Port Hedland Moves About $80M of BHP Iron Ore a Day—and Sunday’s Strike Still Didn’t Shut Shipping Down

Industrial action is real. “Australia’s iron-ore exports stopped” would still be the wrong headline.

Around 150 workers participated across the weekend in the first major industrial action at BHP’s Port Hedland operations in roughly 25 years. Saturday included a 24-hour restriction on ship-loading work, while about 100 workers joined Sunday’s stoppage.

Port Hedland is huge: it handled roughly 75% of iron-ore shipments from the Pilbara region last year, and BHP moves around $80 million of ore through the hub on a typical day.

But scale of infrastructure is not the same as scale of disruption.

BHP said vessels were still being loaded on Sunday, while other miners using Port Hedland—including Fortescue and Hancock Prospecting—were not affected by the BHP labour dispute. Talks are scheduled to resume on August 18.

For commodity markets, the useful trigger is actual lost tonnes, not the dollar value of everything that passes through the port.

A two-day disruption can often be recovered. Repeated action that creates a growing shipment backlog is a different story.

Not Financial Advice.

What would make you treat this as a real iron-ore supply shock: delayed vessels, lower weekly exports or another round of strikes?

Airbnb’s Best AI Number Wasn’t a Benchmark—It Was 16% Lower Support Cost per Booking

This is what AI ROI looks like when somebody finally attaches it to a business metric.

Airbnb says customer-support costs per booking fell about 16% year over year, helped partly by improvements to its AI assistant. The company also reported $3.61 billion in Q2 revenue, above the $3.57 billion LSEG estimate, and raised its full-year revenue-growth forecast to at least the mid-teens. Shares jumped 14% to a more-than-four-year high.

I like this AI metric more than another model leaderboard.

Support cost per booking directly touches unit economics. If the assistant resolves more problems without damaging customer satisfaction, every booking becomes slightly cheaper to service.

But cost reduction alone is not enough.

AI can save money by handling simple requests—or by frustrating users until they stop contacting support. The real proof would combine lower cost with stable resolution times, satisfaction scores and repeat bookings.

Airbnb is also expanding hotel supply, with room nights growing at three times the rate of homes, adding another test for whether AI support scales across a broader product mix.

Not Financial Advice.

Which AI metric would you trust most: lower operating cost, faster product launches or direct revenue generated by AI?

Unitree Is Pricing at $9.04B While Q1 Adjusted Profit Fell 52.6%

Robot demos are finally meeting public-market maths.

Chinese humanoid-robot maker Unitree priced its Shanghai IPO at 150.8 yuan per share, valuing the company at about 61 billion yuan ($9.04 billion). It plans to raise 6.1 billion yuan by issuing 40.45 million new shares on the STAR Market.

The growth story is real.

Unitree’s 2025 revenue more than quadrupled to 1.7 billion yuan, and humanoid-robot revenue reached 867.8 million yuan, overtaking its four-legged robot business.

Now the harder number: adjusted Q1 profit fell 52.6% as spending on R&D and marketing increased.

That is the public-market question humanoid robotics has avoided while the sector was mostly private.

Can impressive motion demos become repeatable, profitable industrial deployments?

Unitree also flagged US-market risk as restrictions on foreign-made advanced robots tighten. Existing products have approvals, but future models may face additional barriers.

The next metric I want is not another dancing robot video. It is revenue per deployed humanoid and evidence that customers reorder after the pilot stage.

Not Financial Advice.

Would you value Unitree on today’s robot sales—or the much larger labour market humanoids are supposed to attack eventually?

3.4% Is Wednesday’s CPI Forecast—Not Wednesday’s CPI

The market is already trading a number that has not been released yet.

The US Bureau of Labor Statistics will publish July CPI on August 12 at 8:30 a.m. ET. Economists polled by Reuters currently expect headline inflation around 3.4% year over year and core inflation around 2.5%. Those are consensus forecasts, not official data.

For this event, I would use KIA before release rather than after the candle has already moved.

My research prompt would create three scenarios:

Below consensus: what changes in Fed expectations?

Near consensus: which underlying components matter most?

Above consensus: which assets are most exposed to renewed tightening fears?

KIA supports online search and Deep Research, so it can organise expectations, previous inflation data and current macro commentary into one pre-event checklist. I would still use BLS as the final source for the actual number once it drops.

The weak July jobs report reduced confidence in further tightening. Hot inflation could challenge that repricing almost immediately.

Not Financial Advice.

Which CPI surprise would create more volatility now: 3.1% or 3.7%?

450% Assumed Leverage Met a Crowded Tech Trade—and Quant Funds Lost 5% in July

Being right about AI did not protect everyone from being positioned exactly like everyone else.

JPMorgan says global hedge funds gave back almost 3% in July as crowded technology positions unwound. Asia-Pacific stock pickers lost an average 9.4%, while global quantitative-equity funds fell 5%.

The leverage number deserves attention.

JPMorgan estimated quant strategies were running roughly 450% assumed leverage, the highest among the strategies it tracked. Hedge-fund borrowing overall remained near the upper end of its five-year range.

This is the danger of a crowded winner.

When everybody owns similar momentum trades, the first sell-off creates losses. Those losses reduce risk budgets. Reducing risk means selling more. Suddenly a fundamental correction becomes a positioning event.

Crypto has the same mechanism, only faster. Futures liquidations and collateral requirements can compress days of deleveraging into hours.

The useful signal is not just whether AI or crypto fundamentals remain bullish. Watch leverage + concentration + falling liquidity together.

