Capital Expenditure

When the people building the infrastructure start selling pieces of it, that's not confidence, that's balance-sheet management. Data-centre developers are reportedly racing to offload billions in stakes to sovereign wealth funds and infrastructure investors, spreading the risk of a capex cycle that assumes AI demand keeps compounding at today's rates. It's a sensible hedge if you're the seller and a bet on someone else's optimism if you're the buyer. Watch which sovereign funds show up on the buy side, because that tells you who's still willing to underwrite the AI infrastructure story at scale.

Seoul has announced an 880 billion dollar investment programme in semiconductors and AI infrastructure, combining state incentives with commitments from Samsung and SK Hynix across advanced packaging, next-generation DRAM, and domestic AI compute capacity. To put the number in context: it is roughly three times the UK's entire annual government capital spending. The plan is a direct response to the US CHIPS Act and Taiwan's own expansion, and it accelerates the demand cycle for lithography and advanced packaging tools, which matters immediately for anyone holding ASML or ARM. The second-order risk is that Korean subsidies further compress margins for non-subsidised fabs trying to compete on cost, including the European foundry capacity the EU has spent four years trying to build.

A combined ten-year investment pledge of $1.3 trillion from Samsung and SK Hynix is the kind of number that makes the UK's semiconductor strategy look like a rounding error. The capital is targeted at advanced memory, logic, and AI chip production, with the implicit understanding that Korean fabs intend to hold their position in HBM supply regardless of what Sophon-style architectural disruption does to demand curves. For UK policymakers, the relevant benchmark is that the entire British industrial strategy budget is measured in single-digit billions. For institutional investors, the read is that Korean semiconductor equities are pricing in sustained pricing power in HBM that the Sophon PFG-1 story, if it scales, directly challenges. These two stories are running in parallel this week and they point in opposite directions for the memory supply chain.

BHP's Jansen potash project is becoming the mining industry's most expensive lesson in greenfield cost discipline. The company has flagged a $2.3 billion impairment and simultaneously disclosed that Stage 2 costs have risen from $4.9 billion to $6.9 billion, a 41% increase on an expansion that was only approved in 2023, as Morningstar's deal coverage confirms. Stage 1 tells the same story: approved at $5.7 billion in 2021, now sitting at $8.4 billion. BHP US shares fell about 2% on the news. The strategic rationale for Jansen, that potash provides diversification away from iron ore and long-term fertilizer demand exposure, has not changed. But the first production is still pencilled for mid-2027, and investors now have to ask whether the economics at $15 billion-plus of combined capital still justify the original thesis, especially if Jansen's supply volumes push global potash prices down by the estimated 7% some analysts project at full ramp.

Alphabet has launched its first primary equity offering in over 20 years, targeting $85 billion through new Class C shares to fund aggressive AI infrastructure investment. The raise reflects a strategic shift as hyperscaler capex is projected to exceed $600 billion in 2026, with 75 percent tied to AI data centers, chips, and networking. Order books were heavily oversubscribed, forcing Google to upsize from an initially smaller target. The dilutive equity raise signals management believes AI capex requirements now outstrip even Google's massive cash generation capabilities.

Chinese companies proposed a record 161 GW of new coal power in 2025, even as Beijing committed to peak emissions before 2030 and phase down coal after 2025. Global Energy Monitor data shows 291 GW still in the pipeline despite clean energy meeting all net power demand growth as coal generation actually fell. The disconnect reflects local economic stimulus priorities, with coal mining companies moving downstream into power generation to lock in demand before potential restrictions. China accounted for 93% of global coal construction starts in 2024, adding 78 GW in 2025 alone while the rest of the world added just 19 GW.

Saudi Arabia's Public Investment Fund will cut foreign assets from 30% to 18% of its $925 billion portfolio as megaproject costs force a strategic retreat from global buying sprees. PIF Governor Yasir Al-Rumayyan signaled the pivot at October's Future Investment Initiative, while Neom's $1 trillion budget faces scaling back amid slower foreign investment inflows than originally projected. The kingdom is shifting focus to AI partnerships, including talks with Andreessen Horowitz on a $40 billion AI fund, as Vision 2030's delivery deadline forces hard choices between global prestige projects and domestic transformation. U.S. Equity holdings already dropped from $35 billion to $20.6 billion between end-2023 and June 2024.

The newly redesigned rocket was fueled and seconds from launch when SpaceX called a halt, marking another setback for a system the company has spent $15 billion developing. Friday's attempt represents the third-generation architecture and a crucial test for SpaceX's broader commercial strategy, which the company warned is highly dependent on Starship's success. The scrub follows a pattern of delays that have pushed back deployment timelines for Starlink's larger satellites and NASA's lunar missions.

Morgan Stanley's Asia team is calling a new industrial supercycle as the region's fixed investment heads toward $16 trillion by 2030, driven by AI infrastructure, energy transition, and defense spending. The thesis centres on hardware over software: major chipmakers' annual capex rising from $105 billion in 2025 to $250 billion by 2028, with Asia controlling the critical supply chains from TSMC's foundries to SK Hynix's memory. Unlike previous cycles built on property speculation, this one is anchored by structural technology shifts that require massive physical infrastructure.

The arithmetic is brutal: Alphabet, Amazon, Meta, and Microsoft will collectively burn through $650 billion in 2026, mostly on AI infrastructure that generates no immediate revenue. Meta hiked its capex outlook to $145 billion yesterday and promptly shed $950 billion in combined market value across the four companies. Microsoft reported a 66% quarterly jump in spending, while Amazon plans $200 billion on data centers. The scale dwarfs Belgium's GDP and makes 21 other major US firms look quaint with their combined $180 billion. Investors are finally asking the obvious question: where are the returns?

Meta stock fell 13% this week after the company raised 2026 capex guidance to $145 billion, driven by AI infrastructure and a Texas data center that ballooned from $1.5 billion to $10 billion. Legal defeats in New Mexico and Los Angeles compounded investor anxiety, while the company delayed rolling out a new AI model due to performance issues. Meta is simultaneously cutting hundreds of jobs to contain costs while pursuing what executives call "frontier ambition" in superintelligence. The 28% drop from recent highs reflects a broader reckoning: investors are no longer willing to fund massive AI bets without clear monetization timelines. Analysts still see 62% upside to $850, but that requires faith in returns years away.

Amazon raised quarterly capex to $31.4 billion, implying an annualized pace exceeding $118 billion as AWS chases AI demand. The company plans $200 billion for data centers and specialized chips, including $50 billion for US government AI infrastructure with 1.3 gigawatts of power capacity. CFO Brian Olsavsky confirmed AWS drives the largest share of spending, with elevated levels expected through 2026. Amazon issued $12 billion in bonds last year to fund the buildout, positioning against Alphabet's $85 billion capex target. The scale reflects hyperscalers' recognition that AI infrastructure is winner-take-all: fall behind now, lose the cloud wars permanently.