Actualités financières en direct

AI model developer Anthropic has revealed that over 25% of its research and development efforts are driven by artificial intelligence itself. This disclosure highlights the company’s increasing reliance on AI for frontier-model development. However, …

AI model developer Anthropic has revealed that over 25% of its research and development efforts are driven by artificial intelligence itself. This disclosure highlights the company’s increasing reliance on AI for frontier-model development. However, concerns have been raised by former researchers regarding the diminishing role of human oversight in these processes. Anthropic, known for its Claude series of AI models, continues to push the boundaries of AI capabilities with its latest releases in the Claude 5.5 family. The company’s strategy to heavily integrate AI in R&D may influence market perceptions and valuations. Key Takeaways Anthropic’s decision to leverage AI for over a quarter of its R&D suggests a strategic shift towards more AI-driven processes. Market pricing appears to interpret this increased reliance on AI as consistent with potential valuation growth, though concerns about reduced human oversight persist. The market’s view on Anthropic’s valuation by December 31 shows a range of outcomes, with significant support for higher valuation scenarios. What to Watch Market participants will likely monitor upcoming announcements from Anthropic regarding further AI model advancements and potential strategic partnerships. Developments such as new funding rounds or enhanced collaborations with major tech companies like Amazon and Google could influence market expectations. Observers will also be attentive to any shifts in regulatory landscapes or public sentiment that could impact Anthropic’s valuation trajectory.

Checkout pages now commonly accept Bitcoin or USDC alongside Visa. In some countries, shoppers cannot complete a card payment at all. Merchants selling digital goods or expensive items internationally have noticed that a crypto …

