U.S.-China Relations
NextFin News - Alphabet has raised its 2026 capital expenditure forecast to $195 billion–$205 billion, the third upward revision in as many quarters, betting that AI demand will keep compounding fast enough to absorb the largest infrastructure build-out in its history. The market's verdict was immediate: shares fell 4.2% in after-hours trading on July 22, to $327.40 from a regular-session close of $341.91, even as revenue and cloud growth beat estimates. The problem behind the sell-off is that the supply-chain hedge meant to protect that bet — the "China+1" push to shift hardware production into India and Vietnam — is running into its own arithmetic: the alternative hubs now sit inside the US tariff wall, and the inputs that power the AI build-out itself remain overwhelmingly concentrated in China.
The capex ratchet: demand is real, but the bill is arriving first
The sequence tells the story. In February, Alphabet set 2026 capital spending at $175 billion–$185 billion. In April, after Google Cloud revenue grew 63% year over year to $20 billion and the committed backlog nearly doubled to $462 billion, the range moved to $180 billion–$190 billion. In July, finance chief Anat Ashkenazi lifted it again to $195 billion–$205 billion — above the roughly $188 billion analysts had modelled — citing "an acceleration in the delivery of capacity to meet growing demand."
The demand side is not an illusion. Second-quarter revenue reached $119.8 billion, up 24% year over year, the company's twelfth straight quarter of double-digit growth. Google Cloud accelerated to $24.8 billion, up 82% year over year, with operating income of $8.8 billion — more than triple the $2.8 billion from a year earlier. The backlog climbed to $514 billion, more than $50 billion higher than the previous quarter, and management expects to recognize just over half of it as revenue over the next 24 months. Search advertising, often written off as vulnerable to AI disruption, grew 17% to $63.3 billion, and YouTube advertising rose 13% to $11.1 billion.
But the spending is arriving ahead of the cash to fund it. Second-quarter capital expenditure reached $44.9 billion, roughly double the $22.4 billion from the same period a year earlier, and first-half capex totalled $80.6 billion. Against $39.1 billion of operating cash flow in the quarter, that produced negative free cash flow of $5.9 billion — the first such quarter in the period tracked by independent analysts. Trailing-twelve-month free cash flow stood at $53.3 billion, down sharply from the run rate investors had underwritten. To pre-fund the build-out, Alphabet went to the equity market in June with an $80 billion raise, including a $10 billion private placement from Berkshire Hathaway — its first equity issuance in two decades.
"We're still in a supply-constrained environment," Ashkenazi told analysts on the July earnings call. "We are seeing very strong demand both from external Cloud customers as well as across the business."
Alphabet is not alone in the spending race, but it is leading it. Meta has guided 2026 capex to $115 billion–$135 billion, nearly double its 2025 spend of $72.2 billion. Microsoft reported $37.5 billion of capex in its latest quarter and has said spending will rise further as component pricing lifts its bill. Amazon's 2026 capex is expected to climb roughly 18% to about $146.6 billion, according to consensus estimates compiled ahead of its report. Alphabet's $195 billion–$205 billion range sits above every peer, and each upward revision has been met with investor caution rather than enthusiasm. The market is beginning to treat capex not as a pure growth signal but as a claim on future margins — and the July sell-off showed what happens when that claim grows faster than the cash flow backing it.
China+1's arithmetic broke when the tariff wall moved
Here is where the two halves of Alphabet's story collide. On the hardware side, the company has been pursuing the same supply-chain diversification as its peers. Reporting in 2026 indicated Google plans to move all Pixel phone, smartwatch and earbud production out of China starting in 2027, consolidating manufacturing in India and Vietnam. Google has not publicly confirmed the plan, but the direction is consistent with its earlier move to begin Pixel assembly in India in 2024, when the company was targeting annual Pixel shipments above 10 million units.
