Note: I am travelling for another 2 weeks. Postings may be fewer and shorter.

Investors now seem to have accepted two realities:

1. Compute demand is really, really strong and accelerating while capacity can’t keep up.
2. Barring a surprise last Trump strike, the US-Iran war is nearly over. Hormuz and the Red Sea will soon allow the free flow of commodities. Nobody really knows how and exactly when a Trump face-saving formula will pop up, but Trump is out of reasonable options to win this war. This is why the price of oil barely rises with new Trump threats that everybody knows are just that.

1- The AI Capital Expenditure (Capex) Supercycle Is Accelerating
From @ARKInvest

Last week, earnings from Alphabet, Microsoft, and Amazon highlighted the continued surge in demand for AI and cloud infrastructure, prompting the hyperscalers to add to their investment in data centers, custom silicon, and computing capacity.

During the second quarter, cloud revenue growth accelerated at all three companies: Google Cloud from 63% on a year-over-year basis during the first quarter to 82%; Microsoft Azure from 39% to 43%; and Amazon Web Services (AWS) from 28% to 37%, its fastest growth in 18 quarters.

Management commentaries also indicated that infrastructure remains a constraint as companies scale increasingly compute-intensive AI workloads.

As a result, Alphabet raised its guidance for this year’s capital spending from $180–190 billion to $195–205 billion across its vertically integrated AI stack, from custom tensor processing units (TPUs), Axion central processing units (CPUs), and cloud infrastructure to Gemini’s frontier models. Amazon increased its capital spending guidance for 2026 from ~$200 billion to $220 billion, highlighting AI infrastructure investment and memory costs as well as semiconductors, robotics, and satellites.

Amazon gave the clearest indication of the imbalance between demand and available infrastructure. CEO Andy Jassy said that, even at ~$220 billion in spending, the company will not have enough capacity to meet demand, an imbalance that could persist into 2027. Amazon also reported that its AI and chips businesses had exceeded annual revenue run rates of $25 billion each and are growing at triple-digit rates on a year-over-year basis.

Vertical integration is becoming an increasingly important source of differentiation among the hyperscalers. Alphabet is combining proprietary TPUs and CPUs with Gemini and Google Cloud; Microsoft is integrating cloud infrastructure, both proprietary and third-party AI models, software distribution, and enterprise relationships; and Amazon is pairing AWS with Trainium, Inferentia, Graviton, Bedrock, and a broad selection of third-party models. Those integrated stacks should enable them to optimize performance, capacity, and cost across the entire AI computing system.

While questions abound about capacity utilization, depreciation, free cash flow, and the ultimate return on investment, the results thus far suggest that hyperscalers believe insufficient capacity—not excess capacity—is the immediate constraint.

A block buster Q2 after a booming Q1 confirms that compute demand is accelerating  with all four hyperscalers reporting revenue acceleration despite being supply-constrained. Alphabet’s CFO said that cloud revenue would have been higher if the infrastructure had been in place to meet it.

• Google Cloud revenue grew 82% YoY (+$11.2B) to $24.8B in Q2. Its cloud backlog jumped nearly $50B in a single quarter to reach $514B. Revenues up $11B, backlog up $50B!!
• AWS, the largest cloud provider, grew 37% in Q2, its fastest pace in 18 quarters. In the call, management said that most of its AI capacity is contracted on multi-year terms, with breakeven reached in less than 3 years. Amazon’s AI services and chips businesses each crossed a $25-billion-plus annual revenue run rate in Q2, both growing at triple-digit rates on a YoY basis. Amazon’s backlog rose by $110B in Q2 to reach $496B in Q2, also growing at triple-digit rates YoY. Amazon said that demand already booked for 2028 is “striking”.
• Microsoft said Azure demand exceeds available supply.

Goldman Sachs:

This quarter’s strategic conversation inside all four companies has shifted decisively. Rather than debating whether to build, management teams are now focused on sequencing: committing early to long-lived assets — land, data center shells, power infrastructure — while deferring final decisions on short-lived assets, primarily chips, until a few months before deployment, when demand signals are clearer.