Checkout pages now commonly accept Bitcoin or USDC alongside Visa. In some countries, shoppers cannot complete a card payment at all. Merchants selling digital goods or expensive items internationally have noticed that a crypto option sometimes closes a sale where the card form does not. Settlement can also be faster than a card payout, with no risk of a chargeback. The unglamorous part is behind the button. The plugin, the API, the admin user, and, in many setups, a wallet that cannot be reversed all become part of the store. If the wallet isn’t secure, a login isn’t protected, someone pays out on the wrong chain, a connector goes unused for months, or half the team can open the payment panel, you have built a hole instead of a feature. Plan for this as you already do for stock, refunds, and payouts. Do it when traffic is low. Why eCommerce Businesses Are Adding Crypto Payments Only a few reasons make it through a finance meeting: You are selling into a country where cards fail Settlement is faster than waiting on a card acquirer Your buyers already pay from their own wallets One processor is dictating the rules for your entire shop This doesn’t make the value of the coin stable unless you change it into a stablecoin as soon as you place the order. Tax offices also treat incoming crypto differently from country to country, and they expect records that card firms already hand you. If you send USDC to an address that expects it on a different chain, the money will be gone. For those selling subscriptions, digital downloads, or expensive goods to customers worldwide, the trade-offs are usually manageable. We can’t accept adding the feature and hoping it will be used later. Understand How the Crypto Payment Flow Works When you’re paying, you can choose to pay with crypto. The store puts together a request: destination plus amount. The shopper pays for the items themselves. Once the network has confirmed the transfer several times, the order will be marked as paid. You can later move that value into a bank account or a treasury wallet. If you use a processor, they often convert and pay you out in regular money, just like a card acquirer would. If you run the wallets yourself, somebody inside the company has to take money out of the checkout wallet and put it somewhere that the website cannot access. The main difference between hiring a processor and doing it in-house is who owns the address creation, confirmation logic, settlement, and storage. Choose Between a Payment Processor and Direct Wallet Management Most shops that don’t do much business are better off with a processor. Integration is faster, the books are cleaner, and the keys never sit on your laptop. If the plugin has a bad toggle, it’s annoying. It is less likely to empty an account before anyone notices. In-house wallets start to make sense when you can see transaction processing fees, when you care about the exact time money moves, or when the business needs you to hold the asset instead of selling it. This route also means your people handle matching payments to orders and every outbound transfer. Be honest about the team, the workload, and how much extra process finance will actually be needed. “We want to be independent” is not a security model. Choose the Right Wallet Setup for Business Funds If you copy your own MetaMask habits into a company, you will make things complicated. The wallet checkout is for taking orders. Don’t use it for company reserves. A custodial provider that holds the keys is usually faster to connect with and easier to recover if someone loses access. If you hold the keys, you have control, and if you make a mistake, it is your problem. Only keep the hot wallet with you if you need it for current orders. Schedule the rest and move it to cold storage or a custody setup the storefront can’t access. If an attacker lands in a plugin, they should have an almost empty balance, not the money made that month. Secure the Checkout and Payment Integration Most incidents start when the shop’s system connects to the payment system. If one part of a system is weak, the whole system can be at risk. Use connectors that still receive updates. Install those updates. Give one job access to the API. Please do not share one admin password around the office. After you change the theme, plugin, or checkout layout, place a test order and watch the funds arrive. A screenshot of a green button is not proof. The site and the payment layer are one surface, even if two different companies built them. Control Who Can Access Crypto Payment Systems The developer who updates the plugin shouldn’t be able to read balances. The support agent who pastes a transaction ID into a ticket shouldn’t be able to send a withdrawal. The person doing month-end doesn’t need to edit gateway settings. If one user holds all those rights, a single stolen password can cause a full incident. Give each person their own login details. Turn on two-factor authentication. It’s more important to think about what a job is like than how long someone has been in it. Seeing a payment is not the same as moving money or changing a payout address. Once the shop is no longer three people, shared admin accounts also wreck any later investigation. Businesses that handle company crypto at scale need a clear permission structure: one person views activity, another starts a transfer, a third approves it. Cryptobanco is built around exactly that model. Use Approval Workflows for Moving Business Funds If a customer pays you and you pay treasury, these are different events. If an incoming payment doesn’t go through, you still have the order and the buyer. If you send money to the wrong account, you can’t cancel it. Treat treasury withdrawals, large sends, edits to settlement destinations, and first payouts to an address you have never used as actions that only two people should do. The person who clicks “send” should not be the one who approves. If someone hacks your account, the thief will stop at the first door instead of making off with your money. Verify Wallet Addresses and Blockchain Networks Carefully If you type the wrong character, you won’t get your money back. If