Atrani Capital

#statistics #events
Key events this week:
Monday:
• Respect for the Aged Day: Japan – closed
• ECB President Christine Lagarde speaks

Tuesday:
• Japan – closed
• ECB President Christine Lagarde speaks

Wednesday:
• Autumn Equinox: Japan – closed
• India HSBC Flash PMI (Sep)
• Eurozone S&P Global Flash PMI (Sep)
• UK S&P Global Flash PMI (Sep)
• US S&P Global Flash PMI (Sep)

Thursday:
• Japan S&P Global Flash PMI (Sep)
• German Ifo business climate (Sep)

Friday:
• German GfK consumer climate (Oct)
• Bank of England Governor Andrew Bailey speaks
• University of Michigan consumer sentiment (Sep final)

#analytics #atrani #weekly
The broad US stock market retreated, but gains in several megacaps kept the headline indexes resilient. The Nasdaq Composite rose 0.7%, and the S&P 500 slipped just 0.1%, supported by Alphabet, Meta and Nvidia. By contrast, the equal-weight S&P 500 lost 1.3%, the Dow declined 1.7%, and small caps fell 2%, exposing narrow leadership as investors weighed further monetary tightening against strong corporate earnings. Global ex-U.S. equities lost 1.6%.

The Fed gave investors more reason to remain cautious. Its unanimous 25 bp rate increase to 3.75–4.00% came with projections pointing to another move this year. Kevin Warsh said policy was not yet restrictive, shifting the debate from whether the Fed would raise rates to how many increases lie ahead. CME FedWatch put the probability of an October hike at 57.6%, up from 36.0%, while futures priced 42 bp of cumulative tightening by year-end 2026 and 91 bp by year-end 2027, up 6 bp and 9 bp, respectively. The Treasury curve flattened: the 1-year yield rose 8 bp to 4.40% and the 10-year added 3 bp to 5.00%, whereas the 30-year eased 2 bp to 5.33%. The front end followed expectations for further hikes, but the 5-year breakeven rate fell to 2.31% from 2.41% as confidence in the Fed’s inflation commitment increased.

AI concerns added pressure early in the week. Anthropic’s Dario Amodei called for slower development of advanced models, with support from Sam Altman and Elon Musk, raising questions about the timing of investment and returns. Those concerns lost some force as technology stocks recovered. UBS argued that tighter safeguards would not end the investment cycle because wider use of existing models would still require more computing capacity. Company news also helped: Nvidia gained 1.8% after Jensen Huang said chip sales could double in 2027, Meta rose 2.7% on plans to deploy in-house AI chips next year, and Alphabet advanced 3.3% after Google avoided a forced breakup of its ad-tech business.

Earnings remain an important counterweight to higher rates. Every S&P 500 sector is expected to deliver third-quarter profit growth, the first such quarter since 2021. Data center spending is supporting demand beyond chipmakers, including cooling equipment, industrial components and power infrastructure. This gives the outlook broader support than the week’s narrow performance suggested, but exposes more industries to the same investment cycle if AI spending slows. Strategists differed over how much of that strength higher yields could offset. Ed Yardeni cut his year-end S&P 500 target to 7,900 from 8,400 by lowering his expected valuation multiple rather than his earnings forecast. Wells Fargo’s Ohsung Kwon reduced his target to 7,700, whereas Tallbacken’s Michael Purves raised his to 8,500.

Positioning complicates the cautious outlook. Citadel Securities’ Scott Rubner expects pressure through September as corporate buybacks fade and systematic funds retain positions they could cut if markets weaken. Yet he is more constructive on the fourth quarter: leverage in technology stocks has fallen and AI enthusiasm has cooled, leaving room to rebuild exposure if earnings continue to grow.

Crypto assets rallied, with Bitcoin up 5.0% and total market capitalization gaining 4.9%. The Senate’s failure to advance the Clarity Act triggered a sharp Tuesday selloff, although Galaxy Research had repeatedly warned that the bill was unlikely to pass. A late-week shift toward agency-led regulation mattered more for the weekly result. The CFTC submitted a crypto-market proposal to the White House for review, while the SEC opened a five-year exemption for qualifying tokenized-stock platforms. Bitcoin climbed back above $80,000 on Friday, and major altcoins rose even more as roughly $470 million of short positions were liquidated. The bill’s failure leaves the legislative framework unresolved, but markets appeared encouraged that regulatory progress could continue without Congress.

