After Jensen Huang's exit, TSMC reportedly asks AI chip buyers to pay cash upfront
Days after Jensen Huang stepped down as Nvidia CEO, TSMC is reportedly tightening credit limits and asking AI chipmakers to pay cash upfront to protect itself from growing financial risks in the AI boom.

Just days after Jensen Huang stepped down as Nvidia CEO on November 12, the global chip industry is already feeling the heat. According to Taiwan’s Commercial Times, the world’s largest contract chipmaker, TSMC, is changing how it takes AI chip orders. Instead of relying on trust and credit, TSMC is now checking how financially strong its buyers are before giving them factory space.
Citing industry sources, the report says TSMC sets a credit limit for each client based on its balance sheet, past loss rates, competitiveness, and business outlook. If an order crosses that limit, the buyer must pay the extra amount in cash.
Some companies are even paying upfront just to book wafer capacity. TSMC also does not allow clients to rent out or auction reserved capacity, except in rare photomask licensing cases.
Why TSMC is worried about AI money
The strict new rules are a response to "circular financing" in the AI boom. In simple words, AI firms take loans, invest in each other, buy expensive chips, and promise to pay back using future profits.
Under Huang, Nvidia acted like a safety net for this system, putting over $100 billion into stakes and financial backstops for partners like OpenAI, Anthropic, and CoreWeave, alongside $118 billion in guarantee commitments. With Huang gone, interim CEO Colette Kress has promised a more "disciplined" approach to investments.
TSMC Chairman C.C. Wei has previously said the company checks cloud firms' spending, financial health, power supply, and component availability before planning capacity, even sending staff to inspect US data centre construction in person.
Custom chipmakers feel the pinch
The biggest impact is on custom AI chip designers, also known as ASIC firms. These companies must pay TSMC as soon as wafers are made, wait for advanced packaging, and only get paid when finished chips reach buyers. As orders grow, this cash gap gets harder to manage, meaning big AI projects could move ahead while smaller ones wait longer in line.
To cope, Taiwanese ASIC firms have been raising massive funds. MediaTek priced a $3.9 billion zero-interest overseas convertible bond on August 31, with Nvidia buying $3.5 billion of it.
Is Nvidia in trouble?
Not right now. On November 17, Nvidia posted record Q3 revenue of $115.8 billion, up 103 per cent from last year, and projected $130 billion for the next quarter. It also reported $390 billion in purchase commitments, primarily for multiyear memory procurement.
Kress said Nvidia’s customers, suppliers, and partners were “100% behind us,” and confirmed that TSMC executives had visited Nvidia’s offices that week and were fully engaged in helping address supply-and-demand gaps.
Still, investors are nervous, with TSMC’s stock closing down 3.45 per cent on November 19. For nearly 30 years, Huang and TSMC founder Morris Chang shared a rare partnership built on personal trust, Chang even asked Huang to lead TSMC in 2013. Without Huang at the helm, TSMC's message to the AI world is simple: show us the cash.

