raye

I write about financial markets, AI, crypto, and emerging economies.

Engineer by day. Analyst by night. L/S Trader by conviction.

[raye @rayemarkets]

"a journey of a thousand miles begins with a single step." - Lao Tzu

[raye @rayemarkets]

Microsoft is planning an artificial intelligence data-center pipeline that could add roughly 26 gigawatts of compute capacity, putting its expansion on a scale comparable with the entire global data-center construction pipeline from only a year ago. The buildout spans owned and leased facilities, multiple accelerator platforms including Nvidia, Advanced Micro Devices, and Microsoft’s own Maia chips, and increasingly aggressive power procurement through renewables, natural gas, and nuclear. The constraint is no longer simply access to graphics processing units. Grid connections, transformers, cooling systems, permitting, and financing are becoming equally important in determining how much announced capacity actually becomes operational.

Personal View

I think this strengthens the case that artificial intelligence infrastructure is becoming an energy and industrial-capacity cycle. Once campuses move into the gigawatt range, electricity availability starts determining how quickly compute can be deployed. Chips remain critical, but power generation, grid equipment, cooling, and electrical infrastructure become part of the same artificial intelligence supply chain.

The part I would watch most closely is the gap between announced gigawatts and energized gigawatts. Microsoft has already cancelled or deferred some projects while expanding elsewhere, which suggests that execution and power access are becoming more valuable than headline capacity targets. Companies that can secure power early and actually turn infrastructure into usable compute should have a meaningful advantage over those sitting on large development pipelines.

$MSFT

[raye @rayemarkets]

The U.S. Treasury expanded its long-duration debt buyback to as much as $6 billion, triple the previous $2 billion ceiling, but ultimately accepted only about $5.2 billion despite receiving more than $10 billion of offers. The operation was meant to support liquidity in older 10-year to 20-year Treasuries, yet the market treated it as a test of whether Treasury could contain the long-end selloff. Instead, the 10-year yield moved toward 5% and the 30-year reached around 5.37%, as rising oil prices, persistent inflation, heavy government borrowing and fiscal concerns continued to overwhelm the relatively small intervention.

Personal View

I think the important signal is not that Treasury missed the $6 billion maximum. It is that even a much larger buyback barely changed the direction of yields. Buybacks can improve liquidity and temporarily absorb some long-duration supply, but they cannot solve a market that increasingly wants higher compensation for inflation risk, fiscal deficits and enormous Treasury issuance.

I would watch the 10-year around 5% much more closely than the size of individual buybacks. If yields remain near that level despite Treasury intervention, it suggests the bond market is repricing the equilibrium cost of capital rather than reacting to a temporary liquidity problem. That becomes increasingly restrictive for equities, housing, corporate financing and government interest expense at the same time.

[raye @rayemarkets]

Stablecoin activity accelerated sharply after the GENIUS Act introduced a clearer United States regulatory framework for payment stablecoins. Adjusted quarterly transfer volume rose from roughly $2.3 trillion in Q2 2025 to more than $3 trillion in Q3, then climbed toward $4.5 trillion by Q1 2026 before easing slightly in Q2. The important point is not simply the growth rate, but the timing. Stablecoin usage was already rising before the legislation, while regulatory clarity appears to have coincided with a much faster expansion in real transaction volume, even after filtering out bots and artificially inflated activity.

Personal View

I see this as evidence that regulation is becoming an adoption catalyst rather than a constraint. Stablecoins are moving closer to financial infrastructure, especially for payments, settlement, treasury operations, and cross-border transfers. Once banks, fintech companies, exchanges, and corporations have clearer legal rules, the addressable market becomes much larger than crypto trading alone.

At the same time, I remain more bullish on the stablecoin sector than on every existing stablecoin company. Regulation legitimizes the product, but it also lowers the barrier for banks and major financial institutions to issue their own tokens. Transaction volume can keep growing rapidly while competition compresses economics for incumbent issuers. The sector can win even if today's dominant companies do not capture all of that growth.

ethereum:0xdac17f958d2ee523a2206206994597c13d831ec7 ethereum:0xa0b86991c6218b36c1d19d4a2e9eb0ce3606eb48

[raye @rayemarkets]

Nasdaq Ventures is investing $100 million in Payward, Kraken’s parent company, at a $21 billion valuation, while expanding a broader partnership around tokenized equities. The strategic piece is more important than the valuation. Nasdaq wants to bring regulated public equities onto blockchain infrastructure without losing issuer rights, compliance, governance, and market surveillance, while Kraken provides crypto-native distribution and settlement infrastructure. Payward already has meaningful traction through xStocks, and similar partnerships with Deutsche Börse and London Stock Exchange Group show that major traditional exchanges are increasingly experimenting with blockchain-based trading and settlement rather than leaving that market entirely to crypto platforms.