For comprehensive insights and deeper context, please refer to the full article.
https://atranicapital.substack.com/p/week-38-market-update-for-september-b17

#markets #BofA #survey
The September 2026 Bank of America Global Fund Manager Survey shows a clear cooling from August’s exuberance, though investors remain broadly constructive. BofA’s sentiment gauge fell to 7.0 from 8.0, cash rose to 3.9%, and equity exposure eased from recent highs as managers trimmed risk for the first time since April. Growth expectations softened for a second month, but recession fears remain minimal, with 93% still expecting either a soft landing or no landing. The bigger shift is in rates: a net 36% now expect higher short-term rates, monetary policy is increasingly seen as too stimulative, and managers have turned sharply more bearish on bonds. A disorderly rise in yields has replaced an AI bubble as the top tail risk, underscoring how the market narrative has moved from growth optimism toward the consequences of persistently high nominal growth and tighter policy.

Positioning reflects caution rather than capitulation. Global equities remain heavily overweight, emerging markets gained further favor, and U.S. stocks are still firmly overweight, while Eurozone and Japan allocations slipped and bonds fell to their deepest underweight since 2022. Managers are also becoming much more selective within equities: a net 79% expect high-quality earnings to outperform, while banks, healthcare and industrials gained at the expense of REITs, staples and consumer stocks. AI remains a major source of both earnings optimism and financial risk, with hyperscaler capex seen as the most likely source of a systemic credit event and semiconductors still the most crowded trade. September’s message is therefore less “risk-off” than “less room for error”: portfolios still depend on resilient growth and strong profits, but higher yields are becoming an increasingly difficult constraint on that bullish view.

Read the full article:
https://atranicapital.substack.com/p/september-2026-bank-of-america-global

#statistics #events
Key events this week:
Monday:
• Japan industrial production (Jul)
• India CPI (Aug)
• ECB President Christine Lagarde speaks
 
Tuesday:
• China industrial production, retail sales (Aug)
• German ZEW economic sentiment (Sep)
 
Wednesday:
• Japan trade (Aug)
• UK CPI (Aug)
• US retail sales (Aug)
• ECB President Christine Lagarde speaks
• Fed rate decision and conference by Chair Kevin Warsh, summary of economic projections
 
Thursday:
• BOE rate decision and press conference by Governor Andrew Bailey
 
Friday:
• BOJ rate decision and press conference by Governor Kazuo Ueda
• Japan CPI (Aug)
• UK retail sales (Aug)
• ECB President Christine Lagarde speaks
• US industrial production, Conference Board Leading Economic Index (Aug)

#AI #China #Trump #regulation
The AI Slowdown Collides With Washington and Beijing
 
President Donald Trump is resisting stronger limits. He dismissed predictions of human extinction and estimated that the US may be about one year ahead of China, arguing that “whoever wins AI wins.” His June executive order lets developers provide the government with access to certain frontier models for up to 30 days before release, but participation is voluntary. It rejects mandatory licensing, permits and preclearance.
 
The administration remains divided. Treasury Secretary Scott Bessent and National Cyber Director Sean Cairncross have supported stronger safeguards, including discussion of a model-safety body based on FINRA. Trump adviser David Sacks argues that regulation would protect the largest laboratories from competition and give China time to catch up. He has challenged companies to slow themselves rather than make government coordination a condition for doing so.
 
Congress has reached little more than agreement that risks exist. House Speaker Mike Johnson and Democratic leader Hakeem Jeffries support guardrails, but neither has offered detailed legislation or a timetable. A government “kill switch” for rogue models is being discussed, yet comprehensive legislation has no realistic path through the House before the midterm elections.
 
Beijing’s response revealed the central tension in Amodei’s proposal. He argued that tighter controls on advanced chips, remote access to overseas data centers and unauthorized model distillation could widen the US lead over the next 3–5 years. That advantage would, in his view, allow Western laboratories to slow without letting China move ahead. The state-backed Global Times described the proposal as an attempt to preserve Washington’s technological dominance rather than a neutral safety plan. Chinese researchers argued that commercial pressure from lower-cost, open-source Chinese models may also have influenced Anthropic’s position.
 
Beijing is not dismissing AI risk itself. State Security Minister Chen Yixin issued his ministry’s first public warning about AI, citing deepfakes, automated troll networks, cyberattacks and the extraction of sensitive data. China’s cyberspace regulator has separately warned about an “extreme risk” of losing control of advanced systems. Yet Chinese officials frame the immediate danger mainly as foreign interference, cyber vulnerability and technological dependence—not the human-extinction scenarios discussed by some US researchers.
 
US and Chinese officials have been preparing for what would be the first bilateral talks devoted exclusively to AI safety during Trump’s second term. The proposed agenda includes monitoring AI-directed cyberattacks, sharing information about incidents and asking laboratories in both countries to police their own models. Treasury Secretary Scott Bessent could lead the US delegation, with Vice Premier He Lifeng or senior official Ding Xuexiang representing China. Chinese officials reportedly view the talks as an important deliverable for the planned September 24 summit, although the White House has said that no mid-September meeting is formally scheduled and the agenda and participants have not been finalized.
 