Not Financial Advice.

Which breaks a crowded trade faster: bad fundamentals, rising volatility or leverage forcing good investors to sell anyway?

A FINRA Licence Wasn’t Enough: Bunq’s US Bank Charter Just Got Rejected

Fintech can scale software globally. Banking licences do not travel that easily.

The US Office of the Comptroller of the Currency rejected Dutch fintech Bunq’s application for a national bank charter, citing significant supervisory and compliance concerns.

The regulator questioned several practical pieces of the US plan: how Bunq would be capitalised, whether management had enough experience with unsecured credit cards, and whether the proposed bank could operate safely and reach profitability in the US market.

That is interesting because Bunq already secured a FINRA-approved broker-dealer licence in 2025 allowing it to offer investment products in the US. A securities licence did not automatically translate into approval to become a bank.

This is a useful framework for fintech—and crypto.

“Regulated” is not one universal status. Brokerage, banking, custody, payments and digital assets can each require different permissions, capital structures and compliance systems.

Bunq says it plans to address the OCC’s concerns and continue pursuing a US presence.

The story is not “America banned Bunq.” It is that the current banking plan failed the regulator’s test.

Which is harder for global fintechs to localise: the product, the balance sheet or the compliance culture?

Sila Did Not Receive $1.4B in Cash—It Received a Conditional Loan Commitment

One word separates a huge financing headline from money already in the bank: conditional.

Sila says it has secured a conditional US Defense Department loan commitment of up to $1.4 billion to expand silicon-carbon anode manufacturing at its Moses Lake, Washington facility and develop battery-cell capacity for specialised applications.

The financing is intended to support domestic battery supply chains serving areas including energy storage, AI data centres, industrial equipment and defence applications.

But “commitment” is not the same as “disbursement.”

Projects typically have to satisfy financing conditions before the full amount becomes available. Sila separately raised $300 million in private capital last month for its planned Phase 2 expansion.

The AI connection is easy to overlook.

AI infrastructure needs batteries for backup power, grid balancing and increasingly energy-dense equipment. That pushes the investment story upstream—from GPUs and data centres into the materials sitting inside energy-storage systems.

The signal I would watch is production capacity actually commissioned at Moses Lake, not the maximum loan headline.

Not Financial Advice.

Is energy storage becoming AI infrastructure in its own right—or will power generation remain the much bigger bottleneck?

Switch at an $80B Valuation? I’d Make KIA Separate the Report From the Filing

The biggest number in the Switch IPO story is still a reported target—not a confirmed offering price.

Bloomberg reported that data-centre operator Switch has confidentially filed for a US IPO. Reuters could not independently verify that filing, although Reuters previously reported that Switch had hired banks for an offering that could raise as much as $10 billion at close to an $80 billion valuation including debt. The size, valuation and timing can still change.

This is where I would use KIA as a research filter.

My prompt would ask it to label every claim as:

Confirmed by company

Reported by named publication

Reported by anonymous sources

Still under discussion

KIA supports online search and Deep Research, which makes it useful for organising a messy IPO timeline before checking the strongest claims against original reporting.

Switch is valuable because AI needs power, cooling and connectivity. But an $80 billion headline tells us what an IPO might value that infrastructure at—not what public investors have agreed to pay.

Not Financial Advice.

Would an $80B Switch IPO validate AI-infrastructure valuations—or become the market’s first serious stress test for them?

Germany’s China Trade Deficit Hit €55B—and German Exports to China Fell More Than 12%

China used to be Germany’s second-biggest export market. It is now ninth.

German exports to China fell to less than €37 billion during the first half of 2026, while imports from China increased 8.9% to €91.8 billion. That widened Germany’s bilateral trade deficit to about €55 billion, up from €40 billion a year earlier.

The direction of dependency is changing.

For years, Germany’s China story was about BMWs, machinery and industrial equipment flowing east. China has invested heavily in domestic manufacturing and increasingly produces equipment and technologies it once imported.

German companies are now dealing with the reverse pressure: weaker Chinese demand for European products and stronger Chinese competition at home.

There is another wrinkle. Germany’s overall exports still grew 3.7% in the first half. This is not simply “German exports collapsing.” It is a specific deterioration in one of Germany’s historically most important industrial relationships.

For markets, trade exposure should be judged by direction—not merely total volume.

Not Financial Advice.

Is Germany de-risking from China—or is China simply becoming less dependent on German industry faster?

China’s PPI Slowed to 3.5% While Consumer Inflation Fell to Just 0.5%

Factories still have pricing pressure. Chinese households barely do.

China’s producer prices rose 3.5% year over year in July, slowing from 4.1% in June and missing the 3.8% consensus. Headline consumer inflation eased to a six-month low of 0.5%, while core CPI was only 0.9% and food prices fell 1.5%.

That combination exposes China’s two-speed economy.

Upstream manufacturers are dealing with higher costs, particularly across mining and raw materials. Businesses selling into the domestic consumer economy still face weak pricing power.

That gap can hurt margins.

A company cannot always pass a 3%–4% rise in production costs to customers whose spending power supports barely positive consumer inflation. Reuters also notes that weak property conditions and job insecurity continue to weigh on household demand.

Beijing has promised faster fiscal spending, but economists expect a lag before that money reaches demand.

So I would watch the gap between PPI and CPI, not either number alone. If producer inflation stays elevated while consumer inflation remains near zero, somebody in the middle absorbs the pressure.

Not Financial Advice.

Which tells you more about China right now: factory prices at 3.5% or consumer inflation at just 0.5%?