Personal View

I see this as another sign that tokenization is moving from a crypto narrative into actual financial-market infrastructure. Nasdaq is not trying to replace the existing equity market. It is trying to make the same regulated securities more portable, globally accessible, and potentially tradable for longer hours using blockchain rails. If that model works, the strongest advantage may sit with exchanges that can combine regulation, liquidity, custody, and on-chain settlement in one system.

I am also more bullish on the tokenization sector than on any single current platform. Nasdaq, Deutsche Börse, and London Stock Exchange Group working with Payward shows that traditional exchanges are willing to adopt crypto infrastructure when it improves distribution and settlement. That also means competition will get much harder for standalone crypto companies because regulated incumbents can bring existing issuers, institutional liquidity, and trusted market infrastructure directly on-chain.

$NDAQ

[raye @rayemarkets]

Meta’s artificial intelligence buildout is moving far beyond the compute needed for its own advertising and model workloads. Rothschild & Co Redburn estimates Meta’s total compute capacity could rise from roughly 2.5 gigawatts in 2025 to more than 14 gigawatts by 2030, with an increasingly large portion classified as excess capacity available for an artificial intelligence cloud. Wells Fargo estimates that monetizing just 1 gigawatt of this capacity at roughly $20 billion of revenue per gigawatt could generate $20 billion of revenue and about $14.6 billion of net income, equivalent to roughly $5.69 of earnings per share or 16.3 percent accretion to its fiscal year 2027 consensus estimate. The key shift is that Meta’s enormous artificial intelligence capital expenditure could eventually become not only an internal infrastructure cost, but also a separate commercial compute business.

Personal View

I think this changes how Meta’s artificial intelligence spending should be valued. If excess compute can actually be resold at attractive utilization rates, part of what currently looks like aggressive capital expenditure starts behaving more like investment into a revenue-producing infrastructure asset. That could significantly improve the return profile of Meta’s artificial intelligence buildout and reduce concerns that spending is running too far ahead of monetization.

The more interesting point is that Meta could gradually become both a major consumer and seller of artificial intelligence compute. It already has advertising demand to justify massive internal infrastructure, so any capacity built above peak requirements creates optionality. The upside is meaningful, but the 85 percent operating margin assumption shown here is extremely aggressive, so I would treat the $14.6 billion earnings estimate as an upside scenario rather than a base case.

$META

[raye @rayemarkets]

Saudi Arabia told OPEC that its crude production fell again in August to its lowest level since 1990, as the Iran conflict continued to disrupt exports through the Strait of Hormuz and force more barrels toward the Red Sea route through Yanbu. The weakness is spreading across OPEC, with August production falling about 640,000 barrels per day to 19.71 million barrels per day even though several members were supposed to increase supply. OPEC+ has now kept October production targets unchanged, but the bigger problem is that higher quotas cannot solve disrupted shipping and export infrastructure.

Personal View

I think this is much more important than Saudi Arabia simply producing less oil. Saudi Arabia normally acts as the global oil market's shock absorber because it has enormous spare production capacity, but spare capacity becomes much less useful when the bottleneck is moving the barrels out of the country. If Hormuz remains constrained while the Red Sea route also faces security risks, the market loses part of the supply buffer it normally relies on during geopolitical shocks.

What makes the setup even more interesting is that OPEC just cut 2026 oil-demand growth to only 380,000 barrels per day, yet Brent still surged above $107. This means the current oil rally is increasingly being driven by physical supply scarcity rather than strong global demand. As long as Saudi exports remain constrained, I think oil can stay structurally expensive even with softer economic growth, keeping energy inflation and interest-rate pressure elevated.

[raye @rayemarkets]

Tech profit margins have structurally separated from the rest of the United States market. The chart shows technology, media, and telecom margins rising from roughly 10 percent in the early 1980s to above 23 percent today, while the ex-tech market has only moved from around 6 percent to about 9 percent. The gap has widened sharply since 2020, reflecting the scalability of software and digital platforms, high incremental margins, stronger pricing power, and increasingly concentrated earnings among the largest technology companies.

