Wistron Won a Massive Nvidia Order

The AI hardware boom has a working-capital problem. Wistron made it visible on Tuesday.

Shares of Taiwan’s Wistron Corporation fell more than 6% on Tuesday, September 8, 2026 after the Nvidia supplier priced a $1.47 billion global depositary receipt offering to fund raw material purchases. The business case for the raise is not in dispute: in August, Wistron said AI server demand from cloud computing firms and business customers continued to exceed supply. Its second-quarter net profit rose 128% from a year earlier, while revenue climbed 64%. Demand is real. The problem is paying for it before customers settle their bills.

Wistron is inside the supply chain building AI server systems for Nvidia, where demand has been growing faster than much of the industry can safely provide. More orders means more components, more factories, and more working capital, before customers eventually pay up. In July, Wistron opened its first U.S. manufacturing facility, a $700 million AI smart facility in Fort Worth, Texas. The site produces NVIDIA GB300 Grace Blackwell Ultra Superchips and is expected to add production of NVIDIA Vera Rubin Superchips. Capital is going out the door at pace.

At roughly NT$186.24 per underlying share, the issuance came in 5.5% below the NT$197 level at which Wistron shares ended Monday’s session. Reuters reported the final price landed near the bottom of the marketed range of $58.67 to $59.79 per receipt. Pricing at the floor of the range is a signal that investor appetite was not exactly queuing around the block. The new shares represent about 7.29% of Wistron’s outstanding shares before the issuance.

That is the market’s message in plain terms: when a low-margin assembler issues nearly 7.3% dilution to buy GPU components, investors price the dilution first and the order book second. This is not a verdict on Nvidia’s demand. TrendForce said in August that hyperscale cloud providers were ramping AI spending and lifting server shipments, while Wistron and other manufacturers expanded GPU-rack fulfillment. The research firm forecasts global AI server shipments to rise nearly 31% in 2026. The cycle is intact.

What is shifting is where the skepticism lives. Investors who were happy buying Wistron on revenue growth are now asking a harder question: how much of that growth actually flows through to shareholders after you account for dilutive capital raises, expensive components, and hyperscalers who set the pricing terms? Wistron is committing large sums to factories, inventories, and expensive components. It also operates in a business where powerful hyperscale customers can influence pricing and margins.

Watch how peers respond. Foxconn, Quanta, Celestica, and Jabil all carry some version of this tension: accelerating AI server revenue on balance sheets that weren’t built for the working-capital intensity of GPU-rack assembly at scale. Wistron is the first name this cycle to put that tension on the front page. It will not be the last.

The Trader’s Lesson

Booming revenue and falling stock prices can coexist when the mechanism of growth consumes capital faster than the market values the returns. Before adding to any AI supply-chain name, traders should ask one question: who bears the working-capital risk, and at what margin? That question has a different answer today than it did six months ago.