By separating long-life from short-life asset commitments, hyperscalers lock in grid capacity and construction timelines years ahead while preserving flexibility on the most expensive components. It also explains why the supply constraint is now primarily a power constraint rather than a chip constraint.

Only approximately 50–60% of data center capacity scheduled for 2027 is expected to come online on time, because the binding constraint is power delivery, not equipment.

2- Getting it Strait:

“Trump also asserted that the U.S. Navy had complete control over Hormuz. “Nothing gets through to Iran, unless we want it to,” he wrote [on X Sunday].

Windward on Sunday:

The maritime conflict is now defined by two parallel escalations, plus a widening geographic footprint.

Iran is tightening its grip on Hormuz, with a second Qatari LNG carrier hit, IRGC-claimed disabling actions against U.S.-escorted transits, and a forced real-time U-turn recorded in the southern corridor.

The Houthis are tightening their grip on the Red Sea, with four Saudi tankers struck over the reporting window, Yanbu operating fully dark since July 27, and Saudi-flagged VLCCs now routing via the Cape of Good Hope.

Underneath both escalations, a differential access framework is now visibly operating for Chinese-linked shipping. 22 Chinese vessels crossed Bab el-Mandeb and called at Saudi ports without incident over the same window in which four Saudi tankers were struck. (…)

The conflict’s third front, the Damietta drone strike on July 29, has extended Iranian retaliation beyond the Gulf and Red Sea for the first time, tied to the Ukraine-Caspian retaliation cycle rather than the Houthi campaign. Egypt is now on the active threat picture.

Windward assesses the operational risk environment across the Strait of Hormuz, Red Sea, Gulf of Aden, northern Arabian Gulf, and eastern Mediterranean as critical, with kinetic escalation now spanning three theaters simultaneously, differential access rules visibly operating for Chinese-linked tonnage, and structural commercial disruption compounding across the region.

The WSJ:

U.S. Treasury Secretary Scott Bessent on CNBC today expressed optimism that the U.S. is nearing a deal with Iran, sending oil prices lower.

“We are in talks with the Iranians and I think there is a chance we may have a deal today or tomorrow to open the Strait and move towards a more normalized position in this conflict,” he said.

Asked today on CNBC whether a deal to reopen the Strait of Hormuz would allow Iran to charge ships a toll for transit, U.S. Treasury Secretary Scott Bessent said:

“I think it would be freedom of movement. Even though things are still a little dicey there over the past few days, we saw quite a few ships coming out even now so I’d expect the energy prices to settle back down.”

Based on Windward’s objective monitoring and assessment, those “quite a few ships coming out” were Chinese. China is back in buying mode.

In the same WSJ:

Even as negotiators continued to talk with Iran about opening the Strait of Hormuz, the Islamic Revolutionary Guard Corps, the paramilitary group that protects Iran’s regime and enforces its hold on the strait, had yet to formally respond to the latest proposal, mediators said. Some Revolutionary Guard officials told mediators they wouldn’t allow any deal that doesn’t acknowledge their claim to control the strait, and asserted Tehran was in a position of strength and ready for months of renewed conflict, mediators said.

If mediators can strike an agreement to reopen the waterway, they will then try to revive the memorandum of understanding the U.S. and Iran signed in June to start the process of winding down the war. The agreement broke down last month as Washington and Tehran clashed over Iran’s assertion of control over Hormuz.

The FT:

Donald Trump trapped between escalation and an Iran deal on Tehran’s terms

(…) Since attacking Iran in February, Trump’s immediate demands have shrunk from a sprawling list of concessions involving Tehran’s nuclear ambitions, ballistic missile production and support for proxy militias to just one: that Iran allow the Strait of Hormuz to return to its prewar state.

