MediaTek target price raised to NT$5,088, with rerating potential from 2nm TPU volume ramp

Overview

The report focuses on incremental opportunities for MediaTek in the Google TPU and AI ASIC supply chain. Morgan Stanley believes 2nm TPU "Humufish" has upside in 2028 volume, ASP, and supply-chain feasibility; Intel EMIB-T is likely to become the main packaging solution, while TSMC CoWoS is used to secure minimum production capacity. The report also points out that 1.4nm TPU v10 "Icefish" and potential turnkey full-rack services may support growth beyond 2029.

Core views

The core view is that AI ASIC demand is sufficient to offset smartphone business headwinds and drive a rerating of MediaTek. The report expects Google TPU to become the main growth driver for MediaTek in 2027-2028, with related revenue CAGR close to 40%-50% from 2025 to 2028. Although smartphone business may weigh on gross margin in 2026-2027, and TPU gross margin is also lower than the 3nm project, a larger 2nm TPU scale could still deliver a 20%-25% operating margin, above the 15%-20% level of the smartphone business.

Analysis framework

The report combines supply-chain research, management commentary, product roadmaps, and a valuation model: first it raises the 2nm TPU mass-production assumption based on supply-chain conditions such as ABF substrates, Intel EMIB yield, bumping, silicon capacitors, and silicon bridge dies; then it revises 2027-2028 earnings forecasts using ASP, gross margin, and operating margin assumptions; finally it derives the target price through a residual income model.

Methodology notes

• Valuation methodsResidual income model

  Derives base-case value from the cost of equity, medium-term growth rate, long-term growth rate, and EPS forecasts.

  The NT$5,088 target price comes from the base-case residual income model; key assumptions are a 9.2% cost of equity, 12.0% medium-term growth, and 3.0% long-term growth.

• Fundamental ForecastSupply Chain Research

  Uses packaging, substrates, yields, ASP, and customer product roadmaps to assess TPU shipments and revenue.

  The report raises the 2028 2nm TPU assumption to at least 2.5mn units and models it at a US$13k ASP.

• Research DataMorgan Stanley ModelWare

  The report's financial metrics and forecasts are based on Morgan Stanley's internal research model.

  Unless otherwise noted, metrics are based on the Morgan Stanley ModelWare framework, and e denotes Morgan Stanley Research estimates.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• MediaTek / MediaTek (2454.TW)

  The report covers the company and the main investment target

  Strengths

  Strong demand for Google TPU and AI ASICs, higher 2nm TPU shipment and ASP assumptions, and potential 1.4nm TPU projects and full-rack services create long-term upside.

  Weaknesses

  Headwinds in the smartphone business in 2026-2027 and margin dilution from TPU could pressure overall gross margin.

  Comparison

  The report believes the current P/E has risen but remains relatively attractive, because TPU upside is not yet fully reflected; the stock should trade more than one standard deviation above its historical average.

  Risks

  Deterioration in smartphone demand, intensifying competition, price competition, slower-than-expected ASIC and new business execution, and greater-than-expected margin dilution.

• Google TPU supply chain

  The core demand source for MediaTek's AI ASIC growth

  Strengths

  The 2nm TPU Humufish is expected to use a larger package and more main dies, with ASP potentially reaching US$12-15k; the supply chain includes ABF substrates, Intel EMIB, bumping, silicon capacitors, and silicon bridge dies.

  Weaknesses

  Large-scale ASIC projects are highly sensitive to packaging yields, substrate capacity, and customer demand timing.

  Comparison

  TSMC CoWoS is used to secure minimum production capacity, while Intel EMIB-T is seen as the main solution to lower costs and expand 2028 output.

  Risks

  Intel EMIB-T execution at scale falls short of expectations, CoWoS capacity is constrained, or customer product roadmaps and demand change.

Key data

• Report date2026-05-25Report published at May 25, 2026 07:53 PM GMT.
• RatingOverweightThe report maintains Overweight and says MediaTek is the Top Pick in Asian technology.
• Target priceNT$5,088Target price raised from NT$2,988 to NT$5,088.
• 2028 2nm TPU shipment assumptionat least 2.5mn unitsMainly based on the Intel EMIB solution.
• 2028 TPU revenue assumptionat least US$37bnIncludes 2.5mn 2nm TPUs at an ASP of US$13k, and 1mn 3nm TPUs at an ASP of US$4.5k.
• Google TPU revenue mix38% in 2027, 63% in 2028Morgan Stanley expects Google TPU to become a core driver of MediaTek revenue.
• Earnings forecast revision0% in 2026, +3% in 2027, +40% in 2028The upward revision mainly reflects improved 2nm TPU demand from late 2027 through 2028.
• Valuation multiple implication38x 2027e EPS, 18x 2028e EPSCorresponds to the new target price of NT$5,088.
• Bull / bear case valuesNT$6,988 / NT$2,588Raised from NT$4,100 and NT$1,520, respectively.
• 2nm TPU gross margin assumptionabout 35%Below the roughly 40% assumed for 3nm TPU, but operating margin is still expected to be around 20%-25%.

Impact & implications

If the 2nm TPU volume ramp materializes, MediaTek would shift from a traditional smartphone SoC company to a platform semiconductor design company with a much higher share of AI ASIC revenue. Its valuation anchor could rerate from a historical roughly 18x P/E to above 35x. For investors, the key variables would shift from near-term handset demand to Google TPU shipments, packaging supply, ASP, margins, and follow-on TPU generation projects.

Risks

• Deterioration in smartphone demand in China and other emerging markets.
• Intensifying competition and price competition, putting pressure on market share or ASP.
• Demand for new products falling short of expectations, causing market share loss.
• TPU or AI ASIC business margin dilution being greater than expected.
• Slower-than-expected development of non-phone businesses such as ASIC, IoT, automotive, and Edge AI.
• Intel EMIB-T, ABF substrates, or advanced packaging supply-chain execution falling short of expectations.
• The AI phone replacement cycle may be too weak to offset the added cost.

What to watch

• The actual production ramp and order visibility for 2nm TPU Humufish from late 2027 through 2028.
• Yield, cost, and capacity performance of Intel EMIB-T in large-scale ASIC projects.
• The roadmap for Google TPU 8, 2nm TPU, and subsequent 1.4nm TPU v10 projects.
• Whether MediaTek will offer turnkey full-rack services after 2028.
• Whether Google TPU revenue as a share of total MediaTek revenue approaches the 2027 38% and 2028 63% forecasts.
• Whether the Android AI agent smartphone replacement cycle starts with high-end Android models.
• Whether smartphone business gross margin pressure is already fully reflected in the share price.