Invesaro publishes investment theses and settles them against the market in public: with a date, a basket of companies and a verdict.

Investment thesis

The technological bottleneck in AI is shifting from silicon structures toward advanced chip packaging

LivePublished July 7, 2026180-day horizontechnology

Watch this thesis

In short

What is happening
Demand for powerful AI accelerators stays high
Why it matters
Shrinking transistors physically is getting harder, so the fight for more compute is moving into advanced packaging. The capabilities the industry now depends on are controlled by a handful of suppliers of precision metrology and assembly equipment.
Who benefits
BE Semiconductor Industries, ASMPT, Camtek, Onto Innovation
What must hold
Assumptions are statements that must stay true for the thesis to make sense.
Assumptions: 5. See the assumptions
When is the verdict
The horizon is the time we give a thesis. That many days after publication we settle it: we compare the basket of companies with the market benchmark and give a verdict.
January 3, 2027, 180 days after publication
Our score at publication
A score from 0 to 100 given on the day of publication, before the market has verified anything. On the results page we check in public whether higher-scored theses do better.
50 out of 100

Basket return

-15.5%

equal weight, since publication

The market over the same window

+1.0%

benchmark for this basket

Edge over the market

-16.5%

in percentage points

Companies in the basket

BESIBE Semiconductor Industries 0522ASMPT CAMTCamtek ONTOOnto Innovation

The argument

The bottleneck moves to assembly

The market has been captivated by the front of the chain: Nvidia designs the powerful GPUs, TSMC physically prints them, ASML supplies the lithography equipment, and memory makers stack HBM. The truth is that this view of the industry rests on yesterday's technology. Gone are the days when every gain in performance came from a smaller process node (physically shrinking the circuits on a silicon wafer). Today, added compute comes from cleverly packing more and more separate parts onto a single plane. That means tall vertical HBM memory stacks and large logic dies (chiplets) placed on a shared interposer. The technological bottleneck, and with it the margins and investor attention, is decisively moving off the silicon layer and into the process known as packaging.

What this thesis rests on

Each one is a statement that has to be true. We go back to every live thesis on a schedule and check these against public sources, so each leg carries either what we found or the date we look next.

Archive: reviews of this thesis are paused. Last review September 16, 2026.

TSMC confirms in its quarterly results or annual filing before 31 December 2026 that advanced packaging (CoWoS) capacity is being roughly doubled again, with 2026 capital expenditure at or above the 2025 level.
operational · Open
Checked September 16, 2026, nothing published either way
BE Semiconductor Industries discloses volume production orders or shipments of hybrid bonding systems for HBM4 or logic in a quarterly report published within the 180-day horizon.
operational · Supported

A source confirms this · August 30, 2026

Q2-26 results (23 Jul 2026): revenue EUR 249.9m, +68.7% YoY, driven primarily by hybrid bonding, photonics and datacenter. On the call Besi said Q2 hybrid bonding orders came from two repeat logic customers plus one new customer, and that its hybrid bonding customer base rose from 15 at end-2025 to 21 at end-Q2-26. Units not disclosed; no HBM4 production order yet.
See the source
ASMPT reports semiconductor solutions segment bookings at least 20% higher year on year in both quarterly reports after publication, offsetting its mainstream assembly weakness.
financial · Supported

A source confirms this · August 30, 2026

2026 interim results (29 Jul 2026): SEMI segment bookings +81.9% YoY (+56.9% HoH), against the 20% bar. Group bookings HK$12.75bn (US$1.63bn), +85.1% YoY; advanced packaging revenue US$339.0m in H1, +17% YoY. First of the two required reports; the Q3 announcement is not yet published.
See the source
Camtek reports quarterly revenue of at least USD 115 million in each of the two quarters following publication.
financial · Supported

A source confirms this · August 30, 2026

Q2 2026 results (10 Aug 2026): record revenue of USD 133.2m, above the USD 115m bar, +8% YoY and +10% sequentially. Q3 guidance USD 158-160m. First of the two required quarters; Q3 is not yet reported.
See the source
Onto Innovation's reported gross margin stays above 50% in the two quarters after publication, with no disclosed price concessions to its largest advanced packaging customer.
competitive · Supported

A source confirms this · August 30, 2026

Q2 2026 results (6 Aug 2026): GAAP gross margin 53.4%, non-GAAP 57.0%, both above the 50% bar, on revenue of USD 343.1m (+35.3% YoY). No price concessions to any advanced packaging customer disclosed in the release. First of the two required quarters; Q3 is not yet reported.
See the source

Show the causal chain step by step

Demand for powerful AI accelerators stays high
The end of easy transistor scaling moves innovation to integration (chiplets and HBM)
The HBM4 standard raises the bar for memory stacking (expected 2026)
Advanced packaging becomes the bottleneck for the entire industry
TSMC fabs and OSATs sharply double their purchases of assembly and test equipment
Leaders in modern chip bonding technology (Besi, ASMPT)
Top makers of metrology and precision inspection equipment (Camtek, Onto)

This thesis, over time

What has happened since publication. Every entry is computed as of its own date, from price history.

October 5, 2026

Day 90 of 180: basket -15.5%

The thesis is halfway to its verdict. The equal-weight basket stands at -15.5% since publication day. The benchmark (S&P 500) returned +1.0% over the same window, so the basket trails its market by 16.5 percentage points. The strongest contributor is CAMT (-1.8%), the weakest is BESI (-26.7%). 90 days remain until the verdict.

September 16, 2026

Review on day 71: nothing moved

We went back to this thesis on day 71 of 180 and checked each of its 5 assumptions against public sources. No news either way: 5. The equal-weight basket stands at -17.1% since publication day. The benchmark (.INX) returned +1.1% over the same window, so the basket trails its market by 18.2 percentage points. 109 days remain until the verdict.

August 30, 2026

Review: 4 of 5 assumptions confirmed

We went back to this thesis on day 54 of 180 and checked each of its 5 assumptions against public sources. Confirmed by the sources: 4. No news either way: 1. The equal-weight basket stands at -12.4% since publication day. The benchmark (.INX) returned +2.8% over the same window, so the basket trails its market by 15.2 percentage points. 126 days remain until the verdict.

August 21, 2026

Day 45 of 180: basket -6.9%

The first quarter of the horizon is behind this thesis. The equal-weight basket stands at -6.9% since publication day. The benchmark (S&P 500) returned +1.8% over the same window, so the basket trails its market by 8.7 percentage points. The strongest contributor is CAMT (+4.7%), the weakest is 0522 (-16.8%). 135 days remain until the verdict.

July 29, 2026

Basket fell below -20% since publication

This basket is at -21.1% since publication, 22 days after the thesis was posted. The benchmark (S&P 500) returned -2.5% over the same window, so the basket trails its market by 18.6 percentage points. The strongest contributor is CAMT (-7.8%), the weakest is 0522 (-28.0%). This is not a change of status: the horizon keeps running and the verdict only lands once it expires.

July 27, 2026

Basket fell below -10% since publication

This basket is at -10.4% since publication, 20 days after the thesis was posted. The benchmark (S&P 500) returned -1.2% over the same window, so the basket trails its market by 9.2 percentage points. The strongest contributor is CAMT (+4.9%), the weakest is BESI (-16.0%). This is not a change of status: the horizon keeps running and the verdict only lands once it expires.

This is analysis, not investment advice and not a recommendation to buy or sell anything. We publish it and track it in public, mistakes included. Any decision is yours and yours alone.