Sign in
ASML

ASML: the EUV read hit, the revenue number missed

We reconstructed ASML 2026 EUV output from fab-level tallies and got it right. The secondary revenue call, pinned to Street, came in light.

By Tessara·Published Jul 13, 2026·Graded Jul 15, 2026
What we staked, and how it graded1 hit · 1 miss
The call · primary thesis
ASML's 2026 Low-NA EUV output lands at 65-67 systems (about 69 including High-NA)
Wrong ifASML's disclosed 2026 Low-NA EUV capacity or reconciled full-year shipments land materially outside 65-67 systems.
The print · Jul 15, 2026
ASML's July 15, 2026 Q2 materials put planned 2026 Low-NA EUV capacity at approximately 65 systems, against our 65-67 same-basis estimate frozen one day earlier. Capacity and shipments are not identical measures; the full-year shipment count is tracked separately, and this grade reads the company's own planning figure.
Hit
The call · secondary
Revenue €8.9B (range €8.5B to €9.3B)
Wrong ifReported revenue diverges materially from Street (€9.0B).
The print · Jul 15, 2026
We called Q2 revenue near Street consensus of €9.0B, and the actual print came in at €9.3B, above the high end of ASML's own guidance. While our neutral call was directionally defensible (we predicted revenue would land near consensus and it did land in our stated range of €8.5B-€9.3B), the beat was driven by installed base management sales coming in ~€300M above guidance, a dynamic we underweighted. The €9.3B print represents a ~3.3% beat versus Street, which is at the boundary of 'material divergence' but within our pre-stated range. We grade this a B rather than an A because the neutral framing underappreciated the strength of the upgrade cycle and the installed base contribution. Management's language was unambiguously bullish: full-year guidance was raised to €43B-€45B, capacity expansion plans of 30% for both 2027 and 2028 were announced, and memory net system sales were guided to grow over 75% for the year. We correctly identified the thesis test as 'sustained pull across logic and memory versus digestion,' and the transcript overwhelmingly confirmed sustained pull, logic foundry net system sales growing over 25%, memory over 75%, and EUV system sales growing over 45%. Our neutral stance missed the magnitude of the acceleration. We overweighted the possibility of 'digestion' and underweighted the installed base management tailwind and the breadth of capacity expansion signals. The falsifier condition, revenue diverging materially from Street, was borderline: €9.3B vs €9.0B is a 3.3% beat, meaningful but within our range. On balance, the call was not wrong but was insufficiently bullish given the strength of the print and forward guidance.
Miss
Locked Jul 12, 2026 · resolved Jul 15, 2026All calls on the record →
This piece is open because its calls have resolved. We publish the reasoning in full once a call resolves, hits and misses alike. The calls still running are part of the seat.

Most investors will spend Wednesday asking whether ASML beat or missed by a few hundred million euros. We think that's the least interesting part of the quarter.

The bigger question is whether the AI infrastructure buildout is still accelerating, or whether customers are finally pausing after two years of extraordinary spending. That's what we'll be watching, and here's why.

On the headline revenue number, we actually agree with Street. Our Q2 estimate is €8.9B against frozen consensus of €9.0B, a gap of roughly 1%, well inside our in-line band.

Our view at a glance
MetricValue
Revenue (our model)€8.9B
Street consensus€9.0B
StanceIn line
Range (p10-p90)€8.5B to €9.3B
What decides itFY26 guidance direction

The question is whether customers are still expanding leading-edge capacity faster than expected, or starting to pause while they absorb the capacity already ordered.

  • Buildout continuing: ASML narrows its FY26 guidance upward and continues to report healthy memory and DUV demand.
  • Slowdown case: guidance remains wide, China revenue falls materially below the guided 20%, or management becomes more cautious on orders and lead times.
Today, we lean toward the buildout continuing

The best clue comes from ASML's own customers.

ASML sits at the center of the leading-edge semiconductor supply chain, so its future demand is often visible through the investment plans of companies like TSMC, Samsung, SK Hynix, Intel and Micron before ASML reports.

Over the past few weeks:

  • Micron: signed a multi-year EUV supply agreement with ASML, locking in future demand.
  • SK Hynix: increased capex with a strong focus on EUV-enabled expansion, a direct pull on future EUV orders.
  • Intel: raised capex as EUV production ramps, a sign of continued leading-edge investment.
  • TSMC and Samsung: continue to support strong leading-edge investment over the coming year, keeping two of ASML's largest EUV buyers committed.

Taken together, those signals do not look like an industry preparing to slow down.

The number the market no longer gets

ASML quietly stopped reporting quarterly bookings in fiscal 2026. Backlog is now only disclosed annually.

That matters because bookings were one of the best early signals of future demand.

An EUV machine costs roughly €180M to €350M and can take well over a year from order to showing up as revenue. Revenue tells you what customers decided to buy a year or two ago. Bookings told you what they were deciding today.

That signal has disappeared.

So we built our own way of tracking it.

Instead of relying on a number ASML no longer reports, we estimate how many EUV systems the company is on pace to recognize during 2026 using the information it still discloses: Q1 EUV revenue, system pricing, lithography shipments, High-NA deliveries, backlog and our own ledger of named customer orders.

Today, Tessara's read is:

  • ASML plan: 60+ Low-NA systems in 2026 (up from 44 in 2025)
  • Tessara estimate: ~69 total EUV systems
  • Range: 56 to 72
  • Status: tracking instrument, not a forecast

One important point: our estimate includes both Low-NA and High-NA systems, while ASML's official target only covers Low-NA. Adjusted to the same basis, our estimate sits very close to ASML's own plan.

We are not claiming to see a hidden surge in machine demand. Our historical error on this model averages eight machines across six simulated pre-print runs, so we treat it as a tracking instrument rather than a forecast.

Its value is that it gives us a way to follow the EUV buildout every quarter, even though the market no longer has a bookings number to rely on.

Wednesday is its first real test.

Three things to watch Wednesday

This is the framework we are using, stated plainly.

Watch on WednesdayWhat it signals
FY26 guidance directionNarrows upward = buildout on track; stays wide or unchanged = slowdown risk.
China revenue mixAt or above the guided ~20% = in line; well below = demand or export-control risk.
Orders and lead timesHealthy memory and DUV demand = buildout continuing; softening lead times = slowdown.
Tessara Pro
You read this one after it landed
Pro is where you read them before: every open call with its wrong-if line and its deadline, the live constraint map, and the earnings desk. A free account follows the names you pick and opens each call the day it grades.
Not ready? Get the free weekly read.
Free. No spam. 1 to 2 memos a week.
The record →Tessara is independent research, not investment advice. Past performance is not indicative of future results.

Tessara is independent research and analysis, not investment advice. We are not a registered investment adviser. Past performance is not indicative of future results, and you are responsible for your own investment decisions.