Sungrow targets 2027 for volume production of solid-state transformers for AI data centers

China's Sungrow (阳光电源), one of the world's largest solar inverter manufacturers, says the explosive growth in AI compute is making solid-state transformers (SSTs) the inevitable power standard for data centers. The company made the claim in minutes from an investor meeting cited by 36Kr.

Why AI data centers need SSTs

According to Sungrow, the explosive growth in AI compute poses fundamentally new challenges for data center power architecture, and solid-state transformers (Solid State Transformer, SST) are the optimal infrastructure solution for scaling further. The key advantage of SSTs is that they eliminate energy losses at intermediate power conversion stages. Traditional multi-stage transformer setups lose some power at every step between voltage levels, and given how much energy modern AI clusters consume, those losses add up to a significant cost line and extra heat load.

• Overseas cloud service providers (CSPs) already rate Sungrow's SST products highly
• The company is in talks with leading North American cloud providers to jointly develop product specifications
• 2026: small batches of prototypes and testing with key customers
• 2027: first volume orders
• 2028 and beyond: transition to large-scale production for the next generation of compute chips

When to expect volume shipments

Sungrow plans to build out its own testing infrastructure to boost customer confidence in the reliability of its SST solutions. Under the announced timeline, 2026 will focus on small-batch samples and field trials with major customers. The first volume orders will follow in 2027, and from 2028 the company will move to large-scale production, timed to match the growing power demands of the next generation of compute chips.

Who Sungrow is

Sungrow is one of the world's largest manufacturers of solar inverters and energy storage systems, with a long track record of supplying power electronics to the energy sector. Its move into data center power infrastructure shows how quickly rising demand for AI compute is pushing energy equipment makers to expand their product lines well beyond traditional solar generation.

Power as AI's new bottleneck

Individual AI data centers now being built by hyperscale cloud providers already draw hundreds of megawatts and are approaching gigawatt scale — an order of magnitude more than typical data centers of the previous generation. At that scale, even a few percent of conversion losses translates into tens of megawatts of wasted power, plus extra heat that the cooling system has to remove. That is why power electronics — transformers, uninterruptible power supplies, power distribution systems — are becoming a focus of engineering innovation in their own right, on par with the accelerators themselves.

What it means

Energy infrastructure is becoming as much of a bottleneck for AI as the chips themselves. Power electronics makers like Sungrow are redirecting R&D toward data center needs, and the timeline for volume SST shipments — 2027-2028 — effectively lines up with forecasts for further compute buildouts at the leading cloud providers.