The logic of that move was always tariff arbitrage: keep China for the deep supply chain, assemble the final product somewhere the US levy is lower. That logic no longer holds. Under the reciprocal tariff regime announced in 2025, Vietnam faces a 46% US levy, India 26% and Taiwan 32%. China's rate has oscillated through the cycle — from 10% to as high as 125% and back toward 10% — but the broader point stands: the tariff wall now encompasses every major link in the Asian electronics chain. "The 'China Plus One' strategy has been severely undercut by Trump's tariffs that have by now encompassed every U.S. trading partner," said Eswar Prasad, professor of international trade and economics at Cornell University. "The viability of rerouting output and restructuring supply chains through countries such as Vietnam and India... has been shattered by the latest round of tariffs."
The transmission mechanism is straightforward. A China+1 strategy works only if the marginal unit produced outside China reaches the US consumer at a lower landed cost than the marginal unit produced inside China. When Vietnam's levy sits at 46% and India's at 26%, that condition fails for a wide range of goods. The diversification still reduces geopolitical concentration risk, but it no longer delivers the tariff relief that justified its cost premium — which means companies are paying more, for less protection, than the strategy promised.
The alternative hubs also lack the scale to absorb a meaningful share of the shift. India's manufacturing imports from Western countries grew at a 6.3% compound annual rate between 2014 and 2023, compared with 12.4% for Vietnam and Thailand combined — a corrective to the narrative that India is the runaway winner of the diversification race. Vietnam, the more established electronics assembler, recorded $165 billion in electronics exports in 2023, a deep ecosystem anchored by Samsung, Intel and Foxconn, but still small next to China's. Apple's decade-long shift of iPhone production to India remains the benchmark for what is possible — and for how long it takes. Even there, the move reduces geographic concentration in final assembly without eliminating upstream exposure to Chinese components and materials.
The AI build-out itself still runs through China
This is the second-order risk that a $200 billion-plus capex plan cannot escape, and it is the one the capex guidance does not price. Building AI capacity is not just buying chips; it is building data centres, power infrastructure, battery backup, cooling systems, racks and networking gear. China accounts for around 30% of global manufacturing output, and its share rises to near-commanding levels in the inputs that matter most for that build-out: more than 80% of global smartphone production, over 80% of battery cell production, 70% of electric vehicles and more than 80% of solar panels, according to international energy and trade data. Battery cathode and anode active materials — the chemistry inside every grid-scale storage system that backs up a data centre — are concentrated even further, at roughly 85% and 90% respectively.
Moving final Pixel assembly out of China does not remove that upstream dependency. A data centre financed by today's capex guidance will be built with steel, solar arrays, battery racks, inverters and cabling whose supply chains are still heavily China-centric. Alphabet's hardware diversification and its AI infrastructure build-out are therefore moving at different speeds: one is a multi-year, assembly-level migration; the other is a two-year spending sprint into a supply chain that has not diversified at the same pace. The company is, in effect, spending $200 billion to build capacity whose physical inputs remain exposed to the very concentration risk the China+1 strategy was meant to hedge.
Google's own executives have flagged the bottleneck. When asked what keeps him up at night, CEO Sundar Pichai answered "compute capacity," pointing to "power, land, supply chain constraints." Google's AI infrastructure chief, Amin Vahdat, has said the company must double its serving capacity every six months to meet demand — a cadence that leaves little room for supply-chain friction. That cadence is what makes the mismatch dangerous: a six-month doubling schedule cannot wait for a five-year supply-chain rebuild.
Cyclical hedge versus structural reality
The right way to read this is to separate the two forces at work. China+1 is, at its core, a cyclical and tactical hedge: it is about dispersing risk so that a lockdown, an export control or a single tariff line does not halt the entire chain. It is a portfolio decision, and like any hedge it carries a cost. China's manufacturing ecosystem, by contrast, is structural — a regime-level depth that does not self-correct or migrate quickly.
The evidence for the structural read is in the numbers. China retained the top position in the Asia Manufacturing Index for the third consecutive year in 2026, scoring a perfect 100 on domestic supply-chain strength — a measure of its ability to internalise production inputs and minimise external dependency. New Zealand's Ministry of Foreign Affairs and Trade noted in a June 2026 assessment that China's manufacturing output exceeded $4.6 trillion in 2024, nearly 60% higher than the United States, and that the country accounts for around 30% of global manufacturing output. Its sector shares in smartphones, batteries, EVs and solar are not marginal advantages; they are choke points that take years to replicate elsewhere, because they rest on the co-location of raw-material refining, component production, assembly and logistics in a single geography.