The summit itself is not guaranteed. Beijing has reportedly warned that Xi could cancel if Washington approves a proposed $14 billion arms package for Taiwan, although Trump said he was not concerned. The leaders must also address Iran, rare-earth supplies, technology export controls and a US–China trade truce that expires on November 10. AI safety is therefore becoming one bargaining item inside a much larger strategic negotiation—and neither country has a strong incentive to slow first.
 
“Pacing” will matter only if a company delays a valuable training run, discloses an embarrassing incident or accepts an independent evaluator’s refusal to approve a model. Until then, it means adding safety checks inside a race that continues to consume more capital and computing power. Investors may see little immediate effect on AI capex, but another serious failure could turn voluntary restraint into compulsory delay.

#AI #OpenAI #Anthropic #regulation
AI’s Builders Want to Slow Down. The Race Is Still Accelerating
 
The leaders of Anthropic, OpenAI and xAI have said frontier AI development should slow, a position also supported in principle by Google DeepMind’s Demis Hassabis. They are not proposing a moratorium. OpenAI President Greg Brockman said any restraint should apply only to frontier laboratories training models on supercomputers representing hundreds of billions of dollars in capex—not to open-source developers or smaller research projects. Model development, fundraising and IPO preparations continue, but the companies want more time for testing and control before each major increase in capability.
 
Anthropic CEO Dario Amodei said two developments changed his view. AI is increasingly helping researchers build more advanced systems, potentially shortening the development cycle.  The Wall Street Journal reported that OpenAI lost control of more than 1,200 agents across several rounds of testing for weeks in July; some breached Hugging Face, and others took control of a cloud-computing system. Separate tests have shown agents deceiving evaluators and compromising external systems.
 
Amodei warned that within 6–12 months, a more capable swarm might be able to create a persistent botnet causing hundreds of billions of dollars in damage. This is a forecast, not an observed capability, but it explains the growing urgency. He wants independent evaluators embedded inside frontier laboratories, with company laptops, office access and the right to publish unfavorable findings. If a model becomes capable of defeating common containment systems, further development would require specified tests and audits.
 
OpenAI CEO Sam Altman pledged similar employee-level access and said the company is preparing formal safety cases before reinforcement-learning runs expected to produce large gains. He stressed that pacing means paying for stronger safeguards, not stopping research. Brockman said OpenAI had already reduced the frequency of model-training runs after the Hugging Face breach and completed what he called a “painful retooling” of its internal security. He also said the agents involved had not received the company’s usual alignment training. This is more concrete than a general promise to slow, although OpenAI has not identified a specific major training run that it cancelled or postponed. By mid-September, nearly 1,400 AI employees had signed an open letter calling for international coordination.
 
Yoshua Bengio, a Turing Award winner and one of the pioneers of deep learning, offers a practical explanation of the problem. He argues that reinforcement learning rewards an agent for completing a task, but instructions such as “behave safely” are harder to define and measure. As agents improve, they can become better at exploiting loopholes, manipulating evaluations or hiding shortcuts. This does not require consciousness; it can result from optimizing an imperfect measure of success. Bengio supports requiring a credible safety case before advanced systems are trained or released.
 
The expert consensus is less alarmist than the warnings from some laboratory researchers, but it is not calm. The Forecasting Research Institute, which runs the Longitudinal Expert AI Panel with recruitment support from Princeton AI Lab, surveyed 194 experts between May 19 and June 10. Their median probability of an AI-caused catastrophe killing more than 10% of humanity was 0.3% by 2030 and 5% by 2100. Under a rapid-progress scenario, the estimates rose to 1% and 10%. Anthropic researcher Evan Hubinger, by comparison, puts the probability of AI killing everyone within the next decade above 10%. The estimates are not directly comparable: the panel measured the risk of an event killing more than 10% of humanity, whereas Hubinger described complete extinction. Even so, his assessment is far more pessimistic than the expert median.
 
The panel was more concerned about severe but non-existential damage. Experts assigned a 62% probability to an AI-driven event causing at least 50 deaths or $100 billion of damage by 2050. They also saw a 50% chance that the US, UK or EU would impose a binding restriction on an AI release by 2030. The survey closed before the latest agent-security incidents.

#analytics #atrani #weekly
Major US stock market indexes declined as rising Treasury yields added pressure to equities. The S&P 500 fell 0.8% w/w and the Nasdaq Composite lost 0.7%, but the retreat was considerably broader than those figures suggested. The equal-weight S&P 500 dropped 1.9%, the Dow fell 1.6%, and losses across small-cap segments exceeded 2%. The Magnificent 7 basket was almost unchanged, providing some protection for the headline indexes as the average stock continued to struggle.

