Overview
At the 2026 ninth International Conference on Photovoltaic and Energy Storage Industries, breaking through power-compute coordination became a key breakout-session theme. “The end of computing is electricity, not electricity price,” one participant noted — but fusing the two is a system-level undertaking requiring generators, users, grid operators, computing firms and equipment makers to move from designing separately to designing jointly. As the discussion put it, “a good solution is compared side by side, not spliced together from individually optimal parts.”
Liu Yi, deputy director of the Tsinghua Sichuan Energy Internet Research Institute, described three mismatches. On construction timing, computing firms wait for confirmed demand before building — by which time power cannot catch up; building power early lacks a computing-side commitment. On iteration, computing equipment, power density and business plans keep changing while power supply, cooling and buildings are long-lived — under-provisioning cannot keep pace, over-provisioning idles capital. On payback, computing ramp-up timing and scale can shift, while power assets require upfront investment recovered over years; if a project is delayed or exits, who bears the dedicated investment becomes a problem.
Background & Interpretation
1. Industry background and policy context
Exponential AI demand is driving a data-center construction wave, making grid stability and efficiency a binding constraint. Power-compute coordination is now in national planning: the computing network is one of the country’s six planned networks, and the 15th Five-Year carbon-peaking action plan calls for greener computing facilities and tighter coordination with renewables. Industry practice is moving from point solutions to systems — the first national standard on power-compute coordination and power-cabin procurement shifting toward system integration both point the same way: the unit of competition is rising from a single chip or machine to systems and clusters.
2. Core drivers and mechanisms
The mismatch stems from different asset lives and decision logics. Computing hardware turns over in years; power infrastructure depreciates over decades. Computing pursues fast deployment and flexibility; power seeks long-term certainty. Liu proposes agreeing five things in advance: how load ramps, in how many phases capacity comes online, how much capacity is committed, what triggers expansion, and how exit is handled — “so early construction has a basis and risks are shared, rather than borne entirely by one side.” On evaluation, he argues for accounting for capacity, efficiency, cost and resilience together: capacity means effective compute supported, not equipment installed; efficiency means resources per unit of service quality; cost means the full bill including compliance and risk; resilience means continuity and recovery under disturbance.
3. Market structure and value-chain impact
Joint design reshapes roles across the chain. Grid operators need clearer load forecasts to justify investment; computing firms need earlier connection certainty to protect timelines; equipment makers must offer flexibly scalable power and cooling systems. The trend favors power-equipment integrators, load-aggregation and virtual-power-plant platforms, and research institutions offering simulation and design tools. For renewable-rich regions, a clear risk-sharing framework helps convert resource advantages into compute investment appeal.
Implications & Outlook
Near term, the debate is shifting from electricity price to solution quality, with joint design likely piloted in new data-center projects. Medium-term variables include whether ramp schedules and capacity commitments enter contracts, whether coordination standards and simulation tools mature, whether cost-sharing mechanisms for dedicated power assets are established, and how fast source-grid-load-storage models replicate in computing scenarios. If incentives align, the timing mismatch between compute rollout and power build-out should ease significantly.
Takeaways for Industry Participants
Computing firms should write grid connection, phased rollout and exit terms into agreements during planning to avoid sunk costs when demand shifts. Power companies can offer load forecasting and flexible expansion as value-added services. Equipment and engineering firms should move from component supply toward system integration, building capabilities in modular power distribution, liquid cooling and heat recovery. The shared goal is earlier commissioning, steadier operation and lower total cost.
This column compiles industry information and shares technical perspectives; it does not constitute investment advice.