you send the right token on Ethereum to someone waiting on Solana, you get the same result. Card rails can sometimes stop you from making mistakes like that. These rails will not. Keep a list of places you regularly pay for, and allow them if the software lets you. Ask someone else to check a new address before you send a large amount of money. When you get your first big payout, send a small amount first and wait. Malware can also replace an address the moment you copy it; checking only the first and last four characters won’t catch a well-made swap. Read the whole text, or copy and paste it from the book instead of the clipboard. Protect Private Keys and Recovery Information Keep your own wallet and private keys; your balance is the key. Phrases you use to recover information, PINs for your hardware, and backup files do not belong in Google Drive, 1Password shared with eight people, or Slack. Two or three named people should know where the physical copies are kept, and those copies should not be connected to the office network. A hardware device stops confidential information from being sent to any laptop that also opens emails. If the balance is large enough that you wouldn’t let one employee transfer it from the bank account, don’t let one employee move it on-chain either. Split-key setups (MPC) or a custody provider are there for that exact reason. Monitor Crypto Payment Activity Start with a baseline, then watch for anything that differs from it. Any money paid to you should match the orders placed. Transfers should always be finished. The outbound activity should be the same as your normal payout rhythm. Report any new payout addresses, new admin users, and logins from unusual places on the same day, not in next month’s spreadsheet. Once a chain transfer has been confirmed, it cannot be reversed. However, many attacks still need a login, permission changes, and a new destination before the money can be moved. If you only notice the problem when you look closely, you are writing an incident report, not preventing one. Plan for Refunds and Customer Support A card refund rides the same rails backward. A crypto refund is a brand-new outbound payment: you send to the customer’s wallet, correct asset, correct chain, address confirmed in writing first. Write the sequence down now and follow it every time: Confirm the customer’s wallet address in writing Check the correct network matches the asset Get sign-off on the refund amount Send a small test transaction first on larger returns Record the transaction ID before closing the ticket Put the policy on the checkout page so the first angry customer isn’t the one who invented the process. Keep Accounting and Reconciliation in Mind Link each payment to an order, an invoice, a refund, or the settlement line. The transaction ID and the timestamp are the join keys. If you need a fiat number for tax purposes, store the rate from the moment you receive the funds. Make sure you connect that to the accounting tool or ERP system while you still have twenty orders, not two thousand. A tab in Sheets is enough for a pilot. It’s not a good channel. People doing audits and tax filings both want a trail they can easily follow back. Create a Crypto Payment Security Checklist Walk this before go-live, then once a quarter. Security control What breaks if you skip it Owner Payment integrations still maintained and updated Stale connectors are a common way into shops Developer / tech lead Multi-factor authentication on every payment login A password by itself is weak on accounts that can move money Anyone with access Personal logins, nothing shared After an incident you cannot reconstruct events Admin / operations Rights limited to the actual job One breach then reaches everything Admin / operations API credentials scoped to a single task Fat keys are a documented target Developer Checkout wallet separate from reserves A drained hot wallet should not take savings with it Finance / owner Second person on outbound treasury payments One compromised user should not finish a send Finance / owner Whole destination reviewed before money leaves A typo here is permanent Finance / operations Recovery material kept offline, physically A digital copy disappears with one stolen login Owner / named custodian Daily eyes on payment activity Finding it at month-end is finding it late Finance / tech lead Refund steps written before the first sale Made-up crypto refunds go wrong Operations / support Transaction ids stored against orders and invoices No join, no audit trail Finance Offline backups and a short incident plan You will need both on a bad day Owner / operations Common Mistakes eCommerce Businesses Should Avoid If you take money from the store into your own wallet, it messes up the accounting and removes every way the company controls what its customers can do. If the hot wallet holds more than current orders require, that extra balance sits reachable from your web infrastructure — and that is exactly what an attacker goes after. If you use one shared admin password, you won’t know who clicked what. Developers who can both write code and transfer funds are often combined into one role. The rest of the list is just as ordinary: nobody double-checks a new payout address, plugins are on last year’s version, the first refund request is also the first time anyone thinks about refunds, and no one writes down who can authorise a transfer. When that person is on leave, or their account is gone, you have a problem. You have two problems at once: a process problem and a security problem. Final Thoughts A crypto button does not make a shop weaker. An improvised crypto button does. Keep the features inside the payment stack that you already know about: connectors that still get patches, wallets that follow how money actually moves through the company, tight access, a second person on outbound funds, someone watching activity, and procedures that exist on paper before the first problem arises. Do that work now, while you still remember how to do it. Open the table above, mark the rows you do not have, and treat those rows as the project. Everything else is just commentary.