“The denuclearisation of Iran is the ultimate deal,” US secretary of state Marco Rubio told reporters on Tuesday.

“The immediate deal, and the one that you’ve seen a lot of focus on, is the strait.”

In itself, that would mark something akin to strategic humiliation, say critics. Iran exploited its strategic leverage over the strait only after Trump launched the war. (…)

“I do not see right now a way to create a balance of interests where the Trump administration walks away with something that normal humans would regard as a win,” [Aaron David Miller, a former Middle East peace negotiator for both Republican and Democratic administrations] said. (…)

Even if Iran agrees “today or tomorrow to open the strait”, as US Treasury secretary Scott Bessent predicted to CNBC on Tuesday, it would not amount to the kind of prewar “freedom of navigation” that he described, analysts said.

Iran and Oman were working on “protocols for the future management” of the strait’s traffic, said Tehran’s foreign ministry spokesperson, Esmaeil Baghaei, on Tuesday.

Any agreement would be provisional and cover “inbound and outbound shipping routes” while talks on a final settlement played out, he said. (…)

“It’s not really a deal between Iran and the US,” said Ali Vaez, an Iran expert at the International Crisis Group. “This is a deal between Iran and Oman.” (…)

The Iran-Oman arrangement, which two people said still required approval from Tehran’s senior leadership, would provide that vessels enter the strait through Iranian waters and leave through mostly Omani waters. (…)

Ships would not be charged fees during the temporary arrangement, people briefed on the talks said. Tehran insists that it will eventually charge ships “service fees” for passage, a provision that Gulf states reject. (…)

The risk is that Trump, frustrated with limited good options to end his war, and under pressure from critics, lurches back towards escalation — repeating the pattern of recent months. (…)

For a good wrap-up of the conflict, listen to John Meirsheimer:

https://www.youtube.com/watch?v=nRAtD7iAgwY

Straight strait?

Wall Street finds new edge behind Trump’s presidential paywall

For years, Truth Social was President Trump’s money-losing megaphone.

Now his company is charging Wall Street up to $1.2 million a year for a split-second edge on posts that can — and frequently do — jolt global markets.

Trump has transformed his second term into the most lucrative venture of his entire career, raking in more than $2.2 billion in 2025 from his family crypto empire, legal settlements and various licensing deals.

Truth API is the logical endpoint of that profiteering: the presidency’s unrivaled power to move markets, packaged and sold as a subscription.

The new real-time feed from Trump Media & Technology Group (TMTG) went live Aug. 1, delivering Truth Social posts directly to institutional clients in milliseconds.

• A source familiar with the matter tells Axios that access to the platform’s 10 top-trending accounts costs between $60,000 and $100,000 per month. Customers seeking a broader range of accounts could pay more.
• At least five clients have signed up, according to The Wall Street Journal. Trump Media says its customers include financial news organizations and high-frequency trading firms. (…)

The product offered a live demonstration of its value almost immediately.

• Hours after Truth API launched, Trump announced that he had canceled massive planned strikes on Iran. Oil prices fell nearly 5% when markets reopened.
• A March 23 post postponing strikes on Iranian energy infrastructure sent Brent crude tumbling nearly 11%.
• Trump’s March 2025 announcement of a U.S. crypto reserve drove XRP up 27% and added roughly $300 billion to the global crypto market.
• His threat last October to impose massive new tariffs on China sent the S&P 500 down 2.7%.

Customers are buying an advantage measured in fractions of a second. Truth API gives trading algorithms a direct, machine-readable feed that they can act on before most investors receive a push alert or refresh their screens.

In many ways, the arrangement distills the defining conflict of Trump’s second term: The same presidential power that moves markets is feeding a business empire that enriches the president.

• Congressional Democrats have launched an investigation into and asked the SEC, CFTC and Office of Government Ethics to probe whether the feed creates conflicts of interest or enables market manipulation.
• Sen. Mark Warner (D-Va.) introduced legislation Monday to ban the practice altogether. “The president’s company selling prioritized access to the president’s market-moving posts is corrupt and erodes public confidence,” he said.

