NEW YORK, January 2, 2026, 07:20 ET
- Nvidia has sought more H200 chip output after Chinese tech firms placed orders topping 2 million units for 2026, sources said.
- Samsung says it is in close talks to supply next-generation HBM4 memory to Nvidia as the AI hardware race intensifies.
- Strategists say 2026 gains hinge on whether AI infrastructure spending translates into profits, keeping Nvidia in the spotlight.
Nvidia has approached Taiwan Semiconductor Manufacturing Co to ramp production of its H200 artificial-intelligence chips after Chinese tech companies ordered more than 2 million units for 2026, sources said. Nvidia has about 700,000 units in stock and has priced China-bound H200 variants around $27,000 per chip, the sources said. TSMC is expected to start work on expanded output in the second quarter, one person said, as Beijing weighs whether to approve imports after Washington recently cleared H200 exports to China with a fee.
The scramble for supply lands as investors reset portfolios for 2026 after the S&P 500 rose more than 16% in 2025, powered in part by AI-linked megacaps such as Nvidia. Strategists say the next leg depends on whether corporate capital spending — known as capex — keeps delivering returns. “If companies start to pull back on the capex … you’re probably looking at more of a flat or even a modestly down year,” said Jeff Buchbinder, chief equity strategist at LPL Financial. Reuters
Nvidia’s chips sit at the heart of the AI boom, supplying the computing horsepower used to train and run large models. That makes its order book and supply chain an early read on whether big spending plans by cloud firms and enterprises are holding up.
The China push also highlights a familiar pressure point for the sector: policy and capacity can matter as much as demand. A sudden change in export approvals, or a bottleneck at a key supplier, can ripple through the AI hardware stack and into earnings expectations.
Samsung Electronics, one of the world’s biggest memory makers, said customers have praised the competitiveness of its next-generation high-bandwidth memory chips, known as HBM4, and reiterated it is in close talks to supply the product to Nvidia. Rival SK Hynix led the HBM market in the third quarter of 2025 with 53% share, ahead of Samsung at 35% and Micron at 11%, Counterpoint Research data cited by Reuters showed.
HBM is a type of stacked memory designed to sit close to processors, helping feed data fast enough for AI workloads. In plain terms, if GPUs are the engines, HBM is the fuel line — and shortages can cap how many servers can be shipped.
A MarketBeat.com analysis republished by Investing.com argued Nvidia is trying to widen its role beyond selling GPUs, pointing to moves into software and broader infrastructure alongside its core chip business. The analysis highlighted the company’s Intel investment and a licensing-and-talent deal with AI chip startup Groq, as well as a bigger push into inference — the stage where trained models generate answers for users — and robotics tooling.
For investors, that strategy cuts both ways. Broader products can deepen customer ties and create recurring software revenue, but they also raise execution risk and put more scrutiny on whether AI spending produces measurable gains.
Nvidia shares were down about 0.5% in premarket trading on Friday, after a volatile finish to 2025 for many AI-linked stocks.
The next catalysts for the broader market arrive quickly, with a U.S. jobs report due on January 9 and inflation and fourth-quarter earnings season close behind — events that can reset expectations for interest rates and valuations.
In the near term, traders are watching whether Nvidia can translate headline demand into shipments without fresh supply constraints, and whether the policy backdrop for China sales stays stable. Developments in memory supply and foundry capacity are likely to be just as important as new product launches.
