Outlook Supplement - Flipbook - Page 17
Forecast
etch processing, positioning us well to
support the next wave of development
activity around the world.
We are seeing particularly robust momentum in Europe, where investments
in advanced packaging are turning
on to support regional manufacturing
production ramps as well as shorten
technology-development cycles. Asia
continues to be another key area of
growth as customers scale both legacy
and emerging packaging platforms.
While the U.S. market remains steady,
these expanding international investments are driving an increasingly global
customer base for ClassOne.
At the same time, AI data-center
buildouts continue to reshape the
requirements for both photonics and
advanced logic packaging. Our customers are actively ramping production
of the next generation of transceivers,
switches, and AI-centric devices—applications that depend on exceptionally
uniform plating performance and
seamless integration of downstream
processes. Across these markets, we are
seeing strong interest not only in our
established Solstice platform but also in
our new Solstice Max, which expands
our reach into 300mm manufacturing.
The system has been very well received,
with multiple orders across regions, and
we are focused on executing installations through the end of this year and
into 2026.
We are also encouraged by customers who initially adopted our
plating systems and are now expanding
into surface-prep steps such as metal
lift-off (MLO) and photoresist strip,
as well as such additional wet-processing capabilities as under-bump
metallization (UBM) etch and bevel
etch. These engagements validate our
strategy of delivering tightly integrated,
single-wafer solutions that streamline
process flow and support increasingly
complex packaging architectures.
Looking ahead, we expect 2026 to be
a year defined by continued investment,
www.semiconductordigest.com
deeper collaboration with development
partners, and strong execution on our
growing pipeline. The long-term drivers
for advanced packaging are firmly in
place, and ClassOne is committed to
supporting our customers as they scale
for the next phase of semiconductor
innovation.
Multi-Modal Workflows
and AI Acceleration Are the
Future of Failure Analysis
THOMAS RODGERS, S e n i o r
Director of Market
Str ategy, Head of Business
Sector Electronics, ZEISS
Microscopy
As semiconductor
devices become
more complex and
densely integrated,
the industry
requires failure
analysis (FA)
approaches that
THOMAS
involve orchesRODGERS=
trated workflows.
We are seeing more
correlated, multi-modal approaches
that combine non-destructive solutions
for defect localization (high-resolution
X-ray computed tomography, optical
wavefront/phase imaging, thermal/
infrared, and light and acoustic microscopy) with destructive solutions,
such as focused ion beam (FIB) and
transmission electron microscopy
(TEM), being used for final analysis.
In 2026, we expect to see enhanced use
of artificial intelligence (AI) models
that improve the speed and cost of these
workflows. AI will increasingly get
better at classifying failure types and
their root causes to enable rapid triage
and reduce the bottleneck of scarce,
highly expensive analysis, such as
TEM.
For example, hybrid bonding is promising because it enables chiplets and
heterogeneous integration for advanced
package systems. However, the range
of Cu-to-Cu bond sizes is projected to
push the boundaries to 400nm interconnection pitch. As a result, fault isolation
and FA for IC packages involving
hybrid bonds are expected to face more
challenges.
3D X-ray microscopy (XRM) is
an effective, high-resolution imaging
and analysis tool used for non-destructive defect location, particularly
in package-level failure analysis. Since
XRM delivers submicron resolution,
the emergence of novel high-density
interconnect technologies challenges the
resolution limit of traditional XRMs.
One approach is our 3D X-ray nanotomography technique that is capable of
imaging with 50nm resolution. Using
a multi-modal approach, the correlated
FIB-SEM workflow follows to image interconnect structures or defects at precise
fault regions with nanometer accuracy
based on the 3D nanotomographic XRM
data already acquired. Long scan times
can be a limiting factor for XRM, so
we utilized an AI-powered software,
ZEISS DeepRecon, to speed up X-ray
data acquisition by a factor of four. With
appropriately prepared samples, the developed workflow enables successful and
efficient FA from die metals, RDLs, and
microbumps to hybrid bonds in complex
IC packages. This approach successfully
demonstrates the value of new multimodal approaches.
Looking ahead, we foresee that these
multi-modal datasets will fuel the development of AI-based predictive models
capable not only of identifying defects
and their root causes but also of forecasting their occurrence and underlying
causes from the earliest symptoms.
Test as the Enabler of
the AI Factory
SHANNON POULIN, P r e s i d e n t ,
S e mico n d u c t o r Te s t
Di v i sio n , Te r a d y n e
As we head into 2026, it’s impossible
to ignore that AI is shaping the future
of our industry. The rapid build-out of
Semiconductor Digest Supplement to January 2026
| 15
