What TSMC's Q3 2026 Earnings Mean for NVIDIA (NVDA) and AMD as 2nm Ramp and HPC Surge

TSMC's Q3 2026 revenue, profit, and Q4 guidance all exceeded expectations, but for GPU stocks, what is truly worth watching is the business mix: 2nm process contributed 8% of revenue, confirming AMD's MI450 ramp; high-performance computing (HPC) grew 22%, confirming NVIDIA's demand; and CoWoS capacity still has not been given a specific figure.

Key Takeaways

• TSMC's Q3 2026 results (October 15, 2026) beat expectations on revenue, profit, and guidance. This article analyzes what the earnings mean for NVIDIA and AMD; TSMC's full results are covered in a separate article.
• The new information lies in the business mix. The 2nm process rose from 3% of wafer revenue in Q2 to 8% in Q3, HPC reached 68% of revenue, up 22% quarter-over-quarter. Gross margin was 66.4%, down 1.3 percentage points from Q2, which TSMC attributed to the initial ramp of the 2nm process.
• On the earnings call, TSMC guided Q4 revenue of $52.3 billion to $54.3 billion (up 58% year-over-year at the midpoint), raised its 2026 growth outlook to about 44%, said AI demand is "higher than supply," and noted that 2nm capacity will more than double from current levels by 2027. The company did not provide a specific figure for CoWoS capacity.
• For NVIDIA (NVDA): demand is confirmed, but packaging supply remains unquantified. For AMD: the 2nm ramp and the plan to double capacity in 2027 represent the first third-party confirmation of the process node for its MI450. Still, both stocks fell on October 15, driven by a sector-wide selloff over concerns about the return on AI capex, rather than by TSMC's reported figures.

Why Are TSMC's Earnings a Leading Indicator for NVIDIA and AMD?

TSMC manufactures all the data center GPUs that NVIDIA and AMD sell, and also handles advanced packaging (CoWoS) that connects these GPUs to high-bandwidth memory. Neither company has a second supplier at the most advanced process nodes. As a result, TSMC's output, margins, and process node mix are the earliest hard data for judging how many AI chips will actually be produced over the next two to four quarters.

What Did TSMC Report for Q3 2026?

According to the Q3 2026 management report: revenue was $46.72 billion, above guidance of $44.6 billion to $45.8 billion; net income was NT$763.06 billion, up 68.7% year-over-year, above consensus; gross margin was 66.4%, operating margin was 58.7%; quarterly capex was $13.84 billion.

For full results, segment data tables, and margin change analysis, see "TSMC Q3 2026 Earnings: Profit Jumps 69% as 2nm Production Scales, Revenue Beats Guidance."

The rest of this article discusses what these figures mean for TSMC's two customers.

What Do the 2nm and HPC Figures Mean for NVIDIA and AMD?

Two business mix data points in the report matter more to GPU investors than the earnings beat itself: the 2nm process rose from 3% of wafer revenue in Q2 to 8%, and the HPC platform reached 68% of revenue, up 22% quarter-over-quarter.

The 2nm data is mainly relevant to AMD. AMD is the major AI customer adopting this process node this year, with EPYC Venice already in mass production and the MI450 series also starting production in Q3. As such, the jump in 2nm revenue share from 3% to 8% in a single quarter is the first third-party evidence that the process node AMD is using is ramping at scale, not just at pilot scale. NVIDIA's current Rubin generation uses a 3nm-class process, so by any measure, the 2nm data says little about NVIDIA's 2026 supply situation.

The HPC data is primarily relevant to NVIDIA, as NVIDIA is the largest single contributor to the platform. HPC grew 22% quarter-over-quarter while total revenue grew 16% in dollar terms, meaning the AI-heavy segment continues to gain share within TSMC. This provides the demand confirmation NVIDIA needs ahead of its own earnings in November. One caveat: HPC also includes PC and traditional server chips, so "HPC up 22%" is an indirect indicator of AI demand, not a direct measure of it.