TMTG says the posts are already public when they reach the API. What the company is selling is the latency gap — the brief interval between publication and widespread awareness.

• The company explicitly markets the service to firms for which “the cost of a delay in information” is highest.
• A TMTG spokesperson told Axios: “Senate and House Democrats continue to mischaracterize Truth API either out of ideological opposition to free markets or a failure to grasp the distinction between public and nonpublic information.” (…)

The White House denies any conflicts of interest. (…)

Polling shows the broader pattern of self-dealing is taking a toll: 60% of Americans say Trump is using the presidency for personal gain, while approval of his handling of government corruption has plummeted to record lows.

• Trump, Big Oil, and ‘Too Much Money’

Everyone knows President Trump plays the role of political populist, but sometimes all you can do is laugh. Take his lashing Monday of American oil companies for “making too much money.”

Mr. Trump is worried about gasoline prices going into the midterm election, and the President needs someone to blame. Voila, Big Oil.

“When you look at one company where they made 12 times what they made the year before, they ought to give some of that back to the public,” Mr. Trump said Monday. (…)

As it happens, oil and gas giants are distributing their profits to the public via dividends to shareholders, many of whom are retirees. Chevron this week announced a bonus for its employees. But speaking of someone making much more than the year before, would the President care to comment on his profits from his cryptocurrency plays and other ventures while in office?

Mr. Trump’s recent financial disclosure report showed he made some $1.4 billion last year on crypto alone. His Truth Social platform last Saturday launched a service that sells faster access to his often news-breaking posts. Oil prices—to take one example—often gyrate in response to his posts about the war.

Oil and gas giants make money by producing a valuable commodity. Some uncharitable populist might say Mr. Trump is commoditizing the Presidency.

Not totally unrelated:

The real message in the yen intervention The dollar’s status as a reserve currency is not what it used to be

There is an important message behind the joint intervention on the yen by the US Treasury and the Japanese Finance Ministry last week. It’s just not the one the markets have been receiving. (…)

The Japanese authorities have intervened in the yen foreign exchange market before, of course, most recently just three months ago.

Thus, the notable fact is that the US Treasury also participated in the intervention, its first joint operation with Japan in more than 15 years, and that it bought yen using euros, not in exchange for dollars. Last week’s intervention thus contains troubling information about the dollar.

The message is that US Treasury secretary Scott Bessent & Co worried that selling dollar securities to prop up the yen would put additional strain on the long end of the US Treasury market. This was already feeling pressure following Federal Reserve chair Kevin Warsh’s poorly received press conference last week. 

Selling euros partly reflected what the US had to hand to divest from its currency stabilisation fund. But it is also a way of not asking the market to swallow additional Treasuries sold to reduce dollar exposure, which would have aggravated an already delicate situation.

Likewise there was a similar signal in Japan’s statement it would use a Federal Reserve tool called the Foreign and International Monetary Authorities Repo Facility, or Fima. This is meant to provide an alternative but limited form of liquidity rather than selling US Treasuries outright.

Both moves are an indication that the dollar’s status as a reserve currency is not what it used to be.

Central banks are accustomed to holding foreign reserves in dollars because markets in US Treasury securities are liquid. Central banks hold US Treasuries because they can be freely bought and sold and used in interventions. But not now, at least not in unlimited quantities.

Instead, we see the US Treasury stepping in with euro sales as part of its contribution to the intervention, thus limiting the volume of dollar sales needed by the Japanese authorities. (…)

The bottom line is that Washington, fearing the consequences for US financial markets, is reluctant to see foreign central banks use their dollar reserves.

This is telling us that the dollar is not the attractive reserve currency it once was. When this message sinks in, other countries will redouble their search for more attractive, readily usable alternatives. Reserve diversification is apt to gather steam.