Mark price is a calculated reference price that a crypto platform uses to estimate the fair value of a contract. Last price is simply the price of the most recent trade executed on that …

Mark price is a calculated reference price that a crypto platform uses to estimate the fair value of a contract. Last price is simply the price of the most recent trade executed on that market. The distinction matters most in leveraged futures. Last price shows where traders are actually transacting, while mark price is commonly used in margin, unrealized PnL and liquidation calculations. Key Takeaways Last price is the price of the latest executed trade. is the price of the latest executed trade. Mark price is a calculated reference designed to reduce the influence of temporary local price spikes. is a calculated reference designed to reduce the influence of temporary local price spikes. Index price is a broader spot-market benchmark, usually based on prices from several exchanges. is a broader spot-market benchmark, usually based on prices from several exchanges. Major futures platforms generally use Mark Price or Fair Price in liquidation calculations rather than relying only on the latest trade. Mark Price and Last Price can diverge during volatility, thin liquidity or unusual futures-market conditions. Calculation and PnL-display rules differ by platform. What Is Mark Price in Crypto Futures? Mark Price is a reference price calculated by a trading platform rather than the price of a specific executed trade. It is especially important in leveraged products such as perpetual and expiry futures because exchanges need a reference for margin, unrealized PnL and liquidation calculations. Mark Price is not limited exclusively to derivatives. Some platforms also apply related mark-price mechanisms to margin products. For example, OKX publishes Mark Price rules for both contracts and margin pairs. This article focuses on futures, where the concept has the greatest impact on liquidation risk. Most systems start with an Index Price, which represents the underlying asset across spot markets, and then account for conditions in the derivatives market. The three prices therefore serve different purposes: Index Price: broader underlying spot-market reference. broader underlying spot-market reference. Mark Price: calculated risk or fair-value reference. calculated risk or fair-value reference. Last Price: latest transaction in the futures market. Mark Price should not be treated as the single “true” price of an asset. It is a methodology-based reference built for specific risk-management purposes. Mark Price vs Last Price: What Is the Difference? The main difference is straightforward: Last Price comes from an actual trade, while Mark Price is calculated. Factor Mark Price Last Price Meaning Calculated fair-value reference Most recent executed trade Source Index plus platform-specific methodology Exchange order book Actual transaction? Not necessarily Yes Main use Margin, PnL and liquidation calculations Trading and execution context Reaction to local spikes Designed to reduce their influence Reacts immediately Can differ from the other price? Yes Yes Suppose a large market sell order moves rapidly through a thin futures order book. The latest executed price may fall sharply even though prices on other markets have barely changed. Last Price immediately reflects that trade. Mark Price is designed to depend less on one isolated transaction by incorporating broader market information. This does not mean Mark Price is always less volatile or inherently more accurate. Its behavior depends on the platform’s formula and the market inputs used. How Is Mark Price Calculated? There is no universal formula used by every crypto exchange. A common framework has two components: 1. Establish an Index Price The platform combines spot prices from selected markets into an underlying reference. Weighting methods, constituent exchanges, outlier rules and fallback procedures vary. 2. Account for the Futures Market The platform then adjusts or compares that reference with information from the derivatives market. For example, OKX currently states: Mark Price = Index Price + Moving-Average Basis The basis is calculated from the difference between the contract’s mid-price and the Index Price. MEXC uses a different approach for its Fair Price: Fair Price = Median (Funding Rate Premium, Mid-Price Basis Fair Price, Last Price) This is important because the three inputs are not simply added together. MEXC takes their median. Its Mid-Price Basis Fair Price itself uses the Index Price plus a moving average of the contract’s basis. These differences are why traders should check the methodology of the specific contract they use. Why Do Exchanges Use Mark Price for Liquidation? The main goal is to reduce the impact of brief distortions in a single futures market. Consider this example: BTC perpetual futures are trading near $100,000 . . A temporary liquidity gap produces a trade at $98,800 . . Major spot markets remain around $99,800 to $100,000. For example, traders monitoring the BTC/USDT trading pair can compare spot-market activity with futures prices to see whether a sharp move reflects the broader BTC market or a temporary derivatives-market distortion. If liquidation depended only on the $98,800 Last Price, leveraged long positions near that level could be affected by a short-lived local wick. A Mark Price system instead considers broader market information so that one abnormal trade has less influence on the platform’s risk calculations. Mark Price is used to reduce unnecessary liquidations and manipulation risk. Its liquidation framework also depends on maintenance margin, collateral and PnL rather than Last Price alone. Mark Price does not prevent liquidation when the wider market genuinely moves against a position. It only changes the reference used in the platform’s risk calculation. How Mark Price Affects Liquidation, PnL and Execution These three functions should not be confused. Liquidation On platforms using Mark Price or Fair Price, liquidation risk is assessed using that reference as part of the margin system. A displayed liquidation price is therefore an estimate, not necessarily a permanently fixed level. Changes in margin, maintenance-margin requirements, funding, position size or account mode can change liquidation conditions. Unrealized PnL The convention differs by exchange. Binance states that Mark Price is used for unrealized PnL and liquidation calculations. MEXC likewise states that Fair Price is used to determine unrealized PnL and reduce unnecessary liquidation risk. Bybit’s documentation, updated August 12, 2026, says unrealized PnL is typically displayed using Last Traded Price. Users can also view PnL using Mark Price, while Mark Price is used for liquidation. This is why identical-looking PnL fields can behave differently across platforms. Order Execution Mark Price is not normally where a market order automatically executes. Orders interact with the actual order book. The price you receive depends on available bids and asks, liquidity and slippage. This distinction between a reference price and an execution price is fundamental. Why Can Mark Price and Last Price Differ? The gap usually comes from one or more of four factors: Sudden volatility: futures trades may move faster than the underlying spot index. Thin liquidity: one large trade can push Last Price away from the broader market. Futures premium or discount: derivatives can trade above or below the underlying spot market. Calculation methodology: Mark Price may incorporate basis, funding or averaged market data. Mini Example Assume a leveraged long position shows: Price Type Value Last Price $100,250 Index Price $100,000 Mark Price $100,040 Estimated Liquidation Price $99,900 Looking only at Last Price suggests a $350 gap above the displayed liquidation level. Using Mark Price, the gap is only $140. The example illustrates why monitoring Last Price alone can give an incomplete picture of liquidation risk. The actual liquidation process still depends on the platform’s maintenance-margin and account calculations. Mark Price vs Last Price vs Index Price: Which Should You Watch? Each price answers a different question. Trader’s Question Price or Data to Check What price did the latest futures trade execute at? Last Price What does the broader spot market indicate? Index Price How close is my position to liquidation? Mark/Fair Price plus margin metrics Where could my market order execute? Order book, with Last Price as context Is the futures contract trading at a premium or discount? Compare futures market price primarily with Index Price During volatile periods, monitoring all three can help explain whether a move comes from the wider market or from conditions on a specific derivatives venue. Common Mark Price Mistakes The most important mistakes to avoid are: Assuming the latest candle automatically determines liquidation. Treating Mark Price and Index Price as the same thing. Assuming Mark Price represents an actual executed trade. Assuming every exchange calculates or displays unrealized PnL the same way. Watching a displayed liquidation price without considering margin, maintenance requirements and account conditions. Before using leverage, check which price your platform uses for both liquidation and PnL. Fine Thought Last Price tells you where the latest futures trade occurred, Index Price provides a broader underlying-market benchmark, and Mark Price provides a calculated reference for futures risk management. The distinction becomes especially important when leverage is involved. A position’s liquidation risk may be based on a reference price that differs from the most visible Last Price on the chart. Before opening a leveraged futures position, check the platform’s current Mark Price methodology, PnL rules, margin system and liquidation conditions.

