On September 23, 2026, the Financial Times reported that China's State-owned Assets Supervision and Administration Commission (SASAC) had spent recent weeks surveying how extensively Broadcom's networking switches are deployed across state-owned data centers. The preliminary finding: Broadcom switches could account for as much as 90 percent of the equipment in use in some of those facilities.
SASAC is not a regulator with procurement authority. It oversees China's largest state-owned enterprises. The survey is an inventory — a count of how much of the networking layer inside China's state-backed AI infrastructure depends on a single American supplier.
That count matters because of what switches do. In an AI cluster, switches are the traffic controllers. They route data between thousands of GPUs during model training. A single high-radix switch can carry 51.2 terabits per second of traffic. If the switches stop working, the GPUs sit idle.
How Broadcom Got to 90 Percent
Broadcom's dominance in data center switching did not happen overnight. Its Tomahawk and Trident series of merchant switch chips hold approximately 70 to 80 percent of the global market for commercial switch silicon, according to Dell'Oro Group. That share was built over more than a decade through design wins at Cisco, Arista Networks, and the largest cloud operators, including Microsoft and Meta.
In China, the dependency is even more concentrated. A survey of investor relations platforms in September 2026 noted that H3C (New H3C) , the networking vendor formed from the HPE-Inspur joint venture, is the largest domestic purchaser of Broadcom switch chips and the only Chinese company authorized to buy directly from Broadcom. Every other Chinese vendor buys through distributors, which affects allocation priority and pricing.
The result is a networking layer where the dominant supplier is American, the largest domestic buyer is a joint venture with American heritage, and the alternative options are one to two generations behind on specifications.
What the Alternatives Actually Offer
The list of potential domestic beneficiaries is limited to three companies: Huawei, ZTE, and Centec Networks.
Huawei has been selling high-end switches in China for years. Its CloudEngine XH9230-128DQ, launched in 2024, was the first domestic standard-form switch to carry Huawei's self-developed Pacific X650 51.2T switching chip, with 128×400G ports. At Huawei Connect in September 2026, the company announced the CloudEngine XH9300 series, a 100T/51.2T NPO switch using a proprietary 3.2T optical engine that reduces interconnect power draw by 40 percent.
ZTE's Lingyun 51.2T switch chip is described as one of China's first commercial 51.2T AI switch chips. It uses a 7nm chiplet process, 36×800G ports, and native RoCEv2 lossless networking, designed for ten-thousand-card Scale-up supernodes. ZTE says the chip has completed validation testing with Alibaba, ByteDance, and Tencent, and plans to begin volume external supply in Q3 2026.
Centec Networks, based in Suzhou, is the only Chinese company in the global top six for Ethernet switch chips, with a 2.3 percent share of the global market in 2024, according to QYResearch. Its current revenue comes primarily from 25.6T chips, which are in mass production. The 51.2T chip had not yet returned from fabrication as of August 2026, with revenue contribution expected in the second half of 2027.
The gap is measurable. Broadcom's Tomahawk 5 uses a 5nm process and delivers 51.2 Tbps. Broadcom's Tomahawk 6, announced in June 2025, delivers 102.4T and uses 3nm process technology with co-packaged optics. No Chinese merchant chip has reached volume production at 51.2T. Centec's highest volume product is 25.6T. ZTE's Lingyun 51.2T is entering volume supply in Q3 2026. Huawei's Pacific X650 is in production but is used exclusively in Huawei's own switches, not sold as a merchant chip.
For state-owned data center operators — China Telecom, China Mobile, and China Unicom — replacing Broadcom switches means accepting lower port density, higher power consumption per terabit, or both. The three carriers have already cut their 2026 capital expenditure plans.
What Broadcom Actually Loses
The revenue impact on Broadcom is smaller than the headline number suggests.
China accounted for approximately 30 percent of Broadcom's total revenue in fiscal 2024, according to company filings. But most of that was custom AI accelerators for ByteDance and other non-state customers, not switches. Broadcom's networking business represents a mid-single-digit percentage of its China revenue. Even a full ban on Broadcom switches in the state sector would reduce group revenue by a low-single-digit percentage.
