Bernstein Raises TSMC Price Target on AI and Advanced Packaging Strength

TSMC US8740391003

Analysts at Bernstein have issued a significantly more bullish outlook for Taiwan Semiconductor Manufacturing Company (TSMC), sharply increasing their price target for the world’s leading contract chipmaker. The firm cites the relentless demand for artificial intelligence and a specific advanced packaging technology as the core drivers for this upgraded assessment.

In its latest analysis, Bernstein lifted its price target for TSMC’s Taiwan-listed shares from TWD 1,444 to TWD 1,800. For the company’s US-listed American Depositary Receipts (ADRs), this equates to a new target of $330. With the ADRs currently trading near $257.50, the revised target implies considerable upside potential.

Market strategist Mark Li attributed this optimistic move to a powerful combination of factors: soaring demand for specialized packaging solutions and an anticipated revenue surge of 23% for the year 2026. Bernstein maintains its "Outperform" rating on the semiconductor giant.

CoWoS: The Critical Enabling Technology

A specific packaging technique known as CoWoS (Chip-on-Wafer-on-Substrate) is central to Bernstein’s thesis. This method allows for the ultra-dense stacking of processors and memory, dramatically accelerating data transfer rates—a feature that is absolutely essential for cutting-edge AI chips.

The research house projects that TSMC will expand its monthly CoWoS production capacity to 125,000 wafers by the end of 2026. Bernstein experts estimate that announced projects from Nvidia alone, specifically its Blackwell and Rubin platforms slated for 2025 and 2026, will consume nearly all of this available capacity.

Key financial projections from the report include:

• Revenue Growth: 23% in 2026, followed by 20% in 2027
• Earnings Per Share (EPS): An annual growth rate of 20% through 2027
• Capital Expenditure (Capex): Approximately $47 billion for the coming year
• Margins: An expected moderate decline of only about 150 basis points over a two-year period

Positioned for a Multi-Year Infrastructure Cycle

The analysts’ confidence extends well beyond the immediate horizon. Bernstein calculates that the global build-out of AI data centers will require an expansion of 55 to 60 gigawatts of capacity by 2029. This sustained infrastructure investment is forecast to enable TSMC to grow its AI-related revenue at a mid-40% annual rate.

In the study, TSMC is characterized as a "Quality Compounder" and a foundational investment, given its structural position at the very heart of the ongoing AI infrastructure expansion. The company’s ongoing capacity build-out is seen as a direct response to this unbroken appetite for artificial intelligence capabilities.