Models

CoreValue Research · Model v2026.Q2

TSMC revenue scenarios to 2030

TSM · Data as of · Reviewed quarterly

Bear case
$278B
18.2% CAGR · $131B EBIT
Base case
$338B
23.0% CAGR · $159B EBIT
Bull case
$409B
27.8% CAGR · $192B EBIT

Projected revenue by year ($bn)

Scenario 2026 2027 2028 2029 2030 CAGR
Bear 142 167 196 233 278 18.2%
Base 148 180 220 272 338 23.0%
Bull 153 193 246 316 409 27.8%

Assumptions

Segment FY25 revenue Op. margin Bear Base Bull σ measured
Data Center (HPC) $70.5B 47% 24.7% 30.0% 35.3% 21.3pp
Smartphone $36.0B 47% 5.4% 8.0% 10.6% 10.3pp
Automotive & IoT $12.2B 47% 3.8% 9.0% 14.2% 20.7pp
Other $3.7B 47% 1.3% 5.0% 8.7% 14.9pp

Method

Each segment is compounded from its FY25 actual at a constant annual rate, and operating income applies a fixed margin per segment.

The base case is CoreValue Research's estimate. It is the analyst's own forecast for each line, not an output of the range — everything below only describes how far either side of it we draw the bear and the bull.

The width of the range is measured, not chosen. For each segment we take its realised year-on-year growth over FY2021–FY2025 and compute the sample standard deviation — how much that line's growth has actually varied, in percentage points. That figure is the σ column, and it is printed unmodified so you can always check our arithmetic or rebuild the range at any width you prefer.

Bear and bull are the base case minus and plus a quarter of that σ (FY2021–FY2025). So a segment whose growth has historically swung by 20 percentage points is shown 5 points either side of its base case. We publish a deliberately narrow band: a full standard deviation priced several of these segments at a multi-year decline, which measures how violently a young industry has moved rather than describing a path we would defend. Growth decelerating hard from an exceptional base is the risk we are willing to put our name to; the industry ceasing to exist is not.

These are scenarios, not confidence intervals. The band is not a probability statement, nothing here is calibrated to say the outcome falls inside it 70% or 90% of the time, and a quarter of a standard deviation is a house convention rather than a statistical threshold. It is a disciplined way of saying “this much variation is normal for this business”, and it should be read alongside the assumptions rather than instead of them.

TSMC discloses revenue by platform but reports a single company gross profit, not profitability by platform. All four lines above therefore share one margin, which is the finest split the disclosure supports: FY2030 segment gross profit of US$191.6bn less company operating expense of US$30.2bn retains 84.2%, reproducing the forecast's own FY2030 operating income of US$161.4bn exactly. High Performance Computing almost certainly earns more than the Other line, and the data to separate them is not published. The forecast rests overwhelmingly on that one platform: Data Center revenue compounds at 30% a year from US$70.5bn to US$266.0bn, which would be 78% of FY2030 revenue against 58% today. Smartphone, the second line, is assumed to grow at 8%, roughly in line with the last five years.

Citation — CoreValue Research, “TSMC revenue scenarios to 2030”, model v2026.Q2, August 10, 2026. Base case: $338B revenue and $159B operating income in 2030, assuming Data Center (HPC) grows 30% a year, Smartphone grows 8% a year, Automotive & IoT grows 9% a year, Other grows 5% a year.