TSMC’s 30% Revenue Surge Reveals AI Capex Decoupling From Equity Panic

TSMC reported combined January-February revenue of NT$718.91 billion, a 29.9% year-over-year increase that provides hard evidence of sustained AI chip demand while broader equity markets sold off technology names.

The world’s largest contract chipmaker posted approximately 30% growth for the first two months of 2026, with January revenue jumping 37% to NT$401.3 billion ($12.7 billion). The numbers arrive as semiconductor investors parse a disconnect between infrastructure spending momentum and equity market sentiment: software stocks have declined 25-30% year-to-date on AI disruption concerns, yet the companies funding AI buildouts continue aggressive chip procurement.

The revenue data validates a thesis gaining traction among allocation strategists: current equity corrections reflect geopolitical and valuation pressures rather than weakening AI fundamentals. TSMC expects full-year 2026 revenue growth close to 30% in U.S. dollar terms, supported by capital expenditure between $52 billion and $56 billion, with 70-80% allocated to advanced process technologies serving AI workloads.

Hyperscaler Capex Holds Despite Market Volatility

The semiconductor demand signal aligns with capital expenditure commitments from TSMC’s largest customers. Amazon, Google, Meta and Microsoft plan to invest up to $630 billion in capital expenditures for 2026, about a 62% increase from the record $388 billion in 2025 spending, according to Data Center Richness. Roughly 75% of this spending targets AI infrastructure.

This spending intensity represents a structural shift in technology capital allocation. Hyperscalers now spend 45-57% of revenue on Capex, ratios previously unthinkable for technology companies, resembling industrial or utility firms rather than traditional software businesses. The capital deployment occurs even as Microsoft holds an $80 billion Azure backlog stemming from power constraints, not demand softness, per Introl.

Equity Markets Price Macro Risk, Not Demand Risk

The divergence between semiconductor procurement and software equity performance illustrates a sectoral decoupling. Palantir is down about 22% in 2026, while Adobe, Salesforce, and ServiceNow have seen shares slide 25% to 30%, even as these companies report strong results with AI acting as a catalyst rather than headwind, notes The Motley Fool.

Market participants point to valuation resets rather than fundamental deterioration. Unlike the dot-com bubble of the 1990s, today’s growth is supported by real earnings, according to analysis from Generali Asset Management cited by Funds Society. Fidelity International calls AI “the defining theme for Equity Markets” in 2026, with the BlackRock Investment Institute saying the technology will likely “keep trumping tariffs and traditional macro drivers”.

Memory and Advanced Node Tightness Persists

Chip procurement patterns reveal AI-specific supply constraints that reinforce hyperscaler urgency. Memory revenues in 2026 are likely to reach $200 billion, with growth in demand for HBM3, HBM4, and DDR7 memory causing shortages of consumer memory; prices for DDR4 and DDR5 were up about 4x between September and November 2025, per Deloitte Insights. Samsung, SK Hynix, and Micron could leave DDR5 buyers in a tough spot, with prices surging 30% to 50% every quarter through the end of the first half of 2026, according to industry analysts cited by Sourceability.

Advanced process capacity remains allocated to strategic customers. As more customers commit to long-term capacity agreements with TSMC, allocation available to second-tier clients will narrow; Nvidia and Apple have already secured large blocks of production capacity through the end of 2026. The capacity constraints explain why hyperscalers continue raising capex even as free cash flow comes under pressure.

Semiconductor Market Projections Diverge Sharply

Industry forecasts for 2026 semiconductor revenue span a wide range, reflecting uncertainty about AI’s multiplier effect on total addressable market. The Semiconductor Industry Association reports global sales of $791.7 billion in 2025, a 25.6% year-over-year increase, with 2026 annual sales projected to reach roughly $1 trillion worldwide, according to StartUs Insights. UBS Securities expects the semiconductor market to approach $700 billion in 2025, surpass $1 trillion in 2026 representing year-on-year growth of more than 40%, and reach around $1.18 trillion by 2027.

The variance in projections stems from differing assumptions about AI workload proliferation beyond hyperscaler data centers. Deloitte projects global chip sales will reach $975 billion in 2026, with growth accelerating from 22% to 26% year-over-year. Conservative estimates from the World Semiconductor Trade Statistics place an 8.5% global market increase in 2026, reaching $760.7 billion, with growth expected across all major regions and product categories.

What to Watch

TSMC’s Q1 2026 earnings in April will provide granular visibility into 3nm and 2nm node utilization rates and customer mix shifts. Monthly revenue disclosures through March will clarify whether January’s 37% growth represented Lunar New Year timing effects or genuine acceleration. Hyperscaler quarterly reports in late April and early May will reveal whether capex guidance holds as free cash flow pressure mounts: Amazon faces projected negative free cash flow of $17-28 billion, and Alphabet’s free cash flow will plummet roughly 90% to $8.2 billion.

Memory pricing dynamics bear close monitoring as supply-demand imbalances ripple through non-AI segments. If DDR4/DDR5 prices continue 30-50% quarterly increases, procurement teams at consumer electronics and automotive OEMs will face margin compression that could trigger inventory adjustments. Advanced packaging capacity additions at TSMC and competitors will signal confidence in sustained CoWoS and chiplet demand beyond 2026.

The strategic question for allocators: whether current software equity weakness creates entry points into AI beneficiaries trading at compressed multiples, or whether hyperscaler capital intensity represents an unsustainable overbuilding cycle. TSMC’s revenue trajectory suggests the former, but the power grid constraints limiting Azure deployment hint at execution risk that could delay ROI realization into 2027-2028.