Alibaba's Landmark REIT Approval and HappyHorse AI Trial Signal a New Era of Innovation-Driven Efficiency for China's Platform Economy
Dual Signals of Policy Shift: Alibaba’s REIT Approval and “HappyHorse” Gray-Testing Mark Platform Economy’s Entry into a New “Innovation–Efficiency” Cycle
On March 13, 2026, the Hong Kong Exchanges and Clearing (HKEX) officially confirmed that Alibaba Group may proceed with the spin-off of an infrastructure-focused public REIT on the Shenzhen Stock Exchange—China’s first REIT pilot led by a technology giant, underpinned by self-owned digital infrastructure assets such as data centers. Almost simultaneously, Alibaba Cloud launched gray-scale testing of its internally codenamed large video-generation model, “HappyHorse”: at 720p resolution, its generation cost dropped to ¥0.44 per second—62% below the industry average—and it has already been integrated into core applications including Taobao Live, Quark AI Search, and DingTalk Smart Meetings. Though seemingly independent, these two developments jointly constitute the obverse and reverse of a single policy logic: regulatory focus on the platform economy has systematically shifted—from the “standardization and rectification” phase of 2021–2024—to a new paradigm centered on “promoting innovation, enhancing efficiency, and stabilizing expectations.” This shift is no temporary relaxation, but rather a deep structural realignment grounded in technological evolution, capital-efficiency optimization, and strategic national coordination.
REIT Spin-Off: A Paradigm Revolution—from “Asset Accumulation” to “Capital Circulation”
Conventionally, internet giants’ IDCs, cloud servers, and optical module clusters have been viewed as heavy-asset burdens—subject to rapid depreciation and long payback periods. By contrast, the underlying assets approved for Alibaba’s REIT explicitly comprise its ultra-large-scale intelligent computing centers built in the Yangtze River Delta and Guangdong–Hong Kong–Macao Greater Bay Area—including liquid-cooled clusters, high-speed optical interconnection nodes, and green power supply systems. The pivotal breakthrough lies in the regulator’s formal recognition—first of its kind—that digital infrastructure possesses stable, real-estate-like cash flow characteristics. Under HKEX’s exemption provisions, Alibaba is not required to offer guaranteed allocation rights to existing shareholders—a clear signal of policy openness toward the new model of “professional asset operation + public capital participation.” This is not merely about monetizing assets; rather, it establishes a virtuous cycle of “construction → operation → securitization → reinvestment.” In contrast, Zhaolong Interconnect’s ¥1.079 billion investment in a high-speed interconnect components project (with annual output capacity of 18 million units) represents capacity expansion at the hardware manufacturing level; Alibaba’s REIT, however, transforms existing digital infrastructure assets into tradable, priceable, and leveragable financial instruments. Semiconductor equipment maker Advanced Micro-Fabrication Equipment (AMEC) reported a 197% surge in Q1 net profit (to ¥930 million), of which ¥397 million came from gains on the sale of shares in Top Nanotechnology—illustrating how mature capital-market logic has become within the semiconductor supply chain. Alibaba’s REIT transplants this same logic into the digital infrastructure domain, transforming “computing power” into a quantifiable, tradable, and valuably assessable new factor of production.
“HappyHorse” Gray-Testing: Breaching the Cost Barrier at the Commercialization Inflection Point of AIGC
The ¥0.44-per-second cost of “HappyHorse” is no mere parameter tweak. The model employs a Mixture-of-Experts (MoE) architecture coupled with dynamic frame-rate compression algorithms, enabling real-time 24-frame-per-second video generation at 720p resolution, with inference latency under 350ms. Its technical foundation is deeply integrated with Alibaba’s in-house Hanguang NPU and Feitian AI computing platform scheduling system—boosting per-unit compute utilization to 83% (versus an industry average of ~52%). This cost structure directly shatters the commercialization threshold: Taobao Live merchants can now embed product-video generation into daily operations (cost per video under ¥5); DingTalk users can generate visualized meeting summaries with one click (a 10-minute video costs only ¥26.40). Notably, Sungrow Power Supply reported a 40.12% decline in Q1 net profit—primarily due to pricing wars squeezing margins in the photovoltaic equipment sector—whereas “HappyHorse” achieved further cost reduction amid the ongoing downward trend in compute costs. This affirms AI’s distinctive “marginal cost tending toward zero” law: once model efficiency crosses a critical threshold, application adoption surges exponentially—not linearly. GF Securities posted a 70.73% year-on-year rise in Q1 net profit (to ¥4.707 billion), with AI-powered investment advisory clients in its wealth management business growing by 142%—a microcosm of AIGC’s penetration into B2B services.
Valuation Logic Reset: Three High-Certainty Industrial Chains Ascend
The confluence of policy shift and technological breakthrough is reshaping valuation anchors across the internet sector. The former consumer-internet logic—centered on DAU, GMV, and price-to-sales ratios—is giving way to three high-certainty thematic axes:
First, the “Physical-Layer Dividend” of Compute Infrastructure.
Alibaba’s REIT underlying assets will drive rigid demand for high-speed optical modules, liquid-cooled servers, and smart NICs. Zhaolong Interconnect’s 120-mu industrial park focuses specifically on “high-speed data transmission,” targeting the 800G/1.6T optical interconnect bottlenecks inherent in AI training clusters. AMEC’s R&D expenditure accounts for 31.14% of revenue, and its etching equipment has already entered production lines at leading domestic intelligent computing centers. Such firms are no longer categorized as cyclical stocks—but as the “reinforced concrete” of digital infrastructure.
Second, the “Penetration-Premium” of AIGC Applications.
“HappyHorse” validates the feasibility of scalable, commercial-grade video generation—paving the way for accelerated deployment across education (AI-generated teaching materials), healthcare (visualization of surgical procedures), and government services (short-video policy explanations). Its value lies not in the model itself, but in the GDP增量 unleashed by lowering society-wide content-production barriers—a logic analogous to Sungrow’s growth driver (PV installation volume), yet delivering higher marginal returns.
Third, the “Regulatory Arbitrage” of REIT Expansion.
Following Alibaba’s successful pilot, Tencent, ByteDance, and others—with over 3 million servers collectively—could rapidly securitize their server assets en masse. Drawing reference from U.S. data-center REITs—averaging a 4.2% dividend yield and 28x P/E—versus A-share communications equipment stocks trading at just 18x P/E—the valuation gap itself represents a space for institutional arbitrage. By easing issuance requirements via exemption clauses, regulators are effectively channeling long-term capital into the market to hedge against the high volatility of AI-related investments.
Stabilizing Expectations: Rebuilding Market Consensus—from “Risk Pricing” to “Innovation Pricing”
The deeper significance of Alibaba’s REIT approval and “HappyHorse” gray-testing lies in ending the market’s “uncertainty discount” applied to the platform economy. Between 2021 and 2023, the core driver behind the internet sector’s declining valuation multiple was rising risk premiums caused by policy ambiguity. Today, HKEX exemption provisions, refinements to Shenzhen Stock Exchange REIT rules, and expanded AIGC regulatory sandboxes collectively establish a clearly defined “innovation tolerance boundary.” When the regulatory framework explicitly endorses both “using technology to enhance capital efficiency” (REITs) and “using models to reduce societal costs” (“HappyHorse”), markets can price assets based on discounted cash flow (DCF)—not emotional speculation. AMEC’s R&D intensity exceeding 30%, yet commanding enthusiastic investor interest, foreshadows precisely this logic.
The platform economy has never left center stage—it has simply re-emerged in a more robust, efficient, and sustainable form. This time, code and capital are finally resonating in unison.