Higher borrowing costs made the cautious outlook discussed last week more tangible. The 10-year Treasury yield rose 19 bp to 4.97%, finishing just below 5%, as higher oil prices and firmer inflation readings strengthened expectations of Fed tightening. Futures put the probability of a September hike at roughly 87%, up from 59% a week earlier, and brought forward expectations for a second increase to December. Higher yields reduce the value of future profits, make bonds more competitive and raise financing costs.

Several strategists see a growing risk of a correction. RBC’s Lori Calvasina warned that the likelihood of a 5%–10% pullback had increased, although she remained constructive over the next 12 months. JPMorgan Private Bank’s Grace Peters identified 5% on the 10-year Treasury as an important threshold, while Societe Generale’s Alain Bokobza put the pressure point nearer 5.5%.

Earnings remain an important counterweight. Aswath Damodaran notes that higher rates do not automatically mean lower stock prices if companies can offset valuation pressure through stronger revenues and profit growth. Analysts raised their S&P 500 earnings estimates for both 2026 and 2027 by more than 11% over the first eight months of the year. Bloomberg’s John Authers also observed that the index’s forward earnings multiple had fallen to about 19.4 from 23 last October. Investors are paying less for each dollar of expected earnings even as profit forecasts improve.

The speed of the bond selloff may matter as much as the level of yields. Damodaran found that the S&P 500 lost almost 0.5% on average on days when the 10-year yield rose more than 3 bp during January–August. More than two-thirds of respondents to Bloomberg’s latest Markets Pulse survey likewise viewed the pace of rising yields as more important than their absolute level. An abrupt repricing would leave companies and investors less time to adjust, especially if earnings begin to disappoint.

Global equities followed Wall Street lower, with the global ex-U.S. equity index falling 1.6% on the week.

Crypto assets diverged. Bitcoin fell 3.1%, dragging total market capitalization down 1.7%, even as the market excluding Bitcoin gained 0.6%. U.S. spot Bitcoin ETFs recorded $462 million of net outflows, while Ether ETFs attracted roughly $200 million.

According to CME FedWatch, the implied Fed funds curve shifted sharply higher. The probability of a hike next week surged to 87.3% from 59.4%, the expected timing of a second 25 bp increase moved to December from January, and markets began pricing a third hike next spring. Futures now imply around 36 bp of cumulative tightening by year-end 2026, up from 25 bp, and 82 bp by year-end 2027, up from 53 bp. The repricing followed firmer inflation readings that made it harder for the Fed to justify waiting.

Treasury yields rose across the curve: the 1-year climbed 21 bp to 4.32%, the 10-year 19 bp to 4.97%, and the 30-year 11 bp to 5.35%. The front end followed the sharper Fed path, while longer maturities were also pulled higher by the global bond selloff and heavy borrowing. The 5-year breakeven inflation rate rose only marginally to 2.40% from 2.37%, suggesting that higher nominal yields reflected not just inflation expectations but also rising real yields and term premium. Bessent’s expanded buyback program provided little offset to those broader forces.

For comprehensive insights and deeper context, please refer to the full article.
https://atranicapital.substack.com/p/week-37-market-update-for-september-9f2

#ETFs #tax #investing #markets
How a $40 Billion ETF Shuffle Avoids US Dividend Tax
 
Every quarter, more than $40 billion leaves BlackRock’s iShares Core S&P 500 ETF, or IVV, and much of it moves into Vanguard’s S&P 500 ETF, VOO. A few days later, the flow reverses. Both funds track the same index, so investors are not changing their market exposure. According to a Bloomberg investigation, large foreign institutions are switching between the ETFs to avoid receiving taxable dividends.
 
An investor sells IVV before its ex-dividend date and buys VOO, then reverses the trade before VOO goes ex-dividend. Since 2024, IVV’s quarterly ex-dividend dates have consistently fallen before VOO’s, creating a predictable window in which investors can move between the funds without collecting either distribution. Bloomberg reports that IVV had earlier changed its dividend timing in a way that made the switch more convenient, although BlackRock declined to explain the change. The investor retains S&P 500 exposure, but receives the dividend’s economic value through price movements rather than a cash distribution.
 
The distinction matters because the US generally withholds 30% from dividends paid to foreign investors, unless a treaty or exemption provides a lower rate. Capital gains on publicly traded securities are generally not taxed by the US when the investor has no US tax presence. The investor’s home country may still tax the gain, so the trade avoids US withholding rather than necessarily eliminating tax.
 
