Asia wrap: Alphabet’s capex surge reopens the AI funding question as Oil eyes $100

Alphabet’s capex surge reopens the AI funding question

US futures slipped as Alphabet delivered the kind of earnings report that would once have been enough to carry the broader technology complex higher, only for another large increase in artificial intelligence spending to reopen the question investors can no longer avoid: how much capital must be poured into the AI buildout before the returns begin catching up with the ambition?

Nasdaq 100 futures fell 0.6%, while Alphabet dropped more than 3% in late trading after lifting its 2026 capital-expenditure forecast to as much as $205 billion, above its previous estimate of $190 billion and more than double last year’s outlay. The operating numbers were strong, but the market’s reaction function has changed. A year ago, a larger AI budget was interpreted almost automatically as evidence of a larger future prize. Today, every fresh increase comes with a second question attached: margins, depreciation, electricity costs, financing needs, and how long shareholders must wait before the investment begins earning its keep.

The market is not doubting that AI demand exists. It is beginning to question the economics of chasing it at almost any price.

Alphabet and the other hyperscalers may genuinely have little choice. Falling behind in models, computing capacity or cloud infrastructure could prove strategically fatal, which is why the argument that these companies cannot afford not to spend still carries considerable weight. But necessity is not the same as profitability. A company can be forced to defend its place in the race without guaranteeing that every dollar committed generates an attractive return for shareholders.

The AI arms race increasingly resembles a group of industrial powers building aircraft carriers. Nobody wants to be the first to stop, because doing so risks surrendering the seas. Yet each new vessel becomes more expensive to construct, more costly to operate and harder to justify to the people funding the fleet.

That tension explains why the companies writing the cheques are being marked down while many of the firms supplying the machinery continue to rally. South Korea’s Kospi surged 4.4%, with semiconductor and memory names benefiting from the expectation that Alphabet’s larger budget will translate into stronger demand for chips, servers, networking equipment and power infrastructure.

Asia remains the most direct picks-and-shovels expression of the AI boom. The suppliers receive orders and near-term cash flow, while the hyperscalers absorb more execution risk and shoulder the growing capital burden. The gold rush is still alive, but the bill is increasingly landing on the balance sheets of the companies buying the machinery.

Even within semiconductors, however, the tide is becoming less forgiving. STMicroelectronics plunged after issuing third-quarter sales guidance below expectations, a reminder that AI demand does not lift every chipmaker equally. The market is becoming more selective about who owns the true bottlenecks, who merely benefits at the margin and who is still exposed to weaker parts of the global industrial cycle.

The next layer of the debate is moving from the equity market into credit. The AI buildout is no longer being funded solely from the deep reservoirs of internal cash generation that once made the largest technology companies appear almost financially indestructible. Capital expenditure is rising so quickly that free cash flow is coming under pressure, debt issuance is expanding, and investors are beginning to examine coverage ratios, duration exposure and the amount of future earnings already committed to today’s infrastructure.

Equity investors can spend years debating the destination. Bond investors are usually more interested in who is paying for the fuel.

This does not mean the hyperscalers are approaching a credit crisis. Their balance sheets remain formidable. But the direction of travel matters. As AI-related borrowing increases, the largest technology issuers will capture a larger share of investment-grade supply and compete more directly for capital with governments, industrial companies, and the rest of corporate America. At the same time, higher Treasury yields and rising energy costs make every additional data centre more expensive to finance and operate.

The bond market is therefore beginning to place a visible price on an AI boom that equity investors once treated as though it came with a blank cheque.

IPO supply adds another potential claim on the same pool of capital. The concern is not that new listings are inherently bearish. A functioning IPO market is usually a sign of healthy risk appetite. The issue is the scale of the dollars that may need to be absorbed, even if the number of deals remains relatively ordinary.

Large technology, semiconductor and AI-linked offerings could arrive alongside heavier hyperscaler debt issuance, private-market exits and still-elevated valuations across the existing leaders. Investors may soon be asked to fund infrastructure companies, finance hyperscalers, absorb large new listings, and continue supporting the stocks that have already carried the market higher.

The AI boom is slowly moving from a scarcity of assets to an abundance of funding requests.

That would be manageable in a world of falling yields and cheap energy. It becomes more complicated with Brent crude above $97 and threatening to break through $100. Oil rose for a fifth straight session after attacks on two Saudi vessels in the Red Sea added another potential disruption point to an already unstable Middle East supply map.

The Red Sea matters because investors had been willing to tolerate the Gulf shock on the assumption that shipping routes could adjust and the disruption would remain geographically contained. A second active chokepoint makes that assumption harder to defend. The longer the interruption persists, the greater the risk that higher crude prices begin leaking into freight, inflation expectations, consumer spending and central-bank policy.

For AI, the connection is direct. Data centres consume enormous amounts of power. Higher energy prices raise operating costs, while renewed inflation pressure keeps interest rates elevated and lifts the discount rate applied to profits that may not arrive for years. The sector is therefore being squeezed from both sides of the valuation equation: more money must be invested today, while the future cash flows intended to justify that spending become worth less in present terms.

Foreign-exchange markets have so far remained remarkably calm. Higher oil has not yet produced the broad deterioration in risk sentiment normally required for the dollar to fully benefit, while hawkish repricing outside the US has limited the advantage from rising American front-end yields. High-yielding commodity currencies have consequently remained well supported.

Norway’s krone continues to stand out within G10 and remains my preferred four-month carry trade against the Swiss franc. It offers the combination of energy exposure and a favourable yield differential, but the trade still depends on markets treating higher oil as a terms-of-trade benefit rather than the opening act of global demand destruction.

Carry works beautifully while volatility sleeps. The trouble begins when the market wakes up angry.

The European Central Bank now steps into this unusually delicate backdrop. A hawkish hold remains the most likely outcome, with higher European gas prices and renewed Middle East risk keeping inflation concerns alive. The ECB may seek to preserve the possibility of a September hike to prevent front-end yields from falling too sharply, which should offer the euro some temporary rates support.

But a hawkish central bank cannot completely insulate EUR/USD from a broader deterioration in energy security or risk appetite. The market appears dangerously comfortable with the idea that oil can keep rising without eventually forcing a larger adjustment across equities, credit and currencies. Unless the geopolitical newsflow improves, EUR/USD still looks vulnerable to a move towards 1.1380.

Australia offers the cleaner side of the commodity-currency story. Employment rose by 76,000 in June, unemployment held at 4.4% and earlier data were revised higher. The labour market remains tight enough to heighten the risk that the Reserve Bank of Australia adopts a more hawkish tone, particularly as energy prices threaten another inflation impulse.

That combination of stronger terms of trade and a higher prospective interest-rate path should continue supporting the Australian dollar, with AUD/USD still capable of reaching 0.73 by year-end. The obvious caveat is that the Aussie performs well when commodities represent income, but poorly when they become a tax on global growth.

For now, markets are still rewarding the suppliers, tolerating the borrowers and assuming the energy shock will remain manageable. But the room for error is narrowing. Alphabet’s results show that the AI investment cycle is still accelerating, while the reaction in its share price suggests investors are no longer willing to applaud every increase in spending without examining the cost.

The boom is not ending. It is entering the stage where cash generation, balance-sheet durability and financing discipline matter more than headline capex alone. Oil is raising the fare, the bond market is checking the balance sheet, and the IPO calendar is asking how many more passengers investors are prepared to carry.