NextFin News - Asia's biggest technology and consumer companies are walking into their heaviest earnings week of the season carrying the same uncomfortable split: revenue is growing, but profits are collapsing under the weight of artificial-intelligence spending. Alibaba Group's net income fell about 75% even as revenue rose 9% in the June quarter, and Tencent Holdings nearly tripled its capital expenditure. Now the rest of the region's corporate elite - from BYD to the remaining reporters on the calendar - must show whether the AI boom can actually pay for itself, and whether Chinese consumers are finally coming back.
The Week's Two Questions
The earnings calendar for the week of August 24, 2026 is the densest of the season, and it frames two bets that have carried Asian equities for months. The first is that the AI infrastructure buildout is a durable, structural shift rather than a bubble. The second is that China's economy is stabilizing after years of property-led drag. Both are being tested at once, and the early results are not clean.
Consider the scorecard already on the board. Alibaba reported revenue of 268.95 billion yuan ($40.02 billion), a 9% gain that matched the 268.88 billion yuan consensus, yet net income fell to 10.5 billion yuan from 43.1 billion yuan a year earlier. The reason sits in the capital expenditure line: 67.68 billion yuan, up 75%, as the company poured money into AI infrastructure. Free cash flow swung to an outflow of 44.67 billion yuan ($6.58 billion). Its AI cloud and compute services revenue jumped 45% to 48.44 billion yuan, but the AI Labs and Applications segment posted an adjusted EBITA loss of 13.86 billion yuan, more than four times the year-ago loss.
Tencent, which reported a week earlier, told the same story from a different angle. Revenue rose 11% to 204.8 billion yuan and non-IFRS net profit edged up 9% to 68.4 billion yuan, in line with expectations. But operating capital expenditure hit 51.8 billion yuan, up 190% year on year, pushing free cash flow negative at minus 13.8 billion yuan. Strip out the new AI products and Tencent's operating profit grew 19%; include them and growth was 9%. The AI buildout is not a footnote. It is the difference between a 19% profit engine and a 9% one.
The consumer side of the ledger is equally mixed, and in some cases outright weak. JD.com posted its first quarterly revenue decline since listing - down 2.9% to 346 billion yuan - even as non-GAAP net income surged 21% to 8.9 billion yuan. Kuaishou's shares fell about 4% after its report; total revenue grew just 1.4% to 35.5 billion yuan, profit dropped to 3.2 billion yuan from 4.9 billion, and live-streaming revenue fell 13.5% to 8.7 billion yuan. The one bright spot was Kling AI, whose revenue more than tripled to roughly $117 million - a glimpse of monetization that remains small against the infrastructure bill. Xiaomi held profit steady at 6.2 billion yuan on revenue of 108.9 billion yuan, but its EV and AI initiatives lost 2.6 billion yuan in operating terms and smartphone gross margin compressed to 8.5% under memory-cost pressure.
And the macro backdrop is not helping. China's July retail sales grew just 0.6% year on year, a marked slowdown, while fixed-asset investment for the first seven months fell 6.7%. Industrial output eased to 4.5% in July. The economy grew 0.9% in the second quarter, the weakest quarterly pace since mid-2024. As of August 21, the Shanghai Composite sat at 3,905 points and the Hang Seng at 3,311.91 - markets that have rallied on hope more than on earnings.
The AI Capex Paradox: Revenue Up, Profits Down
The first-order read of this earnings cycle is simple: AI demand is real, but it is not yet profitable. The mechanism behind that statement matters more than the statement itself. Capital expenditure does not hit the income statement all at once. It arrives as depreciation over the life of the data center, as power and cooling costs as soon as the racks go live, and as chip procurement costs that front-load the cash outflow. Revenue from AI services, by contrast, arrives only as customers adopt, scale, and pay. That timing mismatch - cash out now, cash in later - is why a company can post 45% cloud growth and a 75% profit collapse in the same quarter.
There is a second-order layer that most investors are not pricing. The AI buildout is not just a China story; it is a global arms race, and China is running it from behind. Goldman Sachs estimates the combined 2026 AI capital expenditure of China's four biggest spenders - Alibaba, Tencent, ByteDance and Baidu - at roughly $102 billion. The comparable figure for the American group Goldman tracks is on the order of $764 billion. That is an eight-to-one gap. Yet the growth rates have converged: TrendForce expects the Chinese four to grow capex more than 80% year on year in 2026, while the American big four are guiding to roughly 77% growth off a far larger base. When a group one-eighth the size is growing at least as fast as the leader, the gap stops widening - but the catch-up spend has to come from somewhere, and it is coming from free cash flow that used to fund buybacks and dividends.
This is where the structural-versus-cyclical call gets difficult. The AI infrastructure shift is structural: the rules of the technology industry have changed, and compute is now the scarce input the way oil was in the twentieth century. No company can opt out without ceding the next platform. But the pace of the buildout - 80% growth in a single year, front-loaded into the second half - is cyclical. It can be slowed, deferred, or cut if the revenue does not arrive. The risk is not that AI is a mirage. The risk is that the income statement pays for the capex before the revenue line catches up, and that shareholders revolt in the interval.
The US peers are not immune to this logic, but they are further along the curve. Microsoft, Meta, Amazon and Alphabet are guiding to roughly $732.5 billion of combined capex in 2026, with $301 billion already spent in the first half and about $432 billion expected in the second. Wall Street is now modeling nearly $935 billion for 2027, within reach of the $1 trillion milestone. That scale buys optionality. China's giants are spending at a similar growth rate without the same revenue base, which means their margin for error is thinner.
"As we enter into the third quarter of the year, we are making substantial progress towards building a new, AI-empowered Tencent in terms of intelligence, applications and infrastructure," said Pony Ma, Tencent's co-founder, chairman and chief executive officer.