Deposits of Coinbase-issued equities in Aave's new Base market doubled in a week, with Meta and Nvidia making up half the total Wall Street’s favorite tech names are quietly becoming DeFi collateral. Aave v4 …

Deposits of Coinbase-issued equities in Aave's new Base market doubled in a week, with Meta and Nvidia making up half the total Wall Street’s favorite tech names are quietly becoming DeFi collateral. Aave v4 on Base now holds $4.7 million in Coinbase-issued tokenized stocks, and that figure doubled over the past week. Half of those deposits sit in just two names: Meta and Nvidia. Even onchain, the market can’t stop chasing AI-adjacent megacaps. What’s actually happening on Aave’s Equities Hub Aave V4 launched a dedicated Equities Hub on Base, Coinbase’s Ethereum layer-2 network, on September 25, 2026. Eligible users outside the US can deposit tokenized versions of major stocks and borrow USDC against them. Seven stocks are supported at launch, all issued by Coinbase with a “c” suffix: Advertisement Apple (AAPLc) Amazon (AMZNc) Alphabet (GOOGLc) Meta (METAc) Microsoft (MSFTc) Nvidia (NVDAc) Tesla (TSLAc) Users can mix and match these tokens as collateral. Someone holding a little Apple and a little Tesla can stack both into a single position rather than picking one. The hub runs as an isolated market. For now, USDC is the only asset available to borrow. Whatever happens in this corner of Aave is meant to stay in this corner of Aave, walled off from the protocol’s broader lending pools. AI, tech, and the markets they move—in one daily briefing. Daily. Free. Join 34,000+ readers across crypto, finance, and policy. Email address Subscribe free We respect your privacy. Unsubscribe anytime. The guardrails: collateral factors and caps Risk firm LlamaRisk set the initial parameters, and they lean conservative. Collateral factors for the tokenized stocks range between 65% and 79%. A quick translation: if you deposit $100 worth of tokenized shares, you can borrow somewhere between $65 and $79 in USDC, depending on which stock you’re using. The aggregate collateral cap sits near $29 million. The USDC supply cap is $32 million, and the borrow cap is $21 million. Pricing comes from Chainlink’s onchain feeds, which deliver valuations on a 24/5 schedule. The Aave market itself runs 24/7, pausing only during official corporate actions like stock splits or dividends that change what a share actually represents. What this means for traders, lenders and DeFi With half of deposits in METAc and NVDAc, early users appear to be bringing their highest-conviction tech positions onchain. By October 1, total deposits reached $9 million, reflecting approximately $3.8 million in inflows during the week following launch. The scheduling mismatch is a specific risk to watch. Aave’s market never closes, but Chainlink’s price feeds run 24/5, in step with traditional equity trading. That means weekends are a quiet zone where stock prices can’t update. If major news breaks on a Saturday, the onchain price won’t reflect it until feeds resume, and borrowers sitting near their limits may find little room to react. LlamaRisk’s conservative collateral factors look designed partly with that kind of scenario in mind. The US exclusion is a reminder that tokenized equities still live in a regulatory gray zone. Stocks are securities, and lending against them onchain raises questions that American regulators have not fully settled. For now, the product is built for the rest of the world.

The House Financial Services chair argues agency rulemaking can be undone, while a statute cannot be reversed as easily Rep. French Hill has a message for crypto’s regulators: thanks for the effort, but it …