The market reaction was consistent with that arithmetic. Broadcom's stock fell more than 1 percent on September 24 as investors weighed the news, closing at $350.36, according to market data.
The real risk is not the state sector. It is the precedent. If informal guidance hardens into formal procurement rules, the same logic could extend to Broadcom's custom AI accelerator business — the part of its China revenue that actually matters.
The Broader Pattern
The SASAC survey is part of a broader audit of foreign technology in China's AI infrastructure stack. Nvidia's accelerators and networking products are already barred from state-backed data centers. Broadcom's switches remained widely deployed — until now. The survey appears designed to close that gap under the "domestic chips for domestic use" campaign, which aims to expand the adoption of Chinese-made semiconductors across the public sector.
For Broadcom, the immediate financial exposure is limited. For the Chinese operators being asked to switch, the cost is real: lower performance, higher power, and a procurement cycle measured in years, not quarters. For the domestic chipmakers, the opportunity is significant but constrained by the same physics that limit every advanced chip in China — process nodes, yield, and time.
The survey is an inventory count, not a ban. What happens after the count depends on how quickly the alternatives can close the gap.
Sources: Financial Times via The Economic Times (September 23, 2026); Lianhe Zaobao (September 23, 2026); MoneyDJ (September 23, 2026); Edgen.tech (September 23, 2026); Investing.com (September 23, 2026); Benzinga (September 24, 2026); C114 (July 1, 2026); Huawei official announcement (May 22, 2026); Huawei Connect 2026 coverage (September 21, 2026); STCN (August 31, 2026); Sohu (August 7, 2026); Zhengguan News (July 15, 2026); Sina Finance (May 22, 2026); Chinese tech community coverage (September 23, 2026).
Disclaimer
The information provided in this article is for general informational and educational purposes only. It does not constitute legal, financial, or professional advice. The author and publisher are not responsible for any actions taken based on the content of this article. Readers should consult qualified professionals for advice specific to their situation. All trademarks and references to third-party products, services, or organizations are the property of their respective owners. The performance data and benchmarks discussed are based on specific research studies and may not generalize to all use cases or environments. As of the publication date, the AI landscape continues to evolve rapidly, and readers should verify current information independently.
Limitations
This analysis is based on reporting and public data available as of the article date; figures may be revised as sources update.
Forecasts from third-party analysts can change with market conditions.
Cost and pricing examples are point-in-time estimates; actual rates vary.
Country and company comparisons rely on public reporting, not operational data.
This sector moves fast; timelines and deal terms may be updated later.
Company deals and regulatory rulings may evolve; verify current status.
AI infrastructure is changing quickly; claims can become outdated soon.
Sources
- Financial Times via The Economic Times (September 23, 2026)
- Lianhe Zaobao (September 23, 2026)
- MoneyDJ (September 23, 2026)
- Edgen.tech (September 23, 2026)
- Investing.com (September 23, 2026)
- Benzinga (September 24, 2026)
- C114 (July 1, 2026)
- Huawei official announcement (May 22, 2026)
- Huawei Connect 2026 coverage (September 21, 2026)
- STCN (August 31, 2026)
- Sohu (August 7, 2026)
- Zhengguan News (July 15, 2026)
- Sina Finance (May 22, 2026)
- Chinese tech community coverage (September 23, 2026).
The information provided in this article is for general informational and educational purposes only. It does not constitute legal, financial, or professional advice. The author and publisher are not responsible for any actions taken based on the content of this article. Readers should consult qualified professionals for advice specific to their situation. All trademarks and references to third-party products, services, or organizations are the property of their respective owners. The performance data and benchmarks discussed are based on specific research studies and may not generalize to all use cases or environments. As of the publication date, the AI landscape continues to evolve rapidly, and readers should verify current information independently.
Limitations: This analysis is based on reporting and public data available as of the article date; figures may be revised as sources update.; Forecasts from third-party analysts can change with market conditions.; Cost and pricing examples are point-in-time estimates; actual rates vary.; Country and company comparisons rely on public reporting, not operational data.; This sector moves fast; timelines and deal terms may be updated later.; Company deals and regulatory rulings may evolve; verify current status.; AI infrastructure is changing quickly; claims can become outdated soon.