AI Chips Outlook: TSMC and ASML Signal Multi-Year Demand Cycle

Investors seeking reassurance about the pace and longevity of artificial intelligence (AI) infrastructure spending have found solace in the latest earnings reports of the world’s largest manufacturer of advanced chips, Taiwan Semiconductor Manufacturing Company [TSM], and its equipment supplier, ASML [ASML].

Both companies reported another quarter of record revenue in Q3 and forecast even higher revenue in Q4. However, investor attention was largely focused on what their management teams had to say about the outlook for AI spending through the end of the decade.

TSMC’s capacity expansion continues

There are few companies with more skin in the AI game than TSMC. The company, which manufactures advanced chips designed by Nvidia [NVDA], Advanced Micro Devices [AMD] and Broadcom [AVGO], spent $40.89bn in 2025 to expand its production capacity and expects to spend another $60-64bn in 2026.

The latest commentary from the industry bellwether suggests that AI-related demand continues to outstrip supply. 

Amid concerns over power availability in the US, TSMC has been sending teams to assess the progress of data centre construction projects. Management said on its Q3 earnings call that the findings strengthened its confidence in AI demand, supporting plans for “significantly higher” capital expenditure between 2026 and 2028 compared with 2023-25.

“TSMC is expanding at an unprecedented pace with nearly 20 fabs worldwide in progress. We are building 13 leading edge and advanced packaging fabs in Taiwan over the next several years, and another five to six fabs overseas,” saidC.C. Wei, chairman and CEO of the company, during the Q3 earnings call.

Wei pointed to emerging trends such as agentic AI and open-source models as drivers of “continued robust multiyear demand.”

However, the company’s aggressive overseas expansion and the ramp-up of its latest 2-nanometre (N2) manufacturing technology have weighed on gross margins, with the pressure expected to persist through the end of 2026.

Revenue, meanwhile, increased by 50.9% year-on-year and 17.6% quarter-on-quarter to NT$1.49tn in Q3, while full-year revenue is expected to grow about 44% y/y in 2026.

ASML juggles demand, capacity constraints

ASML echoed a similarly bullish demand outlook, revealing that customers were looking to place orders for its semiconductor manufacturing equipment for 2029 and 2030.

However, concerns are emerging that ASML’s production capacity could become a constraint on the AI infrastructure buildout after the company said its current supply plans for extreme ultraviolet (EUV) lithography systems would not fully meet demand in 2027 and 2028.

Given ASML’s monopoly in advanced EUV lithography systems, the execution of its capacity expansion plans is becoming increasingly important for the broader AI industry.

During its Q3 earnings call, management said the company was exploring ways to expand its 2028 production capacity beyond its previously disclosed target of a 30% increase.

Analysts scrutinised these plans, questioning whether constraints within ASML’s supply chain could limit production growth, particularly whether precision optics supplier Zeiss could ramp up capacity in parallel.

CEO Christophe Fouquet dismissed concerns about Zeiss becoming a bottleneck, pointing to a 50% increase in EUV production capacity achieved in 2026 as evidence of ASML’s ability to execute further expansion in 2027 and 2028.

“Every number we share with you is the result of us having validated that the supply chain can, of course, follow those numbers,” added Fouquet.

Strong demand and ASML’s dominant position in advanced lithography equipment have also given the company greater pricing power, allowing it to raise prices for its “most critical products”. Management said the impact of these increases would be partially reflected in 2027 and fully realised in 2028.

Additionally, technology upgrade cycles at semiconductor fabrication plants were identified as a potential “swing factor” supporting ASML’s margin expansion in 2027 and beyond.

In Q3, ASML reported record quarterly net sales of €11.7bn, up 56% y/y and 25% sequentially.

For the full year 2026, ASML forecast net sales to be between €44bn and €45bn, compared with €32.7bn in 2025. Gross margin was expected to reach around 56%, up from 52.8% a year earlier.

Hyperscaler earnings, Gartner warning

Investor attention will now turn to the earnings reports of the world’s largest hyperscalers, with a focus on capital expenditure plans, AI adoption rates, computing demand and data centre construction timelines.

Meta [META] is scheduled to report its quarterly results on 21 October, followed by Microsoft [MSFT] and Alphabet [GOOGL] on 28 October, and Amazon [AMZN] on 29 October.

Continued revenue growth at Google Cloud, Microsoft Azure and Amazon Web Services will provide investors with evidence that the hundreds of billions of dollars being invested in AI infrastructure are beginning to pay off, while also highlighting growing demand for computing capacity from AI labs.

Positive results could help ease mounting concerns over the increasingly expensive AI infrastructure buildout, particularly against a challenging macroeconomic backdrop characterised by sticky inflation and rising interest rates.

Research firm Gartner has warned that AI data centre owners could fall significantly short of the estimated $8.5tn in revenue needed to cover announced capital spending, financing costs and a 12% return on invested capital (ROIC) through 2030. 

The firm forecasts cumulative enterprise and consumer AI revenue attributable to AI data centres of no more than $3.6tn over the period.

“Failure to achieve sustainable ROIC on rapidly depreciating assets would almost certainly necessitate write-downs,” Gartner said in a 14 October note, adding that neoclouds remained “most exposed to the imbalance between revenue and capital spending.”

Conclusion

Taken together, the results from TSMC and ASML suggest the AI build-out still has considerable room to run, with both companies now planning capacity and pricing for demand that stretches into 2029 and 2030. The constraint has shifted from whether customers will place orders to whether the supply chain, ASML’s EUV systems above all, can keep pace. The harder question is the return on the capital behind it, as Gartner’s warning makes clear. The coming hyperscaler earnings will show whether cloud revenue growth is large enough to justify the spending. For now, demand is strong and supply is tight, but the pay-off is still to be proven.