Today's Hong Kong market saw a dramatic scene with KNOWLEDGE ATLAS (02513.HK). In the afternoon, its gains expanded to nearly 30%, hitting an intraday high of HKD 658.5, setting a new record since its listing and pushing its total market value past HKD 290 billion. It closed up 13.48% at HKD 635.5. Since its listing on January 8 at an issue price of HKD 116.2 to today's HKD 635.5, KNOWLEDGE ATLAS's stock price has surged approximately 4.5 times in just four months. The catalyst was the recent upgrade of SothisAI, a subsidiary of Sugon, which now fully integrates KNOWLEDGE ATLAS's GLM-5 and Alibaba's Qwen3.5 large models. The market interpreted this as an acceleration in the commercial deployment of GLM-5, igniting investor enthusiasm.
A comparison reveals a stark contrast:
| Target | YTD 2026 Performance | | :--- | :--- | | KNOWLEDGE ATLAS (02513.HK) | ~ +380% | | Hang Seng Tech Index | -11% | | TENCENT (0700.HK) | -24% | | Alibaba (9988.HK) | ~ -9% |
The same Hong Kong market, the same AI sector, the same trading hours—yet KNOWLEDGE ATLAS rose 13% while TENCENT fell 24%. This isn't just a market fluctuation; it signals a major reshuffle in valuation logic. 01 First, Look at the Money: How Much Have Tencent and Alibaba Earned and Invested? Saying TENCENT and Alibaba are "flush with cash" is no exaggeration. Let's examine their recent financials. Alibaba (BABA / 9988.HK)
| Fiscal Year | Revenue | Net Profit | Notes | | :--- | :--- | :--- | :--- | | FY2023 | $126.5B | $10.6B | — | | FY2024 | $130.4B | $11.0B | — | | FY2025 | $137.3B | $17.8B ↑62% | Quarterly Capex +80% |
Alibaba's FY2025 net profit surged 62%, showing a solid bottom line. However, it simultaneously announced a three-year plan to invest RMB 380 billion (approx. $53B) in AI and cloud infrastructure, equating to an annual burn of about $17.7B. Goldman Sachs pinpointed the issue: Alibaba's targeted AI capital expenditure already exceeds its full-year EBITDA. Citi added that to sustain this pace, Alibaba Cloud must grow at least 40% annually—a speed the market widely deems unattainable. Today (May 13), Alibaba officially released its FY2026 Q4 and full-year results, with the market keenly focused on whether cloud business growth can maintain the previous quarter's 36% high rate. TENCENT (0700.HK)
| Year | Revenue | Net Profit | Capital Expenditure | | :--- | :--- | :--- | :--- | | 2023 | $86.3B | $16.3B | $3.4B | | 2024 | $91.8B | $27.0B ↑66% | $10.7B ↑215% | | 2025 | $104.6B | Continued Growth | ~$11.5B (Chip Constraints) |
TENCENT's 2024 net profit was $27.0B, while its capital expenditure skyrocketed from $3.4B to $10.7B, more than tripling. TENCENT also announced a multi-year new infrastructure plan worth RMB 500 billion (approx. $70B). Ironically, its 2025 capital expenditure failed to rise further—not for lack of willingness to spend, but due to an inability to purchase sufficient GPUs. Chip export controls have blocked the path of converting capital into computing power. 02 The Gap: How Much Will U.S. Tech Giants Spend This Year? The contrast becomes stark when comparing with the U.S. side.
| Company | 2026 Capital Expenditure Plan | | :--- | :--- | | Amazon | $200B | | Google (Alphabet) | Up to $190B | | Microsoft | $190B | | Meta | $125B—$145B | | **Four-Company Total** | **~ $725B** |
This figure is approximately 15 times the combined annual AI investment of TENCENT and Alibaba. Google Cloud's Q1 revenue grew 63% year-over-year to $20B, demonstrating that computing power investments are translating into real revenue. In contrast, the monetization paths for TENCENT and Alibaba's AI initiatives remain unclear, leading investors to question the return on these massive expenditures. 03 Tencent and Alibaba Tell a Good Story, But Investors Want to See the Ledger The term "AI" has been mentioned countless times by TENCENT and Alibaba this year. Their Hunyuan model, Tongyi Qianwen, and cloud computing sound impressive. Bloomberg's sentiment: "Show Me the Profits." TENCENT's 2024 net profit was $27.0B, and Alibaba's FY2025 net profit was $17.8B. Neither company's financials look poor. Yet, their combined market capitalization has evaporated by approximately $66B this year. They are profitable and tell a compelling story, yet their stocks fall. The problem lies in investors' inability to calculate the payback period. AI infrastructure is a long-cycle asset. After massive spending, when will it translate into visible incremental revenue from users, cloud services, or advertising? The market sees no clear roadmap. 04 Hang Seng Tech Index Hit Its Worst Point, Down 15.7% Three pressures converged to create this downturn: The Federal Reserve's role. Signals of "no rate cuts in 2026" triggered a swift withdrawal of foreign capital from the USD-sensitive offshore Hong Kong market. Domestic competition. Internet giants are engaged in subsidy wars for AI assistants and price wars in food delivery, cutting into each other's profits in a saturated market, which weakens profit margins and shakes valuation logic. Weak fundamentals. The manufacturing PMI fell to 49.3 in January-February, and consumer recovery fell short of expectations. With a weak overall foundation, it's no surprise stock prices suffered. 05 A More Troublesome Issue in the Supply Chain: Component Shortages Parallel to the stock market narrative, China's AI infrastructure construction hit another wall: a parts shortage. Intel quietly warned major Chinese cloud providers: delivery lead times for some server CPU models have extended to up to six months, with capacity bottlenecks potentially lasting until early 2027. The situation with High Bandwidth Memory (HBM) is even tighter: SK Hynix stated publicly that its 2026 production capacity is essentially sold out, and supply tightness may persist until the end of 2027. The primary reason TENCENT couldn't significantly increase its 2025 capital expenditure was the inability to procure enough AI chips. Even if TENCENT were willing to spend more, the hardware isn't available. Export controls have blocked high-end GPUs, and now even supporting components are becoming scarce, creating an unforeseen chain reaction. The future depends significantly on whether domestic production can bridge this gap. 06 Where Did the Money Go? Chasing "Pure AI" Stocks KNOWLEDGE ATLAS's surge today makes this point crystal clear. Capital no longer wants to pay a premium for conglomerates where "AI is just one business line." It is betting solely on 100% pure AI companies. Firms like KNOWLEDGE ATLAS and MiniMax, listed for less than six months and without stable profits, enjoy valuation multiples far exceeding those of established giants like TENCENT and Alibaba, which generate tens of billions in annual net profit. This logic mirrors the U.S. market: the premium for Nvidia and Palantir isn't for conglomerates. MiniMax and KNOWLEDGE ATLAS are set to be included in the Hang Seng Tech Index on June 8, which is expected to drive over $1B in passive fund inflows, potentially fueling this trend further. Today's surge in KNOWLEDGE ATLAS draws the valuation chasm between old and new AI more clearly than any report. TENCENT and Alibaba are not short on cash or profits. However, they face a harsh dual dilemma: On one side, the four U.S. giants are collectively investing $725B this year, creating a computing power gap of a different magnitude. On the other side, chip controls and component shortages prevent Chinese companies from spending their money effectively; even if they try to catch up, the hardware isn't there. The first phase of the AI narrative was about storytelling. The second phase is about accountability: where is the money going, how will it be earned back, and when? The ledger doesn't lie, but today's market shows it's willing to pay more for the dream of "pure AI." Data updated to the close of May 13, 2026.
Comments