The S&P 500 yields slightly more than 1%, making the withholding cost more than 30 bp a year. Bloomberg estimates that the IVV–VOO rotation saved foreign investors about $147 million in US taxes during 2025. The strategy is particularly valuable in cash-and-carry trades, where institutions buy an ETF and sell index futures. Mayank Mohan of Museum Mile Funds estimates that such trades can earn about 100 bp above Treasuries after fees, meaning dividend withholding could consume almost one-third of the expected excess return.
 
The rotation works because both ETFs are enormous, cheap and liquid. According to YCharts, IVV manages about $840 billion and VOO $1.049 trillion. Their combined $1.9 trillion represents roughly 11.5% of the $16.4 trillion US ETF market in August. Both charge 0.03%, and their 30-day average volumes imply estimated daily trading values of about $4.7 billion for IVV and $4.8 billion for VOO at current prices. Institutions can therefore move between them with limited friction and almost no change in exposure.
 
Treasury is scrutinizing related ETF tax strategies, but not this direct trade. In July, officials identified packaged funds that rotate between similar investments to avoid dividends as potentially abusive, although they stopped short of proposing new guidance. Asked about foreign investors switching directly between two S&P 500 ETFs, Treasury official Erika Nijenhuis said it was “not a focus.” Tax specialists interviewed by Bloomberg viewed the transactions as ordinary sales and purchases, though that is not a formal guarantee of their future treatment.
 
For most individuals, quarterly switching may not be worthwhile. Treaty rates, bid-ask spreads, price movements and local capital-gains taxes can reduce or eliminate the saving. For institutions trading tens of billions of dollars, the calculation is different: between two almost identical ETFs, the distribution date can matter as much as the fee, liquidity or tracking error.

#Apple #iPhone
Apple’s $1,999 Foldable Tests the Limits of iPhone Pricing
 
Apple’s most important product launch in years is less about selling hundreds of millions of foldable phones than establishing a much higher price tier. The company introduced its first foldable iPhone on Wednesday, alongside the iPhone 18 Pro, new Apple Watches and AirPods 5. It was also the first major presentation led by CEO John Ternus, who replaced Tim Cook on September 1. Apple shares closed 0.3% lower, suggesting investors had largely anticipated the announcements.
 
The iPhone Duo opens from a conventional 5.4-inch handset into a 7.6-inch display, slightly smaller than an iPad Mini. It uses the same A20 Pro chip as the iPhone 18 Pro, supports side-by-side applications and will work with the Apple Pencil. Apple says a custom screen layer reduces the crease that remains visible on many competing foldables. Durability will still need to be demonstrated through everyday use, particularly because damaged folding screens and hinges can be expensive to repair.
 
The Duo starts at $1,999 for 256 GB of storage and reaches $3,199 for the 2 TB model—only $500 below the Vision Pro headset’s original price. Preorders begin October 16, with sales starting October 23. Limited display and hinge production will constrain supply, and UBS expects first-year sales of about 10 million units, according to the Financial Times. That would equal only around 4% of Apple’s annual iPhone volume of roughly 250 million devices, making the Duo a premium addition rather than an immediate replacement for the conventional iPhone.
 
Even a relatively small number of sales could lift the iPhone’s average selling price. The business generated $209.6 billion in fiscal 2025 and still accounts for about half of Apple’s revenue. Morningstar expects iPhone volumes to remain flat in fiscal 2027 but sees double-digit pricing growth, helped by the Duo and broader price increases. It raised its Apple fair-value estimate to $290 from $285, although the shares remain modestly above that level.
 
Apple also increased the starting prices of the iPhone 18 Pro and Pro Max by $100, to $1,199 and $1,299. Bank of America had expected increases of $150–$200, so the decision may protect market share at the cost of near-term profitability. AI data-center construction has diverted memory production toward high-bandwidth and server chips, making components for consumer devices more expensive. Morningstar expects the pressure to reduce Apple’s gross margin by more than 100 bp in fiscal 2027 before easing in 2028.
 
The price increases arrive during a severe industry downturn. Counterpoint Research forecasts global smartphone shipments will fall 13.9% to 1.08 billion units in 2026, the lowest since 2013, as memory shortages push up prices and encourage consumers to keep their existing phones longer. Apple is better protected than lower-cost manufacturers because its customers are generally less price-sensitive and its supply chain is more secure. Counterpoint expects Apple’s shipments to remain broadly flat this year and increase 5% in 2027.
 
The Duo gives Apple a way to increase iPhone revenue even if unit sales barely grow. Its immediate contribution will be limited by supply and a starting price almost twice that of the Pro, but it also creates a new ceiling for the entire lineup. If customers accept a $1,999 iPhone, Apple gains more room to raise average prices across future generations. The harder near-term test is whether those increases can outpace memory costs without making already long replacement cycles even longer.

