Nvidia’s China challenge and Türkiye’s place in chip manufacturing

The United States opened the door to conditional sales of Nvidia’s H200 AI accelerator to China. Beijing’s import restrictions and new findings on how Chinese institutions access the chips show the limits of Washington’s licensing policy. In Türkiye, the move from low-volume production to industrial scale depends on turning the 110 nm and 65 nm targets into operating capacity.

China’s Northwestern Polytechnical University bought two Nvidia H200 chips in October while on a US export-control list. According to procurement records reviewed by Reuters on 11 November, the purchase from a supplier in China was valued at 874,700 yuan. The records also include contracts under which the university rented computing capacity on servers powered by Nvidia hardware. They offer little information about how the chips first entered China. [1]

The MegaSpeed allegations and Nvidia’s response

In its 11 November report, Le Monde examined Singapore-based cloud company MegaSpeed through its business ties with Nvidia and Chinese technology companies. The Information, as reported by Reuters, alleged that companies linked to former MegaSpeed executive Alice Huang helped about 240,000 Nvidia chips reach China. The report alleged that some of the hardware served Alibaba and Tencent. Reuters said it could not independently verify the claims. [2][3]

Singapore’s Ministry of Home Affairs said the investigation into MegaSpeed was ongoing and denied reports that Singaporean authorities had conducted a separate investigation into Alice Huang. On 8 November, Nvidia’s board launched an independent inquiry following reports that CEO Jensen Huang had been the target of a foreign intelligence operation. The board said it had no reason at that stage to believe that company secrets or US national security had been compromised. On 10 November, the independent board members appointed outside counsel to conduct the inquiry. [4][5][6]

The two cases illustrate different aspects of enforcement. In a hardware sale, the buyer’s identity and the product’s destination are central. With a cloud service, regulators also need to know who is using the same hardware, from where and for what purpose. The January 2026 US licensing rule also requires safeguards against restricted entities and military end users remotely accessing H200 chips exported under the policy. [1][10]

Washington grants licences as Beijing limits purchases

In October 2022, the United States announced a package of controls restricting China’s access to advanced computing chips and the equipment used to make them. In December 2024, the scope widened to cover certain manufacturing equipment and software as well as high-bandwidth memory (HBM). Under a rule that took effect on 15 January 2026, licence applications to export the Nvidia H200, AMD MI325X and similar products to China and Macau are reviewed case by case, subject to conditions. These include protecting manufacturing capacity available to US customers, the buyer’s security and compliance procedures, and independent testing. Separate restrictions still apply to organisations on the US Entity List. [1][10][11][12][13]

According to Nvidia’s filing with the US Securities and Exchange Commission (SEC) for the quarter ended in late July, US authorities began granting licences in February 2026 for limited H200 sales to specific Chinese customers. Nvidia said restrictions in China meant it completed only a portion of the authorised shipments. Those shipments accounted for less than 1% of its data centre revenue that quarter. The company recorded a $400 million charge for H200 inventory and purchase commitments in the first half of fiscal 2027. It recorded a similar charge of $4.5 billion for the H20 in the first quarter of the previous fiscal year. [14]

On 5 November, Reuters cited two people familiar with the matter as saying that about 600,000 Nvidia AI chips were being held in Hong Kong pending approval for shipment to mainland China. Beijing’s decisions on imports can therefore determine whether a US export licence translates into a sale. Beijing is encouraging the use of domestically developed accelerators, and HBM supply remains a bottleneck for manufacturers such as Huawei and Cambricon. Reuters reported that higher memory costs had pushed up the prices of accelerators made by Chinese suppliers. [7][8][14]

The production chain behind an AI chip

Nvidia’s Blackwell graphics processors are made using the 4NP process of Taiwan-based semiconductor foundry TSMC. High-bandwidth memory, advanced packaging and software help turn those accelerators into systems that can run in data centres. Designing chips, fabricating them on silicon wafers, packaging and testing them, and building servers with imported accelerators each develop different capabilities along the value chain. [8][9][15]

Türkiye aims to bring chip production to industrial scale

The Semiconductor Technologies Research Laboratory (YİTAL) at TÜBİTAK BİLGEM can produce low volumes of integrated circuits using 250 nm CMOS and silicon-germanium (SiGe) processes. Its facilities cover design, mask preparation, wafer fabrication and testing, packaging, and final testing. BİLGEM’s TÜTEL laboratory also provides design training, prototyping and testing support. ASELSAN’s filings on Türkiye’s Public Disclosure Platform (KAP) list TÜYAR in semiconductors and packaging, and ASELSAN Bilkent Mikro Nano in wide-bandgap semiconductors and photonics. [16][17][18]

