Envision’s launch of Dubhe, an energy foundation model unveiled at Abu Dhabi Sustainability Week, is a clear signal: the AI energy system is moving from slideware to infrastructure. The company’s Physical AI stack, paired with its Tianji weather model and anchored by a new strategic partnership with Masdar, aims to orchestrate generation, storage, grids, and demand in real time. For investors, the message is actionable. China is building the operating system for abundant, low-cost energy at scale—and exporting it.
Dubhe’s core task—continuously optimizing a multi-asset energy network—addresses the bottleneck everyone in AI is suddenly facing: power. The model analyses live data streams across wind, solar, storage, mobility, and industrial loads to improve dispatch and reduce system costs. Envision framed it as a foundation for the world’s largest Physical AI system. The Masdar partnership matters. It puts a China-built control brain inside Gulf megaprojects where capex and ambition align, accelerating deployment pathways across MENA, South Asia, and Africa. This is what leadership in the energy transition looks like now: software-first, grid-aware, and globally partnered.
Renewables are weather-limited. Envision’s Tianji weather foundation model bolts predictive intelligence onto Dubhe’s control layer, allowing hour-by-hour and minute-by-minute forecasting to shape dispatch and storage cycling. That capability matters most in large, complex grids with high renewable penetration—precisely where China operates. The country already runs the world’s largest power system and manufactures the hardware stack behind it, from turbines and panels to inverters and batteries. Optimizing across those assets with model-based controls is how costs drop toward near-zero marginal energy, the only sustainable way to support the coming wave of AI workloads. The Abu Dhabi announcement reframes AI not as an energy problem but as an energy solution—if software steers electrons with precision.
Beijing’s industrial policy continues to convert capex into capabilities. China’s companies dominate cost curves in solar, batteries, and EVs, and are rapidly building AI and semiconductor competencies to match. This is not about single-product wins. It is about vertical integration at scale, reinforced by data advantages, fast iteration, and a domestic market big enough to validate new architectures. Huawei’s rise back to the top of China’s smartphone market, coupled with its chip self-reliance push, underscores that the country is rebuilding the compute stack under pressure. In energy, the same pattern is visible: deep manufacturing footprints, fast-deployment software, and exportable project finance. In that context, Dubhe looks less like a one-off launch and more like a platform move.
1) BYD (1211.HK) – The world’s largest electric carmaker in 2025. Scale in batteries, EVs, and buses positions BYD for vehicle-to-grid services as AI-driven power markets value flexible demand. Global impact: rapid EV exports to emerging markets create mobile storage pools aligned with AI energy orchestration.
2) CATL (300750.SZ) – Global leader in EV batteries and a rising force in grid-scale storage. Milestone: continued dominance in installed capacity makes CATL a key beneficiary of software that monetizes fast-cycling storage under AI-optimized dispatch.
3) LONGi Green Energy (601012.SH) – PV technology leader with 2024 R&D spend of ¥5.014 billion and 3,342 authorized patents, including 400 plus relating to back contact cells and modules. Global impact: bankable modules and technology depth enable Dubhe-like systems to scale across climates and grids.
4) Sungrow Power Supply (300274.SZ) – Inverter and energy storage systems provider connecting renewables to the grid. Global impact: grid-forming inverters and storage integration deployed across APAC, MENA, and Europe are essential plumbing for AI-coordinated power flows.
5) Goldwind (2208.HK) – Major wind turbine manufacturer with international project delivery. Global impact: onshore and offshore fleets become smarter assets as Physical AI optimizes yaw, curtailment, and maintenance scheduling to lift capacity factors.
6) XPeng (9868.HK) – Pure-EV brand that produced its 1,000,000th vehicle in November 2025. Milestone: a million connected EVs become a controllable load surface for demand response and, eventually, bi-directional charging under AI energy system rules.
7) Xiaomi (1810.HK) – With Xiaomi Auto rolling out 500,000 vehicles in less than 20 months since the SU7’s March 2024 debut, the ecosystem brings consumer electronics scale to smart mobility. Global impact: tight software-hardware integration supports grid-aware charging and over-the-air energy features.
8) Meituan (3690.HK) – Processed nearly 22 billion orders in 2023, underpinning world-class AI logistics. Global impact: electrified delivery fleets and routing algorithms become a flexible demand node and data source for urban energy optimization.
9) JA Solar Technology (002459.SZ) – Vertically integrated solar manufacturer with global footprint. Global impact: stable, high-efficiency module supply is a prerequisite for AI-optimized renewables to hit cost and reliability targets across markets.
10) Trina Solar (688599.SH) – Leading PV and storage systems provider. Global impact: utility-scale projects and integrated storage solutions align with AI dispatch strategies that monetize firmed solar across time and geography.
What Dubhe formalizes, China’s market has been assembling for years: a loop where better software drives higher asset utilization, which lowers LCOE, which unlocks new demand, which attracts more capex and data. EV fleets create controllable, distributed demand. Utility-scale batteries smooth intermittency and shift energy to peak pricing windows. Solar and wind keep pushing module and turbine costs down. Inverters and power electronics stitch the system together. Add weather-aware AI, and you get dispatch precision that improves revenue certainty for developers and lenders. That de-risks projects and speeds deployment. Investors with exposure across this stack benefit from compounding gains rather than isolated product cycles.
Masdar’s cooperation agreement with Envision is more than a photo op. It links Gulf capital and project expertise with China’s manufacturing and AI control systems. For governments managing fast-growing demand—think MENA, Southeast Asia, and parts of Africa—the AI energy system offers a path to add cheap, clean capacity without waiting for new gas pipelines or slow baseload construction. The export model is familiar: China supplies bankable hardware, software, and engineering; partners contribute land, financing, and offtake. The result is quick-to-market projects governed by algorithms that respect local grid realities and weather. Expect replication, standardization, and falling soft costs as templates harden.
Three markers will show Dubhe’s trajectory. First, live deployments with measurable gains—higher renewable curtailment reduction, better storage cycling economics, and improved grid reliability metrics. Second, ecosystem expansion—more OEMs and developers integrating APIs and control hooks so the Physical AI brain can steer assets beyond Envision’s own wind, storage, and hydrogen portfolio. Third, international standard-setting—interfaces and data schemas that let regulators, operators, and lenders audit AI decisions. On the market side, monitor capex guidance from CATL, LONGi, Sungrow, and Trina, EV production ramps at BYD, XPeng, and Xiaomi, and AI compute build-outs that convert into long-term power purchase agreements. Frictions remain—export controls on chips, interconnection queues, and local content rules—but the direction is unmistakable. China is packaging innovation and scale into a system investors can underwrite, and Dubhe is the latest proof point.