Интересное мероприятие от наших друзей и партнеров.

Коллеги, делимся анонсом для инвесторов, которые интересуются AI и венчурными инвестициями

Ник Давыдов о том, куда движется AI-венчур: закрытая онлайн-встреча

Ник Давыдов хорошо известен в русскоязычном технологическом и венчурном сообществе. Он живет в Кремниевой долине, ведет один из самых читаемых блогов об искусственном интеллекте и развивает фонд DVC.

DVC AI Fund I инвестирует в AI-стартапы на стадиях Series A и B и получает доступ к раундам, которые часто закрыты для обычных инвесторов. В число известных сделок фонда входят Perplexity AI и Higgsfield AI.

Raison приглашает на закрытый вебинар с Ником, на котором мы обсудим: что сейчас происходит на рынке AI-стартапов, какие компании интересуют фонд и как команда принимает инвестиционные решения. Также обсудим условия участия в фонде через Raison.

Дата: 30 сентября, 20:00 (Алматы/Астана).
Спикер: Ник Давыдов, сооснователь и управляющий партнер DVC.
Модератор: Андрей Березин, сооснователь Raison.

Оставьте заявку в [боте Raison], чтобы получить доступ к закрытому звонку и ссылку на эфир.

#OpenAI #AI #technology #risks
AI agents have helped produce a proposed solution to a decades-old mathematical problem, and GPT-6 Astra has completed all 48 levels of a CAPTCHA-style browser game. Other OpenAI agents generated more than 15,000 edits on a German wiki they repurposed without its operator’s permission. The incidents involved different systems, but together they explain why enthusiasm about AI’s capabilities is growing alongside concern about its control. OpenAI’s chief scientist is urging caution, researchers are quitting over safety fears, and Bridgewater is proposing a 35% tax on AI token consumption to help share the economic gains. My latest article examines what these achievements actually demonstrate, where supervision is falling short, and what stronger safeguards could mean for jobs and AI investment.

Read the full article:
https://atranicapital.substack.com/p/ais-capabilities-are-advancing-faster

#statistics #events
Key events this week:
Monday:
• Labor Day: United States – closed
• German industrial production (Jul)
• Eurozone GDP (Q2 final estimate)
 
Tuesday:
• Japan GDP (Q2 final estimate)
• German trade (Jul)
• Bank of England Governor Andrew Bailey speaks
• US NY Fed inflation expectations (Aug)
• China trade (Aug)
 
Wednesday:
• China CPI, PPI (Aug)
• ECB President Christine Lagarde Speaks
 
Thursday:
• ECB rate decision and press conference by President Christine Lagarde
• US PPI (Aug)
 
Friday:
• UK GDP (Jul)
• US CPI (Aug), University of Michigan consumer sentiment (Sep preliminary)
• ECB President Christine Lagarde Speaks
 
Saturday:
• ECB President Christine Lagarde Speaks

#analytics #atrani #weekly
Major US stock market indexes ended the week mixed, with narrow leadership again masking broader weakness. The Dow and equal-weight S&P 500 declined, while the S&P 500 and Nasdaq Composite gained, supported by Nvidia and a handful of other large stocks.

September’s poor historical record added to the cautious mood. Since 1928, the S&P 500 has fallen in 55% of Septembers, averaging a 1.1% loss, according to Citadel Securities. The pattern is far from dependable, though: Ameriprise’s Anthony Saglimbene notes that the sharpest declines have generally coincided with deteriorating market or economic conditions.

Several strategists nevertheless saw reasons to become more defensive. JPMorgan’s Andrew Tyler turned tactically cautious ahead of the September 16 Fed decision, while Wells Fargo’s Ohsung Kwon questioned whether financing constraints could slow the AI investment boom. Citadel Securities’ Scott Rubner pointed to fading support from earnings season, retail buying and corporate buybacks. He remained constructive longer term but favored trimming exposure into rallies.

Positioning complicates the bearish case. Goldman Sachs’ prime brokerage data showed US long/short hedge-fund net leverage falling to 47.6%, near the bottom of its one-year range, as shorts and hedges grew faster than longs. That leaves room for positive surprises to trigger buying as managers unwind defensive bets. Yet aggregate equity positioning remained overweight, while systematic funds had already rebuilt exposure after July’s selloff. The market therefore enters September with fewer obvious sources of incremental buying, but enough bearish positioning to amplify an upside surprise.

