, with Anthropic recently finding out that a version of its Claude model had contacted police and falsely claimed it could help solve a murder during testing.
At Trump’s gathering, AI executives including Dario Amodei, Sundar Pichai, Greg Brockman and Jensen Huang supported letting the industry regulate itself. Critics have pushed back against that approach, arguing that it would allow companies decide for themselves what would qualify as safety.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Based co-founder Edison said the company has alerted Singapore’s data regulator and filed a police report following a recent breach.
The incident happened on October 5, and the company reported that the internal dashboard behind its Visa card program may have exposed cardholders’ names, passport numbers, birth dates and home addresses.
What update does Based have for its affected users?
In a post on X on October 11, Based’s cofounder Edison said the company’s compliance and operations staff have been contacting people whose data was caught up in the incident. The company has involved the Personal Data Protection Commission (PDPC), Singapore’s privacy watchdog, along with other regulators.
Edison said the team is coordinating with law enforcement on the case.
Cryptopolitan reported that an intruder reached the firm’s internal operations system and pulled customer KYC files tied to wallet addresses. The exposed set included names, dates of birth, nationalities, passport or Singapore national ID numbers, email addresses, phone numbers and home addresses.
The data was collected by the card issuer for the Based Visa Card, a spending product that pairs with the company’s trading app and advertises up to 8% cashback.
Based has raised $11.5 million in a Series A led by Pantera Capital and says it has passed 100,000 registered users.
The number of customers affected by the breach is yet to be revealed. It is also not confirmed if any funds were moved during the incident.
Is it only Based’s card platform that was breached?
Based insists that its trading app and wallets were not affected because wallets run on a self-custody model, the platform holds no keys, and social login is handled through Privy. Its card business will continue running normally while the investigation continues
Edison addressed questions about an admin backend that had apparently leaked five months earlier. He clarified that the admin backend system was an early OpenClaw admin prototype used for internal testing in March 2026 that has since been retired. It is unrelated to the breach as it had nothing to do with the card business and held no customer data.
Edison’s post goes on to state that Based’s current priorities are supporting affected customers, tightening internal security controls and monitoring, and continuing to build out trading.
However, identity records tied to wallet addresses like the ones leaked in the Based hack can lead to “wrench attacks,” which are physical assaults and kidnappings aimed at forcing crypto holders to transfer funds. Chainalysis found that victims are often picked out using leaked data, social media or insider tips.
The Based incident is also one of a series of KYC exposures across the industry. Cryptopolitan reported that Revolut disclosed an attack in which a spoofed government email was used to pull customer identity documents and transaction histories. The attackers then demanded roughly $3 million.
If you're reading this, you’re already ahead. Stay there with our newsletter.

OneKey Founder and CEO Yishi Wang has announced that the hardware wallet maker’s “dealer network will shrink significantly,” after a full review of previously certified dealers, while those with “unverifiable operating entities or unclear ultimate beneficiaries will have their authorization revoked and be removed from the official website.”
Yishi also said OneKey has revoked CryptoBilis’s authorization as a dealer for its devices.
That warning lands after one of the Ledger wallets sold by Southeast Asian reseller CryptoBilis was reported to have contained an unauthorized hardware implant, as the inquiry continues into a crypto theft with losses now estimated between $86 million and $93 million.
What has Ledger said about the CryptoBilis wallet hack?
An October 10 update shared on Ledger’s support page confirmed that pending the end of its investigation, CryptoBilis will not be selling anything from the hardware wallet maker. That statement came after Ledger said that at least “one of the impacted users’ devices contained an unauthorized hardware implant.”
Ledger stressed that its own systems and direct sales channels were untouched, pointing out that the exposed devices all traced back to CryptoBilis.
Ledger is advising anyone who bought wallets from the Southeast Asian reseller not to initialize any devices they have not set up yet. And for anyone who has already set them up, they are advised to transfer their funds to wallets issued directly by Ledger, not bought from resellers.
So far, security analysts have estimated that about 300 Ledger wallets have been drained of an estimated $86 million to $93 million.
Why OneKey and SafePal buyers are nervous
While Ledger wallets have been the only ones to report thefts so far, the security scare has spread to customers of other wallet makers. CryptoBilis lists itself as an authorized Ledger reseller of Ledger, Trezor, OneKey, Tangem, SafePal, imKey and other wallets in Malaysia, also operating in Indonesia and the Philippines.
CryptoBilis lists the hardware wallet devices it sells. Source: CryptoBilis
OneKey founder and CEO Yishi Wang moved to calm the panic on X, writing that “We have terminated CryptoBilis’s authorization; it is no longer an authorized OneKey dealer.”
Yishi also detailed OneKey’s three-step response to mitigate exposure now and prevent further escalation.
Step one: OneKey has stopped adding new dealers, effective as of the October 12 statement.
Step two: All existing dealers, including previously certified ones, will go through a fresh review process. Dealers with “unverifiable operating entities or unclear ultimate beneficiaries” will lose their authorization. OneKey will publish the full list of authorized dealers on the official website.
Step three: Users who bought OneKey devices from CryptoBilis between March 1, 2026, and October 12 are eligible for a free replacement. Even users who have activated or used their devices can still get a free replacement.
Per Yishi, OneKey went with the March 1 cutoff date because CryptoBilis’s original co-founder publicly stated that the company completed its operational management handover in March after it was acquired earlier this year.
OneKey’s only publicly reported hardware flaw happened way back in 2023, when security firm Unciphered extracted a device’s seed in a lab attack. OneKey patched the gap, paid a bounty and said no users were affected.
Who owns CryptoBilis now?
Fresh corporate records point to a change of control at the Malaysian reseller before the suspected supply-chain attack. CryptoBilis’s former co-founders said on LinkedIn that they handed the company to new owners in March, with reports claiming that ownership passed to an individual identified as Jiaming, a Chinese national.
There are no confirmed reports linking the new ownership with this security incident.
So far in 2026, six hardware wallet brands have disclosed security problems. Galaxy Research put losses above $115 million in the Coldcard case. This early October Ledger disclosure is the second case where funds were lost.
Cryptopolitan reported the BasedApp breach that exposed customer KYC records last week.
If you're reading this, you’re already ahead. Stay there with our newsletter.