Personal View

This helps explain why technology has kept commanding premium valuations. Investors are not only paying for faster revenue growth, they are paying for a business model where an additional dollar of revenue can generate far more profit than in capital-intensive sectors. As long as these margins remain durable, technology can continue producing stronger earnings growth even without extraordinary economic growth.

The risk is that current expectations now assume this profitability advantage persists. Artificial intelligence spending is pushing capital expenditure much higher across hyperscalers, semiconductors, data centers, and power infrastructure. If those investments fail to produce enough incremental revenue, the margin gap can compress. I would watch margins more closely than revenue growth because any sustained decline would challenge one of the strongest fundamental arguments behind technology's valuation premium.

$VGT $IYW

[raye @rayemarkets]

@TheProfInvestor God candle

[raye @rayemarkets]

Countries with larger funded pension and household savings pools tend to have deeper capital markets because they create a permanent domestic source of long-term capital. Apollo’s chart shows the United States, Denmark, Sweden, the Netherlands and Switzerland sitting far ahead of much of continental Europe in pension assets and household investments relative to gross domestic product, with generally deeper equity markets as well. Germany, France, Italy and Spain remain much weaker on both measures. Europe is now starting to respond through auto-enrollment and pension reforms, including Germany’s push to build a larger funded retirement system, as aging populations increase the need to shift part of retirement financing toward accumulated financial assets.

Personal View

I think Europe’s capital-market problem is partly a savings-allocation problem. European households save a lot, but too much capital still sits in deposits, insurance products and other relatively conservative vehicles instead of being continuously recycled into equities, infrastructure and private markets. A larger funded pension system creates a structural buyer that can remain invested through cycles, which is difficult to replicate through foreign capital alone.

If these reforms succeed, the impact could extend well beyond pensions. Larger recurring institutional inflows could gradually increase equity valuations, improve liquidity, deepen private capital markets and give European companies a stronger domestic funding base. Europe has spent years discussing how to close its investment gap with the United States. Building larger pools of long-duration capital may be one of the most practical ways to start.

[raye @rayemarkets]

Grab is reportedly in talks to acquire Atome Financial, the SoftBank-backed fintech behind Atome and Indonesia’s Kredit Pintar, after previously considering joining a funding round of more than $100 million. Atome has grown into a meaningful regional consumer-finance platform, with about $470 million of revenue in 2025, more than $4 billion of gross merchandise value, and pretax profitability for a second straight year. For Grab, the logic is straightforward: it already has payments, lending, banking exposure through Superbank, and a massive consumer and merchant network. Buying Atome would immediately deepen its position in consumer credit and give it more underwriting data, merchants, borrowers, and loan distribution across Southeast Asia.

Personal View

I see this as Grab moving further away from being just a mobility and delivery platform. The more financial products it controls, the more valuable its ecosystem becomes because payments, deposits, lending, and commerce can reinforce each other. Atome also gives Grab something that is difficult to build quickly: an established credit engine with real scale and operating history.

The part I would watch closely is credit quality. Consumer lending can look extremely attractive during expansion because revenue grows faster than the underlying risks become visible. If Grab can integrate Atome without weakening underwriting discipline, the financial-services business could become a much larger earnings driver. If regional consumers weaken, the acquisition would also make Grab more exposed to defaults and funding costs than its traditional platform business.

$GRAB

[raye @rayemarkets]

Epoch AI’s new AI Chip Users Explorer estimates how much frontier compute is actually being used by major AI labs rather than simply counting hardware ownership. By the end of 2025, OpenAI was estimated at roughly 1.74 million H100-equivalents, Google DeepMind at 1.58 million, Anthropic at 1.19 million, Meta Superintelligence Labs at 996,000, and SpaceX AI at 615,000. OpenAI alone increased its estimated compute capacity around 17 times from 2023 to 2025, while its 2025 usage was split roughly evenly between research and inference.

Personal View

What stands out to me is how concentrated frontier AI compute has become. OpenAI and Google DeepMind are already operating above one million H100-equivalents, with Anthropic not far behind. That makes compute access a real competitive advantage because only a very small group of labs can operate at this scale.

I also think the research and inference split is important. OpenAI was already using about half of its compute for inference in 2025, which shows that running AI products now requires almost as much compute as developing them. The more usage grows, the more these labs need to keep expanding compute just to support existing products, not only to train better models.