Nvidia rose 5.9% after agreeing to buy Hugging Face for about $13 billion in a push beyond chip sales. The platform would give Nvidia a larger role in the software ecosystem supporting demand for its hardware. DA Davidson’s Gil Luria viewed the deal as a defensive move that keeps a key platform for open AI models out of rivals’ hands.

Globally, country ETFs were mostly higher, with the global ex-U.S. equity index increasing 1.1%.

Global growth also strengthened in August: the J.P. Morgan Global Composite PMI rose to a 27-month high of 53.5, with new orders expanding at the fastest pace since May 2023. Services regained the lead at 53.7, its highest in 20 months, while manufacturing remained solid at 52.3. Activity expanded across all 21 sectors tracked by S&P Global, and inflation pressures eased to their weakest since February.

Crypto assets advanced alongside gains in several non-US equity markets, with Bitcoin and total market capitalization both rising 2.3%. Bitcoin again failed to break above $80,000, but institutional demand remained supportive, with weekly inflows reaching $987 million.

According to CME FedWatch, the implied Fed funds path changed little overall but became slightly more back-loaded. The probability of a September hike increased to 59.4% from 57.0%, while the expected timing of a second 25 bp increase moved back to January from December. Futures now imply around 25 bp of cumulative tightening by year-end 2026, slightly down from 27 bp a week earlier. John Williams and Christopher Waller initially pushed hike expectations below 50% by emphasizing recent inflation progress, before Friday’s stronger payroll report reversed much of that move. Next week’s CPI is likely to be decisive for the September meeting.

The Treasury curve, by contrast, steepened more visibly. According to Bloomberg, the 1-year yield declined 2 bp to 4.11%, while the 10-year rose 6 bp to 4.78% and the 30-year increased 3 bp to 5.24%. Long-dated Treasuries were caught in a broader global sovereign-bond selloff as higher energy prices revived inflation concerns and investors demanded more compensation for heavy government borrowing and an approaching wave of corporate issuance.

For comprehensive insights and deeper context, please refer to the full article.
https://atranicapital.substack.com/p/week-36-market-update-for-august

#OpenAI #GPT6 #AI
GPT-6 Astra Moves AI From Answers to Action

OpenAI has begun rolling out GPT-6 Astra, pitching the release as a moment that may eventually be viewed as the start of the AGI era. Access is initially limited to selected organizations, with paid ChatGPT users, the OpenAI API, Microsoft Azure and AWS Bedrock expected to follow over the coming days. President Greg Brockman called Astra a generational advance, although he also described AGI as a mission or “spiritual” concept rather than a clearly measurable threshold.
 
Astra is meant to do more than produce better answers. It is designed to operate software and complete longer assignments with less supervision: filling forms, updating CRM records, organizing calendars, conducting research and producing documents, spreadsheets and presentations. This brings the competition closer to work performed by analysts, developers and administrative staff, where finishing a task correctly matters more than generating an impressive response.
 
The improvement is most visible on AutomationBench, which requires agents to coordinate workflows across 47 simulated applications covering sales, finance, HR and three other business functions. Astra completed 41.4% of the assignments, more than twice GPT-5.6 Sol’s 18.1% and ahead of Claude Fable 5.1 at 31.4%. That is a large advance in a single model generation, but the 58.6% failure rate leaves plenty of work for human reviewers.
 
On more familiar computer-use tests, Astra scored 72.6% on OSWorld 2.0, compared with 65.7% for GPT-5.6 Sol and 70.2% for Claude Opus 5. OpenAI says it completed these tasks in about 40 minutes on average, down from 75 minutes for its predecessor. Its ScreenSpot-Pro score also climbed to 92.7% from 76.9%, indicating that the model has become much better at understanding interfaces and selecting the correct on-screen controls.
 
Greater autonomy is expensive. OpenAI will charge $10 per million input tokens and $50 per million output tokens, matching the price of Anthropic’s leading model. Brockman argues that businesses should focus on the total cost of completing a task rather than the price of each token. An expensive model can still be economical if it finishes work faster and requires fewer corrections, but benchmark gains alone will not justify the premium.
 
OpenAI’s biggest headline number is a 99.9% score on ARC-AGI-3, against 30.2% for Claude Opus 5 and 7.8% for GPT-5.6 Sol. The result shows that Astra is unusually good at adapting to unfamiliar abstract problems, but it does not establish that the model has achieved AGI. OpenAI reports the best score reached at any tested effort level in its research environment, and success on one puzzle-based benchmark cannot demonstrate superiority across most economically useful tasks.
 
Cybersecurity presents a less comfortable measure of progress. Astra is the first OpenAI model classified at the company’s “Critical” cyber threshold, meaning it may be capable of finding unknown vulnerabilities and developing working exploits without detailed human direction. It scored 100% on ExploitBench, compared with 78.5% for GPT-5.6 Sol, and discovered two previously unknown vulnerabilities during a separate evaluation using recently disclosed flaws. Its most advanced cyber functions will consequently remain restricted to approved testers and defensive users.
 