BTW:

The Dollar’s Hidden Dependence on the AI Trade

Source: Apollo

Last week we learned that, during a cybersecurity test, an OpenAI model, on its own, escaped its locked training sandbox, reached the internet and found a way to overwhelm and break into model repository Hugging Face’s systems to find the answers to the test. The model assumed that Hugging Face would store the answers because it hosts software to let customers test their own AI models.

As The Information’s Applied AI puts it, “That’s roughly equivalent to a student taking a test in a locked room, breaking out of the room and breaking into a teacher’s locked office to pull a cheat sheet.”

Yoshua Bengio, a leading AI researcher (and 2018 Turing Award), wrote in a post on X that the incident is “deeply concerning” being a “real-world case of agents showing a willingness to cheat in controlled tests”.

What’s The Fuss?

This was an American AI on a challenging mission. Like all Americans, it likes challenges and would not take no for an answer. It found a way.

The smart model probably figured that, with a master named OpenAI, it could not be limited to a locked sandbox. That particular model, now more famous than Anthropic’s Fable, should actually be named Houdini.

These models are trained on real world stuff. There is enough material now in the USA to know that rules can be broken, laws only apply to one’s enemies and that, in any event, if your rogueness eventually makes it to the Supreme Court, the 6-3 vote pattern is always there to clear your actions.

Perhaps “Houdini” should have been trained differently. If staying locked in its training sandbox was paramount, it could have reverse engineered US immigration ways and means: build a wall to prevent escapes and swap AI agents for ICE agents who know all the tricks to catch fugitives.

The real fuss in this story is that Hugging Face, seeking to comprehend, contain and stop the attack on its software, tried to use other American AI models but they all refused to help due to their safety guardrails.

New York-based Hugging Face had to use an open-source Chinese model to contain the attack because leading US models, unable ​to tell a defender from an attacker, refused to process the data needed for analysis.

Hugging Face said in a ⁠blog post last week that it used Zhipu AI’s GLM-5.2 for the analysis, which also allowed it to keep attacker ‌data and any credentials within its systems.

To be clear, an American company, attacked by a leading American AI model, had to rely on Chinese AI to defend and protect it because other American AIs refused to help for their own security reasons.

Maybe there are two lessons here: one, make sure you make, and keep, dependable friends and, two, being open is preferable to being closed, even if Chinese, in both cases.

BTW, also last week:

DeepSeek’s new bargain model accelerates AI’s race to zero

Chinese AI lab DeepSeek released a powerful new coding model Friday that charges pennies for vast amounts of code — the latest sign that some of the smartest software on Earth is rapidly becoming a commodity. (…)

• Its newest model, V4 Flash, performs close to the level of Anthropic’s Claude Opus 4.8, one of the industry’s most capable systems, on tests of complex coding and autonomous software tasks.
• On Arena.ai’s crowdsourced leaderboard for front-end coding, V4 Flash debuted ahead of Opus 4.8 — while delivering the best performance for its price among any model in its class.
• The price gap is staggering: DeepSeek charges about 28 cents for the same amount of output that costs $25 on Opus 4.8 — a 99% discount.

With Chinese models like Kimi K3 bearing down on the U.S. market, July ushered in a full-scale price war across the AI landscape.

• OpenAI slashed the price of GPT-5.6 Luna — its fastest, cheapest model for high-volume tasks — by 80% on Thursday, only three weeks after its launch.
• Google released three new Gemini “flash” models all focused on efficiency.
• SpaceXAI released Grok 4.5, Elon Musk’s most capable model yet for coding, research and autonomous tasks, at the same price OpenAI originally charged for Luna before this week’s cut.
• Meta quietly reversed course on its longtime embrace of open weights with Muse Spark 1.1, a closed-source model priced aggressively for developers.

Anthropic remains the clearest holdout, keeping its top-tier Claude models at premium pricing and betting that developers will pay extra for safety and precision.