The cyclical counter-case has merit too, and it deserves weight. Tariffs are policy shocks, and policy can be renegotiated. Daniel Newman, CEO of research firm The Futurum Group, has argued that the current tariff levels are unlikely to hold in their present form and that "more fair trade deals" will be struck with partners such as Vietnam and India. If that happens, the China+1 arbitrage partially returns, and the diversification thesis regains its cost logic. There is also the resilience argument: even at a higher cost, a parallel supply chain has option value when a single lockdown can idle the entire chain, as it did during the pandemic.
But even in that scenario, the timing mismatch remains. A tariff deal may come faster than a battery supply chain can be rebuilt in India or a solar panel supply chain can be rebuilt in Vietnam. The hedge is real, but it is shallower and slower than the narrative suggests — and for a company running on a six-month capacity-doubling cadence, "slower" is the operative word.
What would prove this wrong
The strongest case for China+1 is that it was never primarily about cost — it was about resilience, about avoiding a single point of failure after the pandemic-era lockdowns and amid tightening export controls. Even partial diversification has option value, and for a company like Google, whose hardware revenue is small relative to Search and Cloud, the cost of that option may be worth paying.
That argument holds only if the option actually functions. Two signals would validate the diversification thesis: first, if Google completes the Pixel China exit on schedule in 2027 with no material impact on unit cost or gross margin; second, if Vietnam and India capture a rising share of electronics value-added — not just final assembly — over the next four quarters. If the move slips, or if landed costs rise materially, the arithmetic turns against it. On the capex side, the falsifying signal for the margin-compression view is backlog conversion: if more than 50% of the $514 billion backlog turns into revenue within 24 months while gross margins hold, the spending is being absorbed by demand, not wasted on idle capacity.
Outlook: who benefits, who is exposed
In the short term, the capex momentum supports the equipment makers, data-centre landlords and power-infrastructure suppliers that sit in front of Alphabet's spending. The hyperscaler stocks themselves, however, face growing scrutiny: the market has moved from rewarding capex as a growth signal to questioning it as a margin tax, and the July after-hours drop showed how quickly the multiple can compress when the spend outpaces the cash flow.
Over the medium term, the key variable is utilization. If the backlog converts to revenue at the pace management expects — just over 50% of the $514 billion within 24 months — the spend pays for itself and margins can hold despite the higher depreciation load. If conversion slows while capacity keeps coming online, margins compress and the negative free-cash-flow quarters become a pattern rather than an anomaly. The company's own inventory build — hardware related to TPU systems for sale rose from $2.4 billion in December to $10.0 billion in June — is an early marker of how much capacity is being stood up ahead of confirmed consumption.
Structurally, the winners are whoever controls the bottlenecks — compute, power and the critical inputs — rather than whoever assembles at the lowest cost. China-based component suppliers remain exposed to the upside of a build-out they are not supposed to benefit from; brands that have priced their US sales assuming India or Vietnam assembly offers tariff relief are exposed to the downside. The asymmetry is uncomfortable: the companies best positioned to profit from Alphabet's capex are concentrated in the geography Alphabet is trying to diversify away from.
The base case is that capex stays elevated, margins compress modestly, and hardware diversification proceeds slowly and incompletely. The upside case is that AI monetisation outruns the spend and the backlog converts faster than expected, turning today's negative free cash flow into tomorrow's operating leverage. The downside case is a utilization gap coinciding with tariff pass-through, squeezing margins from both sides.
Alphabet is not really betting on AI demand — that part is already visible in a $514 billion backlog. It is betting that the world can build AI infrastructure faster than geopolitics can close the supply chain behind it. That is the real $205 billion question, and the answer will not be in the next earnings call.
Explore more exclusive insights at nextfin.ai.