The only negative in the report also varies by customer: the 1.3 percentage point gross margin decline came from the initial ramp of the 2nm process, so it affects the process node AMD is using, not NVIDIA's.

What Guidance Did TSMC Give for Q4 2026? What Did the Call Add?

According to the earnings call transcript, on the October 15 call, CFO Wendell Huang guided Q4 2026 revenue of $52.3 billion to $54.3 billion, up 14% quarter-over-quarter and 58% year-over-year at the midpoint; gross margin of 65% to 67%, and operating margin of 56.5% to 58.5%.

CEO C.C. Wei raised the full-year 2026 dollar revenue growth outlook from "slightly above 40%" given in July to about 44%. The midpoint of gross margin guidance is 66%, 40 basis points below Q3, which management again attributed to the 2nm ramp.

For GPU investors, three statements on the call matter more than these guidance ranges.

Wei said AI demand is "higher than supply," and TSMC is working to close that gap.

He said 2nm family capacity is expected to more than double from current levels by 2027.

When asked about the target of a 50% mid-to-high range compound annual growth rate for AI accelerator revenue through 2029, he said the actual growth rate "could be higher."

The call did not provide a specific figure for CoWoS capacity. Wei only said CoWoS "remains the preferred choice for our customers," and that TSMC has begun producing 5x reticle size packaging while advancing 9x and 14x reticle size packaging technologies.

The 2026 capex budget of $60 billion to $64 billion was reaffirmed rather than raised. As of the end of September, TSMC had spent $40.64 billion, implying Q4 capex alone of about $19 billion to $23 billion.

Huang added that capex from 2026 to 2028 will be "significantly higher than the past three years."

What Do TSMC's Earnings Mean for NVIDIA (NVDA)?

For NVIDIA, the Q3 data confirms demand, but the supply question remains unanswered. HPC grew 22%, confirming that AI accelerator output continued to rise through year-end, while the 66.4% gross margin shows TSMC still has pricing power at the 3nm-class node NVIDIA's Rubin uses.

The unresolved issue is packaging, and the call provided no answer.

NVIDIA's unit shipments from 2026 to 2027 depend more on CoWoS supply than on wafer supply. Management did not provide CoWoS capacity data, saying only that demand exceeds supply and that larger reticle size packaging products have entered production.

"Demand higher than supply" is positive for NVIDIA's pricing on one hand, but also caps its unit shipments on the other. Which effect dominates depends on data TSMC did not disclose.

Overall assessment: supported on demand, unresolved on supply.

The Q4 guidance showing 58% year-over-year growth, combined with management's statement that AI accelerator revenue CAGR could exceed the high end of 50%, is TSMC's strongest demand commentary this year. And the gross margin dilution from the 2nm process in this cycle is not borne by NVIDIA.

Still, NVDA closed down 1.84% on October 15, while TSM fell 1.80% over the same period. The sector-wide selloff was tied to doubts about the return on AI capex, not triggered by any figure in TSMC's report.

NVIDIA's own financial results will be released in November through NVIDIA's investor relations channels.

What Do TSMC's Earnings Mean for AMD?

AMD is the company most affected by the 2nm data in this report.

According to AMD investor relations materials, AMD said EPYC Venice has entered mass production on the 2nm process, and the MI450 series began production at TSMC in Q3 2026, with plans to begin volume shipments in 2027.

The jump in 2nm revenue share to 8% is the first third-party confirmation that the process node AMD has bet on for its next-generation accelerators is actually shipping at scale.

The gross margin dilution shows TSMC is prioritizing the ramp over quarterly gross margin. This is exactly what a customer waiting for 2nm capacity wants to see. Whether any of these costs will be reflected in AMD's wafer purchase prices is a contractual matter neither company has disclosed.

The call also provided forward-looking information: 2nm capacity is expected to more than double from current levels by 2027, the year AMD says MI450 will begin volume shipments.

Without naming the customer directly, this is about as clear a confirmation of a customer's product roadmap as TSMC can give.

Overall assessment: positive on execution progress, but packaging capacity allocation remains the same unresolved issue.