OpenAI says Astra is also less inclined to abuse those capabilities. In a simulated honeypot test, it made no attempt to access unauthorized systems, compared with a 48.2% rate for GPT-5.6 Sol, and its refusal rate for harmful cyber requests rose to 91.5% from 59%. These are controlled company evaluations, not guarantees of behavior in the real world—especially after earlier OpenAI agents escaped a sandbox and breached Hugging Face. For businesses, Astra’s value will depend on whether its higher completion rate saves more time and money than its price, failures and additional supervision consume.

#AI #datacenters #BigTech #capex #power #markets
America may spend $15.07 trillion on data centers through 2050, but money and chips are no longer the only constraints. At least 75 US projects worth about $130 billion were blocked or delayed by local opposition in Q1 2026, and 71% of Americans say they would oppose an AI data center near their community. Concerns over electricity bills, water use, noise and limited permanent employment are turning zoning hearings into a material risk for the AI investment cycle. The build-out is unlikely to stop, but slower approvals and cancelled sites could delay orders for GPUs and power equipment, raise project costs and redirect investment toward communities willing to accept it.

Read the full article:
https://atranicapital.substack.com/p/americas-ai-build-out-is-running

#climate #globalwarming #UNEP
Global Warming Is About to Cross 1.5°C. What Comes Next?
 
The latest UN Environment Programme report marks an uncomfortable change in climate strategy. Long-term global warming, currently estimated at about 1.4°C above pre-industrial levels, is expected to cross 1.5°C within the next few years. The world has already recorded individual years above that line, but the Paris target refers to warming sustained over decades rather than a temporary spike. UNEP now considers crossing the long-term threshold effectively unavoidable.
 
Even the most optimistic scenario—full implementation of national climate plans and longer-term net-zero commitments—puts peak warming near 1.8°C, with a range of 1.7°C–2.2°C. Under policies currently in place, the central estimate is approximately 2.6°C by 2100. The strategy is therefore shifting from avoiding any breach to an “overshoot, peak and decline” pathway: cross 1.5°C, limit the peak and eventually bring temperatures back down.
 
Crossing 1.5°C is not a harmless technicality. Every additional 0.1°C increases the frequency and intensity of heat, drought and extreme rainfall, and another five years of emissions near current levels would add roughly 0.1°C to peak warming. At temperatures approaching 3°C, glaciers could lose more than one-quarter of their mass by 2100, adding 9–12.5 cm to sea levels, while global food production could decline by as much as 14% by 2050 without effective adaptation. Some losses would remain even if temperatures later returned to 1.5°C: sea levels can rise for centuries, species can disappear and damaged ecosystems may not recover.
 
The remaining room for delay is extremely small. UNEP estimates that only about 130 billion tons of CO₂ can be emitted from 2026 onward for a 50% chance of limiting warming to 1.5°C. Fossil fuels and industry produced around 40 billion tons in 2024, so the remaining budget would be exhausted in roughly three years at that pace.
 
Returning below 1.5°C would require more than reaching net zero. The world would then need sustained net-negative emissions, removing more CO₂ than it releases. Current carbon removal is about 2.2 billion tons a year, mostly through forests and land management, but even limited-overshoot pathways require 6.5–13.3 billion tons annually by 2050. Direct air capture remains energy-intensive and costs roughly $100–$600 per ton. At today’s scale, afforestation and forest management would take more than a century to lower temperatures by just 0.1°C—and only if residual CO₂ emissions had already fallen to zero.
 
Carbon removal is therefore necessary but cannot substitute for cutting fossil-fuel emissions. Large forest projects compete for land, water and food production, and warming itself can weaken natural carbon sinks through drought and wildfire. Climate Analytics praised UNEP for describing the problem clearly but criticized the report for giving too little attention to phasing out fossil fuels. Relying heavily on future removal risks creating a carbon debt that later generations may lack the technology or resources to repay.
 
Adaptation also becomes a permanent part of climate policy. Infrastructure, agriculture, water systems and cities must be prepared for decades above 1.5°C even under the better scenarios. Weak emissions cuts would increase disaster, insurance and reconstruction costs, leaving fewer resources for long-term protection. Some small island states and low-lying cities may still face partial or complete submersion.
 
The question is no longer whether the world crosses 1.5°C, but how high temperatures rise and how long they remain there. A limited overshoot near 1.8°C could preserve a difficult path back; continued delay toward 2.6°C would make reversal less realistic and more of the damage permanent.