The statement is confident. The cash-flow statement is less so. Tencent's free cash flow turned negative. Alibaba's free cash outflow more than doubled. These are not companies in distress - they are printing operating profit from gaming, advertising, and commerce - but they are choosing to convert that profit into GPUs and data centers rather than returning it to shareholders. That is a bet on the future. The market is being asked to trust the timeline.
China's Consumer: Cyclical Soft Patch or Structural Drag?
The second narrative under test is whether Chinese households are spending again. The July data offered little comfort: retail sales up 0.6%, the slowest pace in recent memory, against a backdrop of fixed-asset investment down 6.7% for the year and a property sector that continues to subtract from household wealth. This is the cyclical-versus-structural question in its clearest form.
The cyclical case is straightforward. Retail sales grew 2.6% in the first seven months as a whole, and July's 0.6% could reflect timing effects, a high comparison base, and temporary confidence shocks. If Beijing delivers demand-side stimulus - markets are watching the August Loan Prime Rate signal - consumption could bounce quickly. JD.com's own commentary leans this way: management said marketplace and marketing revenues are expected to accelerate as overall sales recover in the second half.
"Despite near-term revenue headwinds, we achieved strong bottom-line growth, marking a clear inflection in our profit trajectory," said Sandy Xu, JD.com's chief executive officer.
But the structural case is harder to dismiss. Fixed-asset investment is not down 6.7% because of a timing quirk; it is down because property, the traditional engine, is no longer building. Household wealth in China is disproportionately tied to real estate, and a balance-sheet recession does not heal in a quarter. The fact that JD.com - a bellwether for discretionary goods - posted its first-ever quarterly revenue decline while Kuaishou's live-streaming revenue fell 13.5% suggests the consumer is trading down, not trading up.
The two forces are both present, and they pull in opposite directions. The cyclical leg says a stimulus package could lift retail sales back to 3-4% growth within two quarters. The structural leg says the wealth effect from property is gone for good, and consumption will grow at the pace of disposable income, not credit. Investors betting on a V-shaped China recovery are betting against the structural leg. That is the risk in the room.
The Second-Order Question: What If the Revenue Does Not Arrive?
The market has priced the AI narrative as a straight line: capex today, monopoly profits tomorrow. The second-order question is what happens if the line bends. Follow the transmission chain. Capex rises → free cash flow turns negative → buyback and dividend capacity shrinks → the equity multiple compresses because the shareholder-return yield that supported the stock disappears. This is not hypothetical. Xiaomi repurchased roughly HK$11.7 billion of shares year to date; JD.com bought back $1 billion in the first half. Both programs sit on cash flow that is now being redirected to AI and EV infrastructure.
There is also a cross-market dimension. If Chinese tech giants cannot monetize AI fast enough, the buyers of their cloud capacity - the model developers, the enterprise adopters - must be growing even faster, or the capacity sits idle. The Kling AI and Qwen consumer apps are growing, but their revenue contribution remains small relative to the infrastructure bill. The gap between the capex growth rate (80%+) and the revenue growth rate of the AI segments (45% at Alibaba's cloud, strong but smaller than the spend) is the expectation gap that this week's reports must narrow.
The strongest counter-thesis is that this is exactly how platform shifts begin, and that the profit compression is the price of admission. The bulls point to the growth rates themselves: cloud revenue up 45%, marketing services up 22% at Tencent, Video Accounts time spent up 20%. They argue that the companies are buying scale at a moment when their American peers are doing the same, and that whoever blinks first loses the decade. Goldman Sachs has raised its combined capex forecast for the four US hyperscalers to $5.3 trillion for fiscal 2025-2030, and $7.6 trillion for 2026-2031 across compute, data centers and power. In that framing, $102 billion of Chinese AI spend is not excessive - it is the minimum entry fee.
The counter-thesis has force, but it rests on one assumption: that demand keeps compounding. If cloud growth decelerates below roughly 30% year on year, or if free cash flow stays deeply negative for two more quarters while AI revenue growth fails to accelerate, the structural thesis breaks. That is the falsifying signal. It is observable, it is quantifiable, and it will be in the next two quarters of reports.
What to Watch: Scenarios by Horizon
Short term - the next one to two quarters - the direction is set by sentiment and stimulus. A credible demand-side package from Beijing, or a stronger-than-expected August retail print, would lift the entire complex regardless of earnings quality. The Hang Seng, up more than 1% to 3,311.91 as of August 21, is trading on that hope.
Medium term - the next four to six quarters - fundamentals dominate. The base case is that AI revenue continues to grow fast enough to justify most of the capex, but profit margins stay compressed through 2027 as depreciation rolls in. The upside case is that AI agents and enterprise adoption accelerate faster than expected, turning the AI segments profitable ahead of schedule. The downside case is that cloud growth decelerates while capex stays elevated, forcing a cut in shareholder returns and a multiple contraction.
Long term - the structural view - the AI infrastructure buildout is real and durable. Compute is the new scarce input, and the companies that own the capacity will own the platform. But the path is not a straight line, and the companies that overbuilt relative to demand will be the ones that re-rate lower even in a winning industry.
For the week ahead, the calendar matters less than the quality of the answers. BYD reports on August 28, and its Q1 profit already fell 55% as China's EV price war bit. If its Q2 shows the same pressure, the consumer-recovery narrative takes another hit. The AI reporters have already shown they can grow revenue. What this week must show is whether the profit line can stop falling.
The central judgment: this is not a test of whether AI is real. It is a test of whether the income statement can survive the balance sheet's ambitions. The companies that manage the timing mismatch - that grow AI revenue faster than AI costs - will be rewarded. The rest will discover that a structural trend does not protect a cyclical overbuild.
Explore more exclusive insights at nextfin.ai.