[raye @rayemarkets]

Ramp’s September AI Index shows the AI trade is shifting from adoption scarcity to price competition. Enterprise usage is still rising, but effective token pricing has fallen roughly 41% since March, while spending per employee among the heaviest AI users also declined in August. The market is getting more intelligence for fewer dollars, and enterprises are increasingly using cheaper models when frontier performance is not required.

Personal View

This is a warning for model-layer valuations. If pricing falls faster than usage expands, revenue growth can decouple from compute demand, leaving frontier model providers with heavy capital intensity, weaker incremental margins, and a harder path to earning returns on infrastructure spend. The market may have to stop rewarding raw usage growth and start pricing AI companies on monetization efficiency.

For public markets, cheaper intelligence is likely more bullish for downstream beneficiaries than for the model vendors themselves. Software, cloud infrastructure, semiconductors, power, data-center operators, and companies that can convert falling inference costs into higher margins or faster productivity gains may capture more of the economic surplus. The next phase of the AI trade could be less about who owns the best model and more about who captures the savings.

[raye @rayemarkets]

Indonesia is moving deeper into the artificial intelligence infrastructure race after Nvidia-backed Zankore secured a $3.1 billion senior loan to finance a large-scale graphics processing unit buildout, with Nvidia providing revenue-sharing and credit support to help make the project bankable. The financing will support roughly 100 megawatts initially, while Zankore is targeting up to 1 gigawatt of capacity across Southeast Asia. This is no longer just a technology story. Banks are increasingly treating artificial intelligence compute as infrastructure finance, while Indonesia is positioning itself as a regional hub for power-intensive digital infrastructure.

Personal View

I think the most important signal is the financing structure. Once banks are willing to lend billions against graphics processing unit infrastructure, with Nvidia helping absorb part of the commercial risk, artificial intelligence capital expenditure can scale much faster than corporate balance sheets alone would allow. That is bullish for the entire infrastructure chain, but it also means leverage is becoming embedded deeper inside the artificial intelligence cycle.

For Indonesia, this strengthens the case that artificial intelligence investment will increasingly flow into electricity, grids, cooling, fiber, and data-center infrastructure rather than just software. I would watch power availability more closely than graphics processing unit supply from here. Compute can be financed, but electricity, transmission capacity, and land are much harder to scale quickly.

$NVDA

[raye @rayemarkets]

Global capital is starting to return to Indonesia after months of pressure, with foreign investors rebuilding exposure to government bonds, the rupiah recovering from its June lows, and equities moving toward their first quarterly foreign inflow of 2026. The recovery has been supported by Bank Indonesia maintaining attractive real yields, active currency stabilization, stronger foreign exchange reserves, and improving confidence that the worst of the rupiah selloff has passed. But the rebound remains fragile because higher United States Treasury yields and elevated oil prices still create external pressure for an oil-importing economy.

Personal View

I see this less as a full Indonesia risk-on trade and more as the beginning of normalization. Foreign money is returning first to bonds because the yield premium is attractive and policy credibility has improved. Equities are still the second leg of the trade, and sustained foreign equity buying would be a stronger confirmation that global investors are becoming structurally bullish again.

The key variable now is whether Indonesia can keep attracting capital without continuously paying a very high yield premium. If United States yields stabilize and oil falls, the combination of carry, currency recovery, and improving domestic confidence could pull significantly more global capital into Indonesian assets. If those external pressures stay high, this recovery can remain real but narrow.

[raye @rayemarkets]

@cz_binance super bull run is coming

[raye @rayemarkets]

https://t.co/pifqOqGqf6

[Stellar @StellarOrg]

Today, @usbank announced a successful live pilot of USBDC, the bank's proprietary USD-backed stablecoin, on the Stellar network.

USBDC is one of the first bank-issued stablecoins deployed on a public blockchain.

Traditional finance meets innovation on Stellar.

[raye @rayemarkets]

U.S. Bank has moved stablecoins from experimentation toward actual banking infrastructure, completing a live cross-border transaction using USBDC, its proprietary dollar-backed token, on Stellar. More important than the token itself is the architecture: USBDC remains connected to the bank’s existing finance, compliance, risk and operational systems, while supporting minting, redemption, freezing and clawback functions. U.S. Bank is now evaluating the same infrastructure for 24/7 payments, liquidity management, collateral mobility and cross-border treasury operations. Stablecoins are increasingly looking less like a crypto-native product and more like the next settlement layer for commercial banking.

