The US government’s tightening of semiconductor export controls and persistent supply-chain bottlenecks are forcing Nvidia to navigate a complex operating environment even as it pushes ahead with next-generation hardware. The Bureau of Industry and Security updated its chip export guidelines on 31 May 2026, closing a loophole that allowed Chinese companies to access restricted processors through international subsidiaries. The new rules require export licenses for Nvidia’s Blackwell and Vera Rubin architectures when the parent company is based in China or Macau, regardless of where the subsidiary sits. Data centres in Singapore and Malaysia are most exposed — industry estimates suggest “hundreds of thousands” of high-end chips have flowed through those channels since May 2025. Nvidia itself does not expect a material impact on current business, and existing data centres will not be forced to shut down.
Yet while regulators clamp down on one front, Nvidia and its manufacturing partner TSMC have started shipping a technology designed to unclog a different bottleneck inside AI factories. The pair have officially begun deliveries of their jointly developed Spectrum-X co-packaged optical (CPO) switches. Built on TSMC’s COUPE platform, these switches integrate optical modules directly into the chip, cutting electrical paths and slashing latency while reducing power consumption compared with traditional pluggable modules. The hardware achieves a data throughput of up to 400 terabits per second. Select partners are receiving the first units now, with production volumes expected to ramp significantly in the second half of 2026. Nvidia believes the switches will improve profitability for operators of large-scale AI facilities.
The optical-switch rollout is only one piece of a broader infrastructure push. CEO Jensen Huang used the Computex stage to publicly forecast that Marvell Technology could eventually become a trillion-dollar company. Marvell, a specialist in networking solutions that complement Nvidia’s GPU clusters, received a roughly $2bn investment from Nvidia in March to accelerate development of the NVLink Fusion platform. Two more players have now joined the ecosystem. Zhen Ding Technology will supply board-based wireless connections that replace traditional copper cables inside server racks, while Lightmatter is contributing new optical products that can halve the fibre requirements inside data centres.
Separately, demand for Nvidia’s next-generation platform is already showing real momentum. Digi Power X committed $35m to Vera Rubin systems on 3 June. The Rubin GPUs feature 288 GB of HBM4 memory and the Vera CPU packs 88 cores, with first deliveries scheduled for the first quarter of 2027. CoreWeave announced it had become the first cloud provider to fully validate the Vera Rubin NVL72 platform, a system joining 72 Rubin GPUs and 36 Vera CPUs via NVLink 6, designed for the most demanding AI workloads.
Should investors sell immediately? Or is it worth buying Nvidia?
TSMC’s CEO C.C. Wei confirmed on 4 June that the global supply of high-performance chips will continue to lag behind AI-driven demand for years, despite expanded production in the United States. On top of that, memory-chip shortages are adding to the pressure. Nine industry associations warned the US government on 3 June of significant risks to automotive and medical-device supply chains, and SK hynix expects memory constraints to persist until 2030.
The stock market’s reaction to these cross-currents has been mixed. Nvidia shares gave up 3.6% on 3 June, closing at $214.75 on the US exchange, as a broad market rotation and insider sales of roughly 906,000 shares over the past 90 days weighed on sentiment. In European trading, the stock was quoted at €183.00 with a slight daily loss, though it still shows a solid year-to-date gain of around 13%. The relative strength index sits near 50, a neutral reading, and the share price is roughly 10% below its recent all-time high. The consensus price target remains $305.38.
For income-focused investors, the ex-dividend date for Nvidia’s next payout falls on 4 June 2026. Shareholders on record will receive $0.25 per share, paid out on 26 June, while the company’s $80bn share buyback programme continues to run in the background.
Ad
Nvidia Stock: Buy or Sell?! New Nvidia Analysis from June 4 delivers the answer:
The latest Nvidia figures speak for themselves: Urgent action needed for Nvidia investors. Is it worth buying or should you sell? Find out what to do now in the current free analysis from June 4.
Nvidia: Buy or sell? Read more here...