The House Financial Services chair argues agency rulemaking can be undone, while a statute cannot be reversed as easily Rep. French Hill has a message for crypto’s regulators: thanks for the effort, but it isn’t enough. The Arkansas Republican, who chairs the House Financial Services Committee, said in an October 7, 2026, interview that recent SEC and CFTC moves on digital assets fall short of what a real law would deliver. His preferred fix is still the Digital Asset Market Clarity Act, better known as the CLARITY Act, even after the Senate turned it away last month. What Hill actually said Hill’s core complaint is durability. Agencies can write rules, and agencies can rewrite them. A statute is a different animal. He framed congressional action as a “permanent law change”, in contrast to agency policies that can be reversed. Hill argued that regulatory policy is exposed to whoever controls the executive branch. A new administration can bring new priorities, and with them, new rules for the same industry. He also pointed out that the SEC and CFTC have floated different approaches depending on which part of the digital asset market is involved. Those proposals, Hill said, still don’t deliver the clarity and stability the industry needs to grow. Advertisement How the CLARITY Act stalled The bill hit a wall in the Senate in mid-September 2026. The vote was 49-50, with 60 votes needed to move it forward. Rather than wait on Congress, the regulators moved. SEC Chair Paul Atkins and CFTC Chair Michael Selig announced plans to pursue rulemaking using the authority their agencies already hold. The news moving money, markets, and the world—before your day starts. Daily. Free. Join 34,000+ readers across crypto, finance, and policy. Email address Subscribe free We respect your privacy. Unsubscribe anytime. Hill still thinks Congress could get the CLARITY Act done in the upcoming lame-duck session, the window when outgoing lawmakers remain in office before a new Congress is seated. He remains hopeful about that path, despite the September defeat. The calendar is not generous. Only 22 Senate session days remain for the current Congress before it adjourns in 2027. A thin bench at the regulators There is another wrinkle in the agency-led approach. As of October 7, 2026, the SEC and CFTC had seven leadership vacancies between them. That count includes the recent resignation of SEC Commissioner Hester Peirce, who had long been one of the most prominent crypto voices at the agency. What this means for the market Without a statute like the CLARITY Act, firms may keep facing uncertainty over compliance and day-to-day operations. The segment-by-segment approach Hill described raises a separate concern. If the SEC and CFTC each handle different slices of the market in different ways, the result could be fragmented standards rather than one coherent rulebook. The watch list for the coming months is short. First, whether the CLARITY Act gets another shot in the lame-duck session, and whether supporters can find the votes they lacked in September. Second, how fast Atkins and Selig advance their rulemaking, and what those proposals look like for each market segment. Third, how the seven open leadership seats get filled. New appointees could accelerate, slow, or redirect the agencies’ crypto agenda. Congressional action remains a major variable for market sentiment and stability in digital asset trading. With 22 Senate session days left and a 49-50 loss still fresh, Hill is betting that lawmakers can deliver what regulators alone cannot.

Bitcoin's price dropped to $83,000 following a significant sell-off in the cryptocurrency market. More than $714 million worth of crypto positions were liquidated in the past 24 hours. Ethereum and several altcoins, including XRP …

Bitcoin's price dropped to $83,000 following a significant sell-off in the cryptocurrency market. More than $714 million worth of crypto positions were liquidated in the past 24 hours. Ethereum and several altcoins, including XRP and Dogecoin, also experience… Bitcoin price today fell to $83,000 level as the cryptocurrency market came under renewed selling pressure, with more than $714 million in crypto positions liquidated over the past 24 hours.BTC USD o…

Singapore's state investor flags shaky AI returns and stubborn prices even as it plans to expand its own AI exposure One of Asia’s largest investors just named the two things most likely to ruin …

Singapore's state investor flags shaky AI returns and stubborn prices even as it plans to expand its own AI exposure One of Asia’s largest investors just named the two things most likely to ruin next year for global markets: artificial intelligence and inflation. Temasek’s investment chief, Rohit Sipahimalani, singled out the pair as the biggest risks ahead. The firm still plans to put a much bigger share of its money into AI, which is a bit like a lifeguard warning about rip currents while wading further out. What Temasek is actually worried about Sipahimalani’s concern centers on the wave of US capital expenditure flowing into AI infrastructure. Companies are spending heavily on the data centers, chips and power needed to run AI systems. His warning is about what happens if that spending fails to pay off. If the expected returns don’t show up, he cautioned, valuation pressure could trigger significant market disruption. He made the comments in discussions around the Temasek Review 2026, the firm’s annual look at its portfolio. Interviews in July 2026 put US capex front and center as a risk factor. Inflation is the second half of the warning, and the two risks are more connected than they look. Investors have noted that AI-driven demand for semiconductors and energy is feeding into higher prices. Advertisement That inflation, in turn, pushes up investment hurdle rates. A hurdle rate is the minimum return a project needs to be worth funding. When it rises, fewer deals clear the bar. Cautious, but still buying None of this has Temasek heading for the exits. The firm aims to raise AI-related investments to 15% of its portfolio by 2031, up from approximately 6% today. AI, tech, and the markets they move—in one daily briefing. Daily. Free. Join 34,000+ readers across crypto, finance, and policy. Email address Subscribe free We respect your privacy. Unsubscribe anytime. That would mean more than doubling its relative exposure in roughly five years. The strategy pairs that expansion with an emphasis on keeping the overall portfolio resilient against rising inflation, interest rates and broader economic uncertainty. Temasek is doing this from a position of strength. Its net portfolio value hit a record S$518 billion (approximately US$401 billion) as of March 31, 2026, a 10.5% increase from the year before. Singapore’s regulators are sounding similar alarms Temasek is not the only voice in Singapore raising these flags. The Monetary Authority of Singapore, the city-state’s central bank, has issued its own warnings in its Financial Stability Review. MAS pointed to inflated capital costs stemming from AI spending. It also warned of potentially severe fiscal pressures if an AI downturn were to hit. The central bank put numbers to that concern. A stress test it conducted on September 22, 2026, found that 32% of Singapore-listed firms could face significant revenue shocks in a severe AI downturn. A recent report from the ASEAN+3 Macroeconomic Research Office, known as AMRO, highlighted that Asian economies are disproportionately exposed to the possible fallout once the AI boom cools. What this means for investors The most direct takeaway is that big institutional money is starting to separate AI enthusiasm from AI pricing. Temasek’s message is not “avoid AI.” It is “build a portfolio that survives if AI disappoints.” The inflation link deserves close watching. If AI demand keeps pushing up chip and energy costs, it could keep price pressures and interest rates elevated longer than markets would like. For Asian markets specifically, the warnings from MAS and AMRO raise the stakes. Economies tied closely to semiconductor supply chains and AI-related demand could feel a slowdown more sharply than others. The key things to track from here are whether US AI capex begins to produce visible returns, whether inflation in chips and energy eases or persists, and how central banks respond.