Türkiye’s 2027–2029 Medium-Term Programme calls for new facilities to make chips for identity cards, passports and industrial applications. The target is to move from the current 250 nm capability to 110 nm. At his 2 November budget presentation, Industry and Technology Minister Mehmet Fatih Kacır said construction had begun on a new chip plant. A separate investment call, HIT-Chip, has a $5 billion support budget and aims to establish capacity for at least one million silicon wafers a year using 65 nm or more advanced processes. The call also covers investment in packaging and testing. [15][19][20][21]

Trade figures give a sense of the scale of the production challenge. According to UN Comtrade data available through the World Bank’s WITS platform, Türkiye imported about $913 million in digital monolithic integrated circuits in 2025 and exported $43.8 million. Imports were about 21 times exports. The figures cover WITS subcategory 854211. Chips embedded in devices and other semiconductor categories fall outside those figures. [22][23]

Corporate disclosures also show domestic demand for advanced computing hardware. In a 9 November KAP filing, ARD Grup reported a $4.53 million order, including VAT, from a Turkish company for AI infrastructure based on Nvidia GPUs. The order points to demand for advanced computing infrastructure in Türkiye. Publicly available manufacturing and investment documents show no domestic mass-production capacity for data centre accelerators in the same class as Nvidia’s. [15][16][19][24]

Automotive control units, appliance motor drives, power management circuits and sensors offer clear markets for chips made in Türkiye. In those markets, commercial success will depend on production yield, reliability testing, lead times and long-term purchase agreements as much as on the process node. Regular production and sales of qualified components will show whether the incentives have translated into industrial capacity. [15][20]

Sources

Nvidia and China

01 Reuters, 11 November 2026. Procurement records for a Chinese university’s H200 purchase

02 Le Monde, 11 November 2026. Nvidia, China and the race for AI chips

03 Reuters, 10–11 November 2026. Nvidia and allegations of chip shipments to China

04 Singapore Ministry of Home Affairs, 31 October 2026. Statement on the MegaSpeed investigation and Alice Huang

05 Nvidia, 8 November 2026. Board statement on the independent inquiry

06 Nvidia, 10 November 2026. Appointment of outside counsel for the independent inquiry

07 Reuters, 5 November 2026. Beijing’s import approvals and Nvidia chips held in Hong Kong

08 Reuters, 10 September 2026. Chinese AI chipmakers and the cost of high-bandwidth memory

09 Nvidia. Blackwell architecture and TSMC’s 4NP process

US regulations and corporate filings

10 Federal Register, 15 January 2026. Licence review policy for certain advanced computing chips

11 US Bureau of Industry and Security, 7 October 2022. Initial controls on advanced computing and semiconductor manufacturing equipment

12 US Bureau of Industry and Security, 2 December 2024. Controls on manufacturing equipment, software and high-bandwidth memory

13 US Bureau of Industry and Security, 13 January 2026. Licensing conditions for the H200, MI325X and similar chips

14 Nvidia, SEC Form 10-Q. Quarter ended 26 July 2026 and disclosures on the H20 and H200

Manufacturing, investment and trade in Türkiye

15 Ministry of Industry and Technology. HIT-Chip investment call

16 TÜBİTAK BİLGEM. YİTAL’s fabrication processes and laboratory facilities

17 TÜBİTAK BİLGEM. TÜTEL support for design, prototyping and testing

18 KAP, ASELSAN. Business activities of TÜYAR and ASELSAN Bilkent Mikro Nano

19 Presidency of Strategy and Budget, September 2026. Semiconductor manufacturing targets in the 2027–2029 Medium-Term Programme

20 TRT Haber / Anadolu Agency, 13 September 2026. Identity and passport chips and the 110 nm manufacturing target

21 Mehmet Fatih Kacır, 2 November 2026. New chip plant discussed in the 2027 budget presentation

22 World Bank WITS / UN Comtrade. Türkiye’s imports of digital monolithic integrated circuits in 2025

23 World Bank WITS / UN Comtrade. Türkiye’s exports of digital monolithic integrated circuits in 2025

24 KAP, ARD Grup, 9 November 2026. Order for Nvidia-based AI infrastructure