Nepal is moving from rescue to reconstruction, a fresh strike puts Ukraine's energy system back in focus, and the AI boom is getting very expensive. Let's get into it.

Nepal has formally shifted the response to the August 26 Bhotekoshi-Trishuli floods from search-and-rescue toward rehabilitation and reconstruction. Reuters reports 1,455 deaths and 5,285 people still missing, while the government says active searches have ended except for emergencies. The early recovery bill is now approaching $5 billion.

The infrastructure damage is already enormous. A preliminary government assessment puts damage to infrastructure at about Rs 227 billion, including roughly Rs 103.54 billion to private and public buildings, Rs 20 billion to roads, Rs 13.8 billion to motorable bridges and Rs 87.87 billion to drinking water and sanitation. Officials say about Rs 63 billion is needed immediately for reconstruction.

The harder question: where does the money come from? Nepal's growth forecast has already been cut to 3.7% for FY2026/27, with the World Bank expecting reconstruction to help lift growth to 5.2% in FY2027/28. But borrowing for reconstruction, protecting vulnerable households and restoring productive infrastructure all compete for the same fiscal space.

Why this matters: reconstruction is not just a disaster-management exercise. Done well, it can replace damaged infrastructure with more resilient roads, bridges, hydropower and water systems. Done badly, it can become a multi-year spending programme with weak oversight and little productivity gain.

The best test is simple: does every rupee rebuild what Nepal lost—or build something better than what stood there before?

NEPSE closed Wednesday at 2,572, down 6.38 points, while turnover rose to Rs 3.74 billion from Rs 3.13 billion. Only 99 companies advanced against 167 decliners. Banking was the sole sector index to finish higher.

The Nepali Congress votes today on its president, office-bearers and central working committee at the party's 15th general convention. Gagan Thapa faces Kishor Singh Rathore and Shyam Krishna Khadgi for president. The reshaped central committee will have 297 members, including 21 seats reserved for leaders under 35.

Himalaya Airlines is under serious financial pressure, reporting accumulated losses of around Rs 15 billion and outstanding dues of at least Rs 629 million to Kathmandu's Tribhuvan International Airport. The airline says higher fuel, taxes and airport charges have squeezed cash flow; CAAN says unpaid dues could eventually threaten operations.

The World Bank now expects Nepal's FY2026/27 growth to slow to 3.7%, with industry hit hardest by damage to hydropower, transmission and transport infrastructure.

*The data presented below are from the official sites of NEPSE, NASDAQ and London Bullion Market Association respectively.

🇺🇸 Wall Street: after two record-setting sessions, the rally cooled on Wednesday. The Dow fell 1.06%, the S&P 500 0.57%, and the Nasdaq 0.67% as oil and Treasury yields rose. The 30-year Treasury yield touched 5.68%, its highest since 2002.

🛢️ Brent crude: around $101.87/barrel; WTI around $89.66 as geopolitical and supply risks kept oil above $100.

🥇 Gold: spot gold fell to $4,107.88/oz on Wednesday; Comex October gold settled at $4,113.80.

🥈 Silver: Comex silver settled near $59.90/oz.

💱 NRB rates: US$1 at Rs 153.98 buying / Rs 154.58 selling. Domestic hallmark gold was Rs 292,300/tola and silver Rs 4,425/tola.

Russia launched one of its largest recent air assaults on Ukraine on Wednesday, using 70 missiles and dozens of drones. At least 24 people were killed, including children, in attacks across several regions; a residential building in Pryluky was the deadliest hit. Ukraine says energy infrastructure was among the targets.

Ukraine's air defences intercepted more than 80% of the incoming drones, but ballistic missiles remain harder to stop. Kyiv is again pressing for more Patriot air-defence ammunition as both sides increasingly target energy and fuel infrastructure.

Why it matters beyond the battlefield: attacks on energy infrastructure can move through the global economy via electricity shortages, fuel markets and higher government spending. That matters even more now with Brent crude already above $100 and shipping routes under pressure.

A new US-led collaboration is putting $1.8 billion into building the biological datasets AI models need to predict how cells behave. The initiative brings together Biohub, the US Department of Energy, the NIH, Meta, Google DeepMind and Isomorphic Labs.

Meta, Google DeepMind and Isomorphic Labs are collectively committing $300 million. The Department of Energy plans to put in more than $500 million over five years, while the NIH is aligning more than $500 million in existing federal data investments with the project.

The ambition is larger than another biomedical database: create standardized, AI-ready measurements of how cells respond to different conditions, so models can eventually run useful virtual experiments before researchers spend years and millions testing them physically. The first major dataset is expected in about a year.

Why it matters: AI's next productivity jump may come from better proprietary data, not just bigger models. Biology is especially data-hungry because the world does not yet have a clean digital description of how living systems respond to change.

SpaceX is in talks to raise $40 billion to buy Nvidia AI chips, according to people familiar with the matter. The proposed structure is about $10 billion in bank loans and $30 billion in investment-grade debt, with Apollo expected to lead the financing and PIMCO among the firms approached.

That is a remarkable shift in the economics of AI. The old AI story was about finding the best model; the new one is about securing enough chips, power, land and financing to run the model economy at scale.

SpaceX is not alone. Nvidia is working with financial institutions on platforms designed to mobilise hundreds of billions of dollars for AI infrastructure, while the wider industry is increasingly using debt to fund data-centre expansion. Reuters has cited Morgan Stanley estimates that $1.5 trillion of external financing may be needed for AI infrastructure by 2028.

The investment question: extraordinary growth can justify extraordinary capital spending—until financing costs or AI revenue growth stops keeping pace. At that point, the winners may be the companies selling the infrastructure rather than the companies carrying the debt to buy it.

Nepal may need billions for reconstruction. How should it finance the rebuild?

Reply with A, B or C:

Answers just below. No peeking. 👇

Capex (say it: KAP-ex)

Meaning: Short for capital expenditure—money a company spends on long-term assets such as data centers, factories, power systems, or equipment.

Example: “Big Tech’s rising capex shows how much money companies are pouring into AI infrastructure.”

For business/promotions, hit us up at: [email protected]

Made it this far? Reply and tell us what you want unpacked tomorrow — we read every message.