The Bridgewater founder says hyperscalers turning to debt and a rising need for cash mirror the setups before 1929 and 2000 Ray Dalio thinks the AI trade is running out of runway. In an …

The Bridgewater founder says hyperscalers turning to debt and a rising need for cash mirror the setups before 1929 and 2000 Ray Dalio thinks the AI trade is running out of runway. In an October 6, 2026 interview with Bloomberg Television, the Bridgewater Associates founder said the AI market looks like a bubble getting close to its burst point. His reasoning is not that the technology is overhyped. It is that the money behind it is getting more expensive, and some of the biggest spenders now need cash. What Dalio actually said Dalio described current conditions as showing “classic signs” of a bubble. He put the AI boom alongside two famous precedents: the 1929 stock market crash and the dot-com bubble of 2000. The core of his argument centers on the hyperscalers. That is the industry term for the giant firms pouring capital into AI infrastructure at scale. According to Dalio, these companies are increasingly turning to debt financing rather than raising equity. He tied that shift to rising borrowing costs and growing cash needs. The three signals he is watching Dalio laid out specific indicators that, in his view, tend to show up when a bubble starts to unwind. Advertisement The first is forced selling to raise cash. When investors or companies must liquidate assets to cover obligations, prices fall because of need, not because of changed opinions. The second is a surge in new stock supply. Heavy issuance floods the market with shares, which can dilute existing holders and soak up the demand that was lifting prices. AI, tech, and the markets they move—in one daily briefing. Daily. Free. Join 34,000+ readers across crypto, finance, and policy. Email address Subscribe free We respect your privacy. Unsubscribe anytime. The third is rising retail leverage. When everyday investors borrow more to buy into a trend, any downturn gets amplified as margin calls force them out. The bigger debt picture Dalio’s AI warning sits inside a broader concern about borrowing across the US economy. He pointed to federal debt topping $40 trillion as of August 2026. He has suggested the financial system could face a significant debt crisis within the next three years if current trends continue. The connection to AI is straightforward. Higher rates make government debt costlier to carry, and they make corporate debt costlier too, including the borrowing hyperscalers now lean on. Innovation versus valuation Dalio was careful to separate two ideas that often get blended together. He acknowledged that artificial intelligence itself could be transformative. What he questioned is whether AI stock prices reflect the economic productivity the technology will actually generate. In his view, the rush of investment into AI-related sectors may be running ahead of that reality. What this means for investors The most actionable part of Dalio’s warning is the financing shift. When major AI spenders move from equity to debt, they take on fixed obligations that must be serviced regardless of how AI revenue develops. If rates keep rising, those interest costs could pressure margins. That raises the risk of financial strain, and potentially the kind of forced selling Dalio flagged as an unwinding signal. For investors, his three indicators offer something more concrete than vibes. Watching for heavy share issuance, signs of forced liquidation, and climbing retail margin debt gives a checklist rather than a gut feeling.

