- Nvidia is seeking more H200 AI chip output for 2026 as Chinese demand jumps, sources say.
- Investors head into 2026 watching whether AI spending, earnings and rate cuts can justify valuations.
- Supply chains and U.S.-China policy shifts are emerging as day-to-day swing factors for AI-linked shares.
Nvidia has approached Taiwan Semiconductor Manufacturing Co to ramp up production of its H200 artificial intelligence chips after Chinese technology companies ordered more than 2 million units for 2026, sources told Reuters. The company holds about 700,000 units in inventory and expects TSMC to start work on expanded output in the second quarter of 2026, the sources said. The graphics processing units, or GPUs, are used to train and run AI models, but Beijing has yet to approve shipments even after Washington recently allowed H200 exports to China with a 25% fee, the sources said. Reuters
The scramble underscores why AI stocks enter 2026 in a more demanding phase: investors want proof that record spending on chips and data centres can turn into profits, and quickly.
U.S. stocks wrapped up a third straight year of double-digit gains in 2025, helped by enthusiasm around artificial intelligence and interest-rate cuts, but strategists say repeating that performance will require stronger earnings and sustained AI spending. Analysts at LSEG project S&P 500 earnings to rise more than 15% in 2026 after about 13% growth in 2025, while some strategists expect the market’s high valuations to limit upside, Reuters reported. Reuters
Policy decisions are also shaping the AI hardware trade. The U.S. government granted TSMC an annual licence to import U.S. chipmaking equipment into its Nanjing plant for 2026, and Samsung Electronics and SK Hynix received similar licences after a previous exemption regime expired at year-end, TSMC said. TSMC has said the Nanjing site makes mature-node chips and contributed about 2.4% of overall revenue in its 2024 annual report. Reuters
Valuation worries are creeping back into the narrative. Dennis Follmer, chief investment officer at Montis Financial, said AI-related stocks look like a bubble that may deflate through “a misallocation of capital.” Reuters
Constraints extend beyond Nvidia’s chips. The Financial Times reported that shortages of high-bandwidth memory, or HBM — specialised memory chips used in AI servers — could push some consumer electronics prices up to 20% in 2026 as suppliers prioritise AI orders. Samsung and SK Hynix have told customers that orders already exceed their capacity, the newspaper reported. Financial Times
Demand visibility in chips remains unusually strong, and investors are watching whether it spills over to rivals such as Advanced Micro Devices, which sells its own AI accelerators for data centres. Networking supplier Broadcom is also tied to data-centre build-outs that accompany the GPU boom.
For AI stocks more broadly, China is a swing factor that cuts both ways: new demand can lift forecasts, but export rules and local approvals can delay deliveries. That can show up quickly in quarterly guidance and cash flow.
Investors will scrutinize whether AI spending spreads earnings growth beyond the biggest tech names and into the wider market. The stakes are higher in 2026 because high expectations leave less room for missed timelines.
That shift could decide whether 2026 becomes a year of steady gains or sharp rotations within the AI complex. Hardware-linked stocks tend to move fastest when customers adjust spending plans.
In the near term, traders will watch supply-chain signals such as memory availability and foundry capacity, along with any new guidance on capital spending. Small changes matter when expectations are high.
The backdrop is supportive: AI demand is real, and companies keep spending. The 2026 forecast, however, looks less like a straight line and more like a test of earnings, supply and policy resilience.
