Hua Hong Joins SMIC at 7nm: China Has Two Advanced Chipmakers Now

China just broke SMIC's monopoly on advanced domestic chip production. Hua Hong Group's manufacturing arm, Huali Microelectronics, is readying 7nm production at its Fab 6 facility in Shanghai — making it only the second Chinese chipmaker after SMIC capable of sub-10nm work. US-blacklisted GPU designer Biren is already using the line for tape-out testing. If you thought the chip war was a two-player game, the board just changed.

What Hua Hong Actually Is (And Why It Matters)

Hua Hong Group is China's second-largest chipmaker by revenue, majority state-owned, and has historically been known for specialty chips — power semiconductors, MCUs, NOR flash — not the advanced logic nodes that grab headlines. Its Huali Microelectronics subsidiary runs Fab 6 in Shanghai's Lingang Special Economic Zone, a facility that was processing 22nm and 28nm wafers until recently.

The jump to 7nm is not a minor process improvement. It is a generational leap that requires fundamentally different lithography approaches, process chemistry, and metrology tooling. Until now, SMIC was the only Chinese foundry that had demonstrated 7nm capability — and it does so at extremely limited volume, constrained by the equipment it was allowed to import before US export controls tightened in 2022 and 2023.

A second foundry at 7nm means China has redundancy in its advanced node production. In semiconductor manufacturing, single points of failure are existential risks. SMIC going offline — from sanctions, equipment failure, or supply disruption — would have crippled Chinese access to domestic advanced chips. That single-point dependency no longer exists.

Biren: The US-Blacklisted GPU Company Already Testing There

Biren Technology is the clearest signal of what this foundry capability is actually for. Biren is a Shanghai-based GPU designer that makes AI training chips — the kind that compete with Nvidia's A100 and H100 in compute density. The US Commerce Department added Biren to the Entity List in October 2022, blocking it from purchasing Nvidia hardware directly and restricting its access to advanced US chip manufacturing equipment.

Being on the Entity List does not stop Biren from designing chips. It stops Biren from having those chips manufactured at TSMC, Samsung, or GlobalFoundries — the three leading foundries that have operations and equipment subject to US jurisdiction. Before Hua Hong's 7nm line, Biren's primary domestic option was SMIC, which was already operating at capacity and facing its own equipment constraints.

Biren is now confirmed to be conducting tape-out testing at Huali Fab 6. A tape-out is the final design file submission before a chip goes into production — it is the last step before actual silicon. This is not exploratory. Biren is planning production runs at Hua Hong.

The practical consequence: a US-blacklisted AI chip company now has a second domestic foundry option for advanced GPU production. The export controls designed to limit Chinese AI chip development just got a workaround that does not require smuggling or shell companies.

SiCarrier and the Domestic Lithography Play

The lithography question is the one that China-watchers focus on most. ASML's EUV machines — the tools required to pattern chips below 7nm using standard approaches — are completely off-limits to Chinese foundries. The US-led campaign to block ASML EUV exports to China has been one of the most consequential technology control efforts of the past decade.

Hua Hong's 7nm approach does not use EUV. It uses multiple-patterning DUV lithography — running older-generation Nikon and Canon scanners through multiple exposure passes to achieve feature sizes that would normally require EUV. SMIC proved this was possible with its 7nm FinFET process (the one found in early Huawei Mate 60 Pro chips). It is slower, more expensive per wafer, and harder to scale. But it works.

What is new is the equipment enabling it. SiCarrier, a Huawei-backed semiconductor equipment company, has been testing enabling equipment at a Shenzhen facility that is designed to improve DUV multi-patterning yield and throughput. SiCarrier is explicitly developing equipment as a domestic substitute for tools that Chinese fabs can no longer import. Korean semiconductor analysts, citing SiCarrier's trajectory, have projected a "Chinese ASML" by 2030.

That projection may be optimistic. But the direction is clear: China is not waiting for sanctions to be lifted. It is building the equipment supply chain in parallel.

Scale: What "Several Thousand Wafers Per Month" Actually Means

Hua Hong targets several thousand wafers per month at 7nm by end of 2026. To put that in context: TSMC processes hundreds of thousands of wafer starts per month across all nodes. Even TSMC's advanced 3nm and 5nm lines, which represent a fraction of its total capacity, process tens of thousands of wafers monthly.

Several thousand wafers per month from Hua Hong is not threatening TSMC's market position. But that misses the point. The goal is not global semiconductor market share. The goal is producing enough advanced chips to serve Chinese AI infrastructure — specifically the data center GPU market that Nvidia currently cannot legally sell into.

China's broader semiconductor plan, reported in February 2026, targets a fivefold increase in 7nm and 5nm output within two years. That plan requires SMIC and Hua Hong to both scale simultaneously. Hua Hong's readiness at Fab 6 is a prerequisite for that plan, not an isolated announcement.

What This Means for China's AI Independence

The China AI infrastructure story has always had a hardware gap at its center. Chinese AI labs — DeepSeek, Baidu, ByteDance, Alibaba — have been extraordinarily good at doing more with less: optimizing model architectures, reducing inference costs, running training on older Nvidia hardware before export controls closed that window. DeepSeek V4's decision to target Huawei Ascend hardware was the clearest signal that China's top labs are committing to the domestic chip stack.

But committing to domestic hardware only makes strategic sense if domestic hardware production can scale. Huawei's Ascend chips are already being manufactured at SMIC. Adding Hua Hong as a second foundry option for AI-grade silicon — whether for Ascend, Biren, or future designs — is what makes the domestic stack viable as infrastructure rather than a stopgap.

For developers building AI applications in China, this matters in a specific way: it affects the compute pricing floor. If domestic GPU supply increases, cloud inference prices fall and training access broadens. Chinese cloud providers — Alibaba Cloud, Tencent Cloud, Huawei Cloud — compete on AI instance pricing. More domestic chip supply equals more competitive AI compute.

The Export Control Feedback Loop

The US has been tightening chip export controls since October 2022, with major expansions in October 2023 and further restrictions through 2025. The stated goal is to slow China's AI development by denying access to advanced chips and the equipment to make them.

The Hua Hong development is evidence of the feedback loop that critics of export controls have flagged: denial of access accelerates domestic development in a way that open access would not have. China would not have invested the billions required to develop domestic lithography equipment, train foundry engineers for advanced nodes, and build a parallel chip supply chain if TSMC and ASML were available alternatives. The controls forced the investment.

That does not mean the controls were the wrong policy call. It means the timeline assumptions behind them may have been too optimistic. ByteDance's 36,000 Nvidia B200 chips routed through Malaysia showed one workaround. Hua Hong's 7nm line shows another — one that does not require creative logistics at all.

Key Takeaways

• Hua Hong's Huali Fab 6 in Shanghai is readying 7nm production, ending SMIC's status as China's only advanced-node foundry
• Biren Technology, US-blacklisted since October 2022, is already conducting tape-out testing at Huali — planning production, not experiments
• SiCarrier (Huawei-backed) is developing domestic equipment to improve DUV multi-patterning yield, substituting for ASML tools China cannot import
• Target output: several thousand wafers per month by end of 2026 — not TSMC-scale, but enough for Chinese AI chip priorities
• China's broader plan targets fivefold increase in 7nm/5nm output within two years — Hua Hong is a required component of that plan
• The export control feedback loop: restrictions designed to slow Chinese AI chip access have accelerated domestic foundry and equipment investment that would not have happened otherwise