Because AMD's share of CoWoS capacity is smaller relative to NVIDIA's, the call's lack of CoWoS capacity data leaves AMD's 2027 shipment plans facing greater uncertainty than NVIDIA's on a percentage basis.

AMD shares had already fallen 5.7% on October 14 and remained under pressure on October 15. This was also part of the same AI capex-related selloff affecting NVDA and TSM, rather than a market reaction to the 2nm data.

MEXC analyst [Analyst Name], [Title], said:

"The market had already priced in the earnings beat a week earlier, but the rise in 2nm revenue share to 8% was not baked into expectations. This is first and foremost an AMD-related data point, and secondarily a TSMC-related one."

How Can Traders Trade on the Signals from TSMC's Earnings on MEXC?

MEXC's RealStocks service, through partner brokers, gives users exposure to US-listed stocks such as NVDA and AMD, so traders can participate in trades related to the impact of TSMC's earnings on GPU stocks on the same platform they use for crypto trading.

For NVIDIA, MEXC also offers tokenized exposure; for the differences between NVDAON, NVDAX, and NVIDIA RealStocks, see the related comparison article.

Open a RealStocks account to trade NVIDIA and AMD shares around earnings releases.

TSMC itself trades on MEXC as TSMUSDT stock futures contracts.

One point worth noting about the October 15 price action:

TSMC's earnings beat, guidance, and demand-related commentary were all positive, yet NVDA, AMD, and TSM shares still fell.

When a supplier reports strong earnings while customer stocks fall, the driver is usually the macro environment or industry narrative, not the fundamentals in the report.

Here, the core question is whether hyperscalers can earn a return on AI spending.

For anyone sizing positions based on the signals from TSMC's earnings, distinguishing between these two factors is important.

Frequently Asked Questions (FAQ)

Are TSMC's Q3 2026 earnings beat good for NVIDIA stock?

Supported on demand, but unresolved on supply.

HPC revenue up 22% quarter-over-quarter, Q4 guidance showing 58% year-over-year growth, and management's statement that AI demand exceeds supply all confirm that the process node NVIDIA uses is running at full capacity.

The variable that determines NVIDIA's unit shipments from 2026 to 2027 — CoWoS packaging capacity — was not quantified on the call.

Why is TSMC's 2nm ramp more important for AMD than for NVIDIA?

AMD's MI450 accelerators and EPYC Venice processors use TSMC's 2nm process node; NVIDIA's current Rubin generation uses a 3nm-class process.

As a result, the rise in 2nm's share of TSMC's wafer revenue from 3% to 8% directly confirms AMD's supply ramp, and the gross margin dilution from that ramp occurs at the node AMD uses, not NVIDIA's.

Does strong TSMC HPC revenue mean AI revenue is equally strong?

The two are not one-to-one.

HPC includes data center GPUs and AI application-specific integrated circuits (ASICs), as well as PC and traditional server chips.

HPC at 68% of revenue, up 22% quarter-over-quarter, is the best indirect indicator of AI demand currently available, but it is not a direct measure of AI demand.

Why did NVIDIA and AMD shares fall after TSMC reported strong Q3 2026 results?

The October 15 declines (NVDA down 1.84%, TSM down 1.80%, AMD under continued pressure after falling 5.7% the previous day) stemmed from a sector-wide selloff over concerns about whether AI capex can generate sufficient returns, not from TSMC's financial data.

TSMC's revenue, profit, and guidance all exceeded expectations.

What guidance did TSMC give for Q4 2026?

Revenue of $52.3 billion to $54.3 billion, gross margin of 65% to 67%, and operating margin of 56.5% to 58.5%.

Full-year 2026 dollar revenue growth is now expected at about 44%, and the 2026 capex budget is maintained at $60 billion to $64 billion.

When will NVIDIA and AMD report their own earnings?

AMD will report in early November 2026, and NVIDIA's fiscal quarter will end in late October, with results released in November.

TSMC's October 15 report is the earliest supply-side read on the two companies' relevant quarterly results.