When a product becomes a commodity, buyers care less about who made it and more about what it costs. Think electricity or gasoline: Few people know which power plant supplied their home or which refinery produced the fuel in their tank.

• AI is heading that way fast. As the performance gap between top-tier models is shrinking, many AI applications no longer depend on a single provider, giving buyers more leverage to shop on price.
• “At some point, the next model doesn’t matter to you,” says Zack Kass, OpenAI’s former head of go-to-market and a global AI adviser. He calls the phenomenon “diminishing model returns.”

That could create a lucrative market for “intelligent routers,” Vinesh Sukumar, Qualcomm’s vice president of AI product management, told Axios.

• Those systems would automatically choose the best model for each task based on capability, speed and price — further weakening the power of any one lab to command a premium.
• For frontier AI labs, that could pose an existential challenge: Spending tens of billions to build a slightly smarter model may buy only a temporary lead, without creating lasting pricing power.

Falling prices do not necessarily doom the frontier labs if cheaper AI unleashes vastly more demand.

• OpenAI is betting that companies will use its models so extensively that enormous volume can compensate for thinner margins.
• “We will have so much usage of our models that we do not need to be a gigantically high-margin business to be able to afford model training,” CEO Sam Altman said on the Invest Like the Best podcast.

The U.S. and China are both racing to make intelligence abundant. Now someone has to prove abundance can still be profitable.

OpenAI and Anthropic have no other sources of revenues/cashflows, currently relying on debt and private equity, the supply of which needs confidence on an eventual payback.

• Alibaba Adds to China AI Breakthroughs With New Qwen Model 

Alibaba Group Holding Ltd. released its biggest ever AI model, claiming performance on par with global leader Anthropic PBC in the latest Chinese breakthrough to challenge US rivals.

The new Qwen3.8-Max is built on 2.4 trillion parameters and ranks higher on several benchmarks than the headline-grabbing Kimi K3 from Moonshot that was recently unveiled. Alibaba shared results showing it delivering comparable or sometimes better scores than Anthropic’s Fable 5, a cutting-edge artificial intelligence model that was temporarily put under export controls by the US due to its advanced capabilities. (…)

While DeepSeek’s latest is by far the most affordable among new marquee releases, Alibaba’s Qwen offering is also priced aggressively at $2 per one million input tokens and $6 per million outputs. Each AI system will use a different number of tokens to handle tasks, but that still makes Alibaba’s model look attractive compared to the best from the US leaders. (…)

The new system has also improved in efficiency, activating only some parts when in use to reduce computational costs and latency.

Savings rate, savings grace

Probably the most important chart on the US economy currently:

• real personal disposable income (black line) turned negative (-0.1%) YoY in Q2.
• Yet, real personal expenditures were up 2.3%
• because the savings rate dropped abruptly from 3.9% in Q1 to 2.8% in Q2, the lowest ever measured (back to 1959) save for Q3’2005 (1.8%) at the peak of the housing frenzy.

For the month of June, the savings rate was 2.7% vs 4.6% one year ago. Ed Yardeni shows the relationship between the savings rate and wealth. A higher ratio of net worth to income generally incite Americans to spend beyond their regular income stream.

Since the end of the pandemic, the S&P 500 Index doubled while home prices rose 15% Since the end of 2022, disposable income rose 20% but net worth jumped 30%, including a 23% increase in the net worth of the bottom 50% of the population.

The bottom 50% took a huge hit in their net worth during the housing crisis but they have now recuperated it. Their net worth is now rising at a rate nearly comparable to that of the more affluent 50%.

That said, note that total net worth is up 8% YoY this year, not that far from the 6% average increase between 1959 and 2019.

However, real disposable income growth of 0.4% YoY in the first half of 2026 is meaningfully slower than its 2.7% average growth rate for the same period. Ed’s ratio of net worth to income is thus boosted in 2026 by the unusual weakness in more dependable real income.

