Inside Huawei's Plan to Build Its Own DRAM Factory During a Global Memory Crisis

Huawei's reported plan to build a 12-inch DRAM wafer fab with Shenzhen-based Swaysure and the municipal government is a strategic bid to insulate itself from the worst global memory shortage in history, bypass U.S. export controls that cut it off from Korean and American DRAM suppliers, and accelerate its pivot toward becoming China's most vertically integrated device manufacturer. While promising, the plan remains unconfirmed, faces significant technical limitations, and would supply only a fraction of Huawei's overall DRAM needs even at full production.

What We Know About the Huawei-Swaysure DRAM Fab

Huawei has reportedly teamed up with Swaysure (Shenzhen Yiweixu) — an entity with close Huawei ties — and the Shenzhen government to build a 12-inch (300mm) wafer fab dedicated to DRAM production . The planned capacity is 140,000 wafers per month, starting at the 28nm node . That volume would be meaningful enough to offset some of Huawei's DRAM import dependency, though it represents only a fraction of what the company consumes across its sprawling product lines.

Swaysure has recruited experienced semiconductor talent, including an ex-TSMC fab director to run the facility . The plan was first reported by industry tipsters in July 2026 and subsequently covered by tech outlets including Wccftech and Huawei Central . Crucially, Huawei has not officially confirmed the plan, and no groundbreaking or production timeline has been announced .

The Global Memory Shortage: Worst in History

The fab plan lands at a time when the memory industry is facing its worst-ever supply crunch. The scale is staggering:

DRAM Price Explosion

• DRAM contract prices surged 171.8% year-over-year as of Q3 2025 .
• In Q1 2026, prices jumped another 90–95% quarter-over-quarter .
• DDR5 server memory prices have tripled or quadrupled over the past year .
• TrendForce expects conventional DRAM contract prices to rise another 58% in Q2 2026 alone .
• DRAM revenue hit $165.7 billion in 2025, a 73% year-on-year increase .

AI Is Consuming ~70% of Memory Production
Analysts estimate that AI data centers will consume up to 70% of high-end memory production capacity in 2026 . Samsung, SK Hynix, and Micron have all shifted production lines away from consumer DRAM toward High-Bandwidth Memory (HBM) for AI GPUs, creating a supply vacuum for conventional DRAM . HBM now consumes 23% of total DRAM wafer capacity, up from single digits just two years ago .

SK Hynix Warns 2027 Will Be the 'Worst Year Ever'
On July 10, 2026, SK Hynix CEO Kwak Noh-jung warned that the memory industry is heading into its "worst-ever supply shortage" in 2027, and that demand will continue to outstrip supply well beyond 2030 — even with aggressive capacity expansion . He stated, "We expect next year to be the tightest supply year in the industry's history" . Samsung's memory chief Kim Jaejune separately warned in April 2026 of "significant shortages" across memory products continuing through at least 2027, with demand fulfillment rates at record lows .

How Huawei's Fab Fits Into the Shortage

DRAM is Huawei's critical vulnerability. U.S. sanctions block Huawei from buying top-tier DRAM from SK Hynix, Samsung, and Micron . The global shortage makes the spot market prohibitively expensive. A captive domestic DRAM supply is the only durable solution.

Timing matters: Even if the Swaysure fab takes 2–3 years to reach volume production, it would come online during the worst of the 2027–2030 shortage window that SK Hynix itself forecasts .

Entry point at 28nm: 28nm is a mature, widely available node that can produce DDR3/DDR4-class DRAM — not cutting-edge HBM, but sufficient for Huawei's base station equipment, networking gear, older smartphones, and IoT products. This frees up higher-end global DRAM supply for Huawei's premium products.

Huawei's Broader Fab Empire and 'Chinese Samsung' Ambitions

The DRAM fab is just one piece of a much larger strategy. According to South Korean media reports, Huawei is directly or indirectly operating at least seven semiconductor manufacturing companies in China, with at least 11 fabs already in operation . These cover logic chips, AI accelerators, and now memory.

Three fabs in Shenzhen's Guanlan district: One for 7nm smartphone/Ascend chips (Huawei's own), one run by SiCarrier (a state-backed equipment maker spun out of a Huawei lab), and a third — the Swaysure DRAM fab . Satellite imagery from early 2025 shows rapid construction progress at these sites .

LogicFolding breakthrough: In May 2026, Huawei unveiled a novel chipmaking method called "LogicFolding" to shorten the gap with TSMC, aiming to produce 1.4nm chips by 2031 using its own equipment .

AI chip production ramping: Huawei is scaling Ascend AI chip output — roughly 250,000 Ascend 910Cs in 2025, targeting 1.6 million total dies across the Ascend line in 2026 .

Vertical integration goal: Huawei wants to control the full stack — design, fabrication, packaging, and now memory — mirroring Samsung's integrated device manufacturer (IDM) model .

China's Push for Domestic Memory Self-Sufficiency

The Swaysure DRAM fab operates within China's broader "Triple Output" AI Strategy — a state mandate to triple domestic AI processor production by end of 2026 . Key context:

• China's memory self-sufficiency rate remains below 5% for DRAM. Longson (CXMT) in Hefei is the only domestic DRAM producer of scale.
• The Shenzhen government's financial backing of the Swaysure fab signals that the state is willing to fund multiple parallel memory projects to reduce dependency on Korean and American suppliers.
• U.S. export controls (curbs on ASML lithography, KLA inspection tools, and EDA software) make it difficult for any Chinese DRAM fab to move beyond 28nm nodes, but even 28nm DRAM is strategically valuable in a market where prices have quadrupled.

Key Caveats and Risks

• The Huawei–Swaysure fab plan is sourced from industry tipsters and tech media, not from an official Huawei announcement. It should be treated as credible reporting but unconfirmed .
• Starting at 28nm means the fab will not produce the advanced HBM3/HBM4 that drives AI GPU demand — it will address the conventional DRAM shortage, not the premium AI memory segment.
• Even at 140,000 wpm, the fab would supply only a fraction of Huawei's own DRAM needs, let alone China's broader demand.
• Building a DRAM fab is technically complex and capital-intensive. Historical precedents (e.g., Fujian Jinhua's troubled efforts) show that Chinese DRAM projects face persistent challenges with equipment sanctions, yield rates, and talent retention.

Bottom line: Huawei's DRAM fab plan is a defensive offensive — it targets the intersection of the worst memory shortage in history, U.S. sanctions that cut Huawei off from global DRAM suppliers, and China's strategic imperative to build domestic memory capacity. If successful, it would transform Huawei from a company that designs chips but buys memory into a fully integrated device manufacturer, while giving China a second domestic DRAM production base at a time when global memory prices have quadrupled and supply will remain critically tight through at least 2030. But the path is fraught with technical, political, and execution risks that make the outcome far from certain.