Coinbase has announced the acquisition of Deribit, a prominent crypto-derivatives platform, and integrated its products into the Coinbase ecosystem. This move is part of Coinbase’s strategy to enhance its derivatives offerings, making them available …

Coinbase has announced the acquisition of Deribit, a prominent crypto-derivatives platform, and integrated its products into the Coinbase ecosystem. This move is part of Coinbase’s strategy to enhance its derivatives offerings, making them available to U.S. institutional clients through Coinbase Prime. As part of the integration, Coinbase has launched new derivatives products and tools, including USDC-settled contracts and a new matching engine. This development marks a significant consolidation of Coinbase’s spot, futures, perpetuals, and options into a unified platform, potentially bolstering market activity and interest in the Ethereum ecosystem. Key Takeaways Coinbase’s acquisition of Deribit and integration of its products appears to strengthen its derivatives platform. The development suggests a potential boost to Ethereum-related market activity, consistent with increased interest in Ethereum reaching $10,000. Market pricing implies a moderate increase in the probability of Ethereum reaching higher price thresholds by the end of 2026. What to Watch Watch for further announcements from Coinbase regarding new derivatives products or tools, as these could impact market perceptions. Key indicators to monitor include any changes in derivatives volumes on the Coinbase platform and reactions from major financial institutions like BlackRock and Fidelity. Developments in regulatory environments for crypto derivatives in the U.S. could also influence market dynamics, potentially affecting Ethereum price predictions.

The asset manager's digital assets director says tokenized fund shares could be posted as collateral without redemption or multi-day cash waits Collateral has always moved at the speed of paperwork. BlackRock thinks it can …

The asset manager's digital assets director says tokenized fund shares could be posted as collateral without redemption or multi-day cash waits Collateral has always moved at the speed of paperwork. BlackRock thinks it can move at the speed of a blockchain. Nikhil Sharma, BlackRock’s Director of Digital Assets, says tokenized money market funds could be posted as collateral directly. That means no redemption step and no waiting days for cash to settle. Sharma is on the speaker lineup for TOKEN2049 Singapore 2026, set for October 7-8 at Marina Bay Sands. His focus is tokenized money market funds (tMMFs) and stablecoins. What Sharma is actually proposing Under the traditional model, an institution holding money market fund shares typically redeems them for cash before using that cash as collateral. The redemption and settlement cycle can take days. During that window, capital sits idle. The tokenized version skips the detour. Fund shares exist as tokens, so the position itself can be transferred on-chain to a counterparty. The holder keeps earning the fund’s yield, and the counterparty receives an asset it can hold or move. Advertisement BlackRock’s broader aim is to let tMMF positions move directly on-chain without the usual cash settlement wait. BlackRock’s tokenization buildout Sharma took the digital assets director role in December 2025. The news moving money, markets, and the world—before your day starts. Daily. Free. Join 34,000+ readers across crypto, finance, and policy. Email address Subscribe free We respect your privacy. Unsubscribe anytime. On August 3, 2026, BlackRock launched two new tokenized funds: BSTBL and BRSRV. Both are targeting eligibility as reserve assets under the US GENIUS Act, the federal framework for stablecoins. BRSRV is built specifically for stablecoin reserves and digital-native institutions. Both products build on BUIDL, the tokenized fund BlackRock introduced in March 2024. BUIDL currently manages around $2.5 billion in assets. BlackRock has rolled out tokenized share classes for its European UCITS money market funds, which totaled $311 billion in assets under management as of June 30, 2026. For the operational layer, BlackRock is working with J.P. Morgan’s Kinexys platform. The collaboration supports 24/7 peer-to-peer transfer capabilities. Why it matters for institutions and crypto markets The collateral use case is arguably the most practical argument for tokenized funds yet. Capital that stays productive while it secures a trade is better. For institutions managing margin across multiple venues, removing a multi-day cash cycle could reduce the buffer they need to keep on hand. Crypto markets have a direct stake in this. Exchanges, lenders and derivatives venues all need collateral, and tokenized money market fund shares offer a yield-bearing, regulated alternative to holding idle stablecoins. If BSTBL and BRSRV achieve reserve asset status, stablecoin issuers gain a compliant home for their backing assets from the world’s largest asset manager. That eligibility is a target, not a done deal, and investors should treat it accordingly. Firms including Fidelity and Invesco are pursuing similar tokenization efforts. Partnerships and industry trials are part of how these products are reaching professional investors. There are open questions. Collateral only works if the receiving party accepts it, so adoption depends on exchanges, clearinghouses and counterparties agreeing to take tokenized fund shares. Legal treatment, custody arrangements and operational resilience during market stress will all get tested before this becomes routine.