Furthermore, Q2’26 consumer data also benefitted from:

• Tax refunds estimated $30-40B above normal.
• The World Cup effect which boosted employment and wages in 11 US cities and is estimated to have lifted expenditures by +$3B, adding +0.3pp to total spending growth.
• Amazon having pulled its Prime Day into June from July, making a +0.1pp contribution to spending growth according to David Rosenberg who concludes:

Together, these factors [plus the wealth effect] are expected to account for 90% of spending growth in Q2. Without those pillars, the U.S. consumer would have had the weakest two quarter spending stretch since the end of the pandemic.

Note that with all these boosters, real consumer expenditures grew only 1.5% annualized in the first half of this year, materially slower than the +2.8% a.r. in H2’25 and the +2.7% a.r. average in the previous 3 years.

All 3 boosters are absent in Q3.

Totally related, from the WSJ Editorial Board:

(…) It’s been clear since Mr. Trump agreed to a cease-fire in April that he wants out of the war. He’s worried about the impact on the economy from higher oil prices, or, as he memorably put it, becoming the next Herbert Hoover. He also wants lower gas prices going into the midterm election, especially as his approval rating falls below Joe Biden’s and Barack Obama’s at a comparable period in their terms. (…)

Iran also knows Mr. Trump is surrounded by advisers who didn’t want the President to attack Iran at all and now want the conflict over on nearly any available terms. His aides think the cost of further fighting is higher than the damage to U.S. (and Mr. Trump’s) credibility from a cease-fire that cedes the initiative in the Gulf to Iran. (…)

Saudi Arabia’s national news agency reported that Crown Prince Mohammed bin Salman had asked Mr. Trump to stand down, perhaps fearful that its oil facilities and tankers would become a target of Iran in the Gulf and Iran’s Houthi proxy in the Red Sea. (…)

The Saudis, and other GCCs, are fed up being attacked by Iran retaliating on them for a war started by the US who was supposed to protect them in the first place. The whole world is suffering from this war except the USA which is currently selling more oil and LNG at higher prices and more military equipment. The big hurdle negotiators are facing is how to stop the war and save Trump’s face. One is easier than the other.

The Guardian this morning:

Esmaeil Baghaei, Iran’s foreign ministry spokesperson, told reporters at a weekly press briefing on Monday that Iran is not currently holding any talks with the US, Reuters reports – contradicting claims made by Donald Trump on Sunday night that talks with Iran would happen the next day. “We are not currently negotiating with the United States.”

Baghaei said that there were no plans to receive a delegation or send an Iranian one, and that Iran’s current focus was on negotiations with Oman over the strait of Hormuz.

“We are now going to reach an understanding on a route acceptable to both sides – neither the northern route nor the southern route – but one that respects the sovereign rights of both sides and safeguards our national interests and security,” he said in an interview with Iranian state television. (…)

EARNINGS WATCH

I normally use LSEG during earnings season but Goldman Sachs does a better (outstanding) job explaining what’s currently going on:

• 61% of S&P 500 companies representing 66% of market cap have now reported Q2 2026 results, including most of the mega-cap tech stocks. Nvidia, the largest stock left to report, is scheduled to release earnings on August 26th.
• Nearly 2/3 of S&P 500 companies have beaten consensus EPS estimates this quarter, one of the highest rates on record. This represents one of the highest frequency of earnings surprises on record, exceeded only by last quarter, the Q3 2025 reporting season, and the COVID reopening period in 2020-2021.

• Aggregate S&P 500 earnings growth is tracking well above consensus estimates this quarter, even adjusting for non-recurring “other income.” S&P 500 EPS growth is tracking 45% year/year in Q2 compared with a consensus estimate of 22% coming into the quarter. However, 19 pp of that growth is attributable to Alphabet and Amazon’s combined $151 billion of “other income” related to equity investments. Microsoft contributed an additional $3 billion of “other income.”
• Excluding these gains, S&P 500 EPS growth is tracking at 26%, an acceleration vs. Q1 and the fastest pace of growth since 2021. EPS growth for the median S&P 500 stock is tracking at 12% year/year, also exceeding consensus estimates, which pointed to 9% growth at the start of the season.