Personal View

I am bullish on stablecoins as a sector, but increasingly bearish on the idea that today’s dominant stablecoin companies automatically capture all of that growth. If major banks can issue their own regulated dollars directly onto public blockchains, the competitive moat shifts away from simply issuing the token toward distribution, liquidity, interoperability and reserve economics.

The interesting trade may eventually be the infrastructure underneath stablecoins rather than stablecoin issuers themselves. Banks already own the deposits, corporate treasury relationships, compliance infrastructure and payment distribution. If tokenized deposits and bank-issued stablecoins scale, crypto rails could become significantly larger while portions of the economics migrate back toward banks, blockchains and settlement infrastructure.

[*Walter Bloomberg @DeItaone]

U.S. BANK LAUNCHES ITS OWN STABLECOIN

U.S. Bank has successfully tested USBDC, its proprietary dollar-backed stablecoin, in a live cross-border payment between its North American and European operations.

The transaction ran on the Stellar blockchain while remaining integrated with the bank’s compliance and risk systems.

U.S. Bank is exploring stablecoin applications including 24/7 payments, liquidity management, cross-border treasury operations and faster settlement.

[raye @rayemarkets]

The Bank of Korea is warning that leveraged derivatives tied to Samsung Electronics and SK Hynix are becoming a market-stability risk as Korea’s artificial intelligence and memory rally becomes increasingly concentrated. These products mechanically add exposure when chip stocks rise and cut exposure when they fall, meaning momentum can reinforce itself in both directions. With Samsung and SK Hynix carrying enormous weight in Korea’s equity market, a semiconductor correction could trigger forced deleveraging that pushes prices far beyond what fundamentals alone would justify.

Personal View

The structural memory and artificial intelligence demand story remains compelling, but this is where a strong fundamental thesis can become a crowded financial trade. When leverage, index concentration, and derivative rebalancing all point in the same direction, price action becomes much more reflexive and much less connected to incremental changes in earnings.

What I would watch now is how the market behaves when momentum finally breaks. If leveraged exposure starts unwinding at the same time foreign flows turn defensive, Korea could see a much sharper equity correction even without a major deterioration in the semiconductor cycle. The fundamentals can stay strong while the positioning becomes the real source of volatility.

[raye @rayemarkets]

UK long-end borrowing costs have surged back to levels not seen in roughly three decades, with the 30-year gilt yield approaching 6% after spending much of the post-2008 era below 4%. The move reflects more than just Bank of England policy: persistent inflation risk, heavy government borrowing, rising gilt supply, and investor demand for a larger term premium are all repricing the long end. Even if policy rates eventually fall, the bond market is signaling that financing the UK government over several decades is becoming structurally more expensive.

Personal View

I think this is increasingly a fiscal credibility story rather than simply a rate-cycle story. When long yields keep rising while markets still expect eventual monetary easing, investors are effectively demanding more compensation for inflation uncertainty, debt issuance, and fiscal risk. That puts pressure on government interest costs, mortgage pricing, pensions, and long-duration assets.

The bigger risk is a feedback loop. Higher yields raise debt-servicing costs, which worsens fiscal arithmetic and forces either tighter spending, higher taxes, or even more borrowing. If the UK cannot convince investors that nominal growth and fiscal discipline can stabilize the debt path, long gilts can remain expensive even after the Bank of England starts cutting rates.

[raye @rayemarkets]

Japan’s bond market is slowly moving away from full Bank of Japan dominance, but the transition is far from complete. The Bank of Japan still holds roughly half of outstanding Japanese government bonds after years of quantitative easing and yield-curve control, while banks, insurers, pensions, and foreign investors remain much smaller marginal holders. As the Bank of Japan normalizes policy and reduces purchases, the market must increasingly clear through private demand, which means yields need to become attractive enough to pull domestic institutions back in.

Personal View

I think this is one of the biggest structural risks in Japan’s normalization cycle. The Bank of Japan can raise rates, but reducing its balance-sheet footprint is much harder because it has effectively become the anchor buyer of the Japanese government bond market. If private investors demand higher term premiums to absorb new issuance, long-end yields can rise faster than the policy rate itself.

Japan is basically testing whether a bond market shaped by decades of central-bank intervention can return to genuine price discovery without destabilizing government financing. With public debt already extremely large, even a gradual repricing in Japanese government bond yields can tighten financial conditions, pressure fiscal interest costs, and pull Japanese capital back home from overseas assets.