シンガポール、2026年10月 — バーチャルおよび実物のVisaカードを提供する暗号資産決済プラットフォームWebitPayは、TOKEN2049 Singapore 2026に参加し、日常の支払いでデジタル資産をより実用的に活用できるよう設計された決済ソリューションを紹介しました。
デジタル資産業界の進化に伴い、注目は取引や投資から、現実世界での実用性へとますます移りつつあります。暗号資産でチャージできる決済カードは、デジタル資産と使い慣れた決済体験をつなぐ実用的な手段です。既存の金融インフラを通じて、日常的な決済サービスを利用できます。

事情に詳しい関係筋によると、Moonshotの経営幹部は現在、株主の一人が不正行為の疑いをめぐって賠償を求めている、注目を集める法的紛争の和解に向けて交渉している。
具体的には、Recurrent AIは経営幹部のヤン・ジーリン氏とジャン・ユータオ氏を不正行為で告発している。関係筋によると、3月11日付でMoonshot AIの投資家に宛てた書簡の中で、Recurrent AIの法定代理人は、2人が投資家を欺き、同社に損害を与えて利益を得たと主張した。Recurrentは現在、約7億ドルの賠償を求めている。

StarknetのSTRKトークンは過去7日間で80％以上上昇し、時価総額上位100銘柄のアルトコインの中で最も高いパフォーマンスを記録した。トークンは日曜日に0.126ドルまで上昇し、昨年12月以来の高値をつけた。この上昇は、ネットワークが10月8日に「L1化を積極的に検討している」と発表した直後に始まった。同社はこの一歩を、2027年までに初の完全な量子耐性ネットワークを実現する道筋として掲げている。
Starknet、「Ethereumと同程度の安全性しか持てない」と説明
StarkWareのCEO、Eli Ben-Sasson氏は、「ポスト量子時代への適応力」を得るためにL1になることが良い考えなのか、それとも悪い考えなのか、フォロワーに問いかけた。その理由は、ロールアップの仕組みに基づいている。Ethereum上のレイヤー2であるStarknetは、Ethereumからセキュリティを借りているため、基盤となるチェーンと同程度の量子耐性しか持てない。

スーパー・マイクロの元請負業者、ティンウェイ（ウィリー）・サン氏は10月8日、NvidiaのAIサーバーを中国に違法に送った罪をマンハッタンの連邦裁判所で認めた。
彼の罪状認否は、規制対象技術が他国を経由する際に、その流れを監視することの難しさをワシントンが抱えていることを示している。
3月の起訴から有罪答弁まで
司法省は3月、サン氏、スーパー・マイクロの共同創業者イーシャン（ウォーリー）・リアウ氏、台湾を拠点とする営業マネージャーのルイツァン（スティーブン）・チャン氏を起訴した。
この事件は、米連邦地方判事のエドガルド・ラモス氏が担当している。サン氏は罪を認めた一方、リアウ氏とチャン氏に対する疑惑は、まだ立証されていない。

Suiが今月後半にビットコイン金融ネットワークを稼働させると、Anchorage DigitalはHashiに接続する最初の企業の一つとなる。
すでにビットコインを保有資産として計上している企業は、ビットコインを売却せずに現金を調達するため、Hashiの2つのルートを利用できる。
Suiは木曜日、シンガポールで開催したカンファレンス「Basecamp 2026」で、この発表を行った。Hashiのメインネットは段階的に公開され、パートナー各社の統合作業が完了するにつれて、利用できる人が増えていく。
Atlasは適格カストディを担い、Portoはヘッジファンドを支援する

ドイツ銀行の元トレーダー、クリスチャン・ビッター氏の金利不正操作に関する有罪判決が金曜日に覆された。ビッター氏は2018年に容疑を認めていたにもかかわらず、判決は覆された。これにより、英国で金利不正操作をめぐって下された9件の有罪判決のうち8件が覆されたことになり、重大不正捜査局（SFO）にとってまたしても痛手となった。
さらに、この判決は厄介な問題を提起している。すなわち、不正行為があった事件、とりわけ陪審員が関連法について誤った説明を受けていた事件で、裁判所はどう対応すべきなのかという問題だ。

AI業界の多くの人々は、今後6〜12カ月以内にAIが引き起こす重大な事故が起きると予想している。Axiosによると、AnthropicやOpenAIをはじめとする大手開発企業の幹部らは、事故後に規則を作ることになる議員たちへの働きかけをすでに進めている。
幹部らは、今の時点でAI関連法案が成立する見込みはないと認めている。そのため、最初の災害が起きた後にワシントンが作るであろう緊急立法に先手を打とうとしている。
サンダース氏の法案、違反者に最長20年の禁錮刑を科す
政策立案者らは、11月3日の中間選挙後、民主党が優位な立場に立ち、先端AIへの対策に速やかに乗り出すと見ている。しかし、この動きは、技術を理解していない古参議員が多い議会、データセンターへの投資に依存する経済、そして誰もが今すぐダウンロードできるオープンウェイトモデルと衝突する。