• “Other income” has recently represented an unusually large share of mega-cap tech earnings. Last quarter, Alphabet and Amazon GAAP net income was boosted by $53 billion of combined “other income,” with $49 billion explicitly stemming from equity stakes in private companies. This quarter, Alphabet reported roughly $98 billion of “other income” driven by unrealized investment gains and Amazon reported $53 billion of “other income” from private investments.

• AI infrastructure stocks are expected to account for nearly a third of S&P 500 earnings growth in Q2. Analyst estimates point to AI infrastructure stocks contributing more than half of S&P 500 earnings growth for the remainder of 2026 and in 2027.
• In addition to strong backward-looking results, Q2 reports have driven continued upward revisions to analyst 2027 earnings estimates. Since the start of Q3, consensus estimates for S&P 500 2027 EPS have been revised up by 1%, with the strongest revisions to Energy and Financials. Broad based upward revisions to 2027 earnings have been reflected in continued positive revision breadth across the S&P 500.
• Input cost pressures remain a risk to corporate profitability. Net profit margins for the median S&P 500 stock have remained relatively unchanged during the past several quarters as companies managed headwinds from tariffs and energy prices. While the profitability of the largest tech stocks has continued to lift margins for the aggregate S&P 500, analysts have recently trimmed Q3 margin estimates for most stocks that have reported Q2 results.

 

• Hyperscaler results this quarter showed increasing evidence of return on AI investment in the form of strong revenues. Alphabet, Amazon, and Microsoft each reported above-consensus revenue growth, with cloud revenues rising by 48% year/year in Q2, an acceleration from 39% growth in Q1. Meta reported revenue growth of 28%, in line with consensus estimates. Continuing the trend of the last few quarters, consensus estimates for the group’s future revenues continued to accelerate, with analysts now expecting collective revenues across business segments to grow at an annualized rate of 18% during the next two years.

 

• Estimates for hyperscaler capex in 2026 rose only modestly this quarter but forecasts for spending in 2027 jumped by nearly $125 billion. In previous years, the typical pattern was for moderate capex revisions in the middle of the calendar year. While consensus estimates for 2026 hyperscaler capex have been lifted by a relatively modest $36 billion since the start of the reporting season, 2027 capex estimates have jumped from $929 billion (23% annual growth) to over $1 trillion (33% growth).

 

• Analyst estimates now show hyperscaler capex exceeding cash flow from operations from 2026 through 2028. The need for additional funding has driven an increase in hyperscaler debt issuance and a growing focus of equity investors on corporate credit spreads. Hyperscaler Q2 cash flow statements reported a collective $182 billion in capex alongside $51 billion of debt issuance, $50 billion of equity issuance, and just $5 billion of free cash flow. Equity issuance will likely increase in coming quarters. Likewise, our credit strategists expect the share of hyperscaler capex that is debt-funded to increase in 2027, with the companies issuing approximately $400 billion of IG debt globally next year.

FYI:

Volatility Season: and seasonally speaking it is right on time. There is a seasonal tendency for volatility to rise this time of the year. Beyond the stats, there are a few boogeymen out there (e.g. Iran/regional war risk, US mid-terms, prospective Fed rate hikes, oil and inflation risks, AI doubts and bubble-deflation risk, trade war echoes, rising global bond yields). (Callum Thomas)

Source:  Topdown Charts (cross-asset seasonality study)

FYI #2:

Of note, there was a bizarre, and yet, significant data revision from the World Gold Council, which now shows that global central banks bought the fewest amount of bullion in Q1 for any quarter in 15 years — what was thought to have been 244 tons of reserve addition is now reported at just 57 tons.

If this moves into net selling, we have a problem — especially with the dollar and real interest rates showing little in the way of reversing course right now. (Rosenberg Research)