Record revenue signals from the foundry core
TSMC, the world’s largest contract chipmaker, forecast second-quarter revenue between $39 billion and $40.2 billion — potentially its highest quarterly revenue ever.
The company also raised its full-year revenue growth outlook to more than 30% year over year, while signaling that capital expenditure would reach the upper end of its previously disclosed $52 billion to $56 billion range.
CEO C.C. Wei told investors that “AI demand is very strong,” pointing to continued positive signals from cloud service providers and their downstream enterprise customers.
TSMC’s Q1 gross margin reached 66%, its highest level in roughly two decades. That margin strength suggests pricing power remains intact despite the capital intensity required to expand advanced-node production.
For AI chip designers such as Nvidia, AMD and Broadcom, TSMC’s guidance effectively confirms that order books remain full.
Equipment demand reinforces the cycle
On the upstream side of the ecosystem, ASML — the sole supplier of extreme ultraviolet (EUV) lithography systems required for cutting-edge chips — also exceeded quarterly expectations and raised its full-year sales outlook.
ASML CEO Christophe Fouquet said demand is expected to exceed supply “for the foreseeable future,” adding that constraints will likely stretch across markets ranging from AI to smartphones and PCs.
For semiconductor investors, strong results from both the manufacturing and equipment layers reduce the probability that AI chip demand is peaking in the near term.
Giuseppe Sette, co-founder of Reflexivity, noted that ASML’s results present a favorable signal for the broader industry, even amid concerns about overinvestment in AI.
Big Tech spending remains the anchor
The semiconductor surge ultimately reflects aggressive infrastructure expansion by hyperscalers.
Companies including Microsoft, Meta, Amazon and Alphabet are collectively expected to invest more than $600 billion in data center capacity this year.
Despite investor pressure to demonstrate returns on AI spending, capital expenditure plans have not materially slowed.
The nature of chip demand is also evolving. While training large models remains compute-intensive, increasing demand is shifting toward inference chips — hardware optimized to run large language models in production environments.
That transition broadens demand beyond frontier AI labs into enterprise deployment at scale.
Capacity constraints and structural bottlenecks
Explosive demand growth is colliding with structural realities.
The advanced semiconductor supply chain is concentrated among a handful of critical players — TSMC for leading-edge fabrication and ASML for EUV tools.
As a result, production capacity is described as “very tight.”
Chip designers are reportedly pursuing multi-year capacity agreements to secure guaranteed wafer allocations. Equipment lead times remain extended, particularly for advanced lithography systems.
TSMC has responded by accelerating capital expenditure and securing additional land in Arizona to expand U.S.-based production — a move that also aligns with long-term geopolitical diversification strategies.
Meanwhile, ASML continues to ramp output, though scaling EUV system production is inherently complex.
Geopolitical risk, but manageable
TSMC downplayed the risk of immediate disruption from geopolitical instability in the Middle East, noting that critical specialty gases such as helium and hydrogen are sourced from diversified suppliers.
Still, geopolitical diversification remains part of the long-term strategy, particularly as U.S.-China technology tensions persist.
For policymakers, the concentration of advanced manufacturing capacity underscores why semiconductor sovereignty remains a strategic priority.
What this means for startups and investors
For AI startups, the implications are twofold:
Compute availability remains constrained and expensive.
Infrastructure expansion suggests long-term support for AI deployment at scale.
For investors, synchronized strength across foundry and equipment layers signals that the semiconductor upcycle is not solely Nvidia-driven — it is ecosystem-wide.
However, sustained CAPEX at these levels increases sensitivity to any slowdown in hyperscaler spending. Upcoming earnings reports from major cloud providers will therefore be closely watched for confirmation that infrastructure investment remains on track.
The bigger signal
TSMC and ASML sit at the foundation of the global semiconductor stack.
When both raise outlooks simultaneously, it carries weight beyond quarterly earnings beats. It suggests that the AI investment wave is not merely speculative enthusiasm — it is embedded in capital allocation decisions across the technology sector.
The next test will come from Big Tech earnings and forward guidance. But for now, the semiconductor backbone of the AI economy appears not only intact, but accelerating.
