{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreicmdnpolmi6f6uepntksycrpvuchjtf6lrsz7oxys2hgy3iqpu474",
    "uri": "at://did:plc:ls6rbbwjyqeakittcsi3k6x3/app.bsky.feed.post/3mfomhupgo6u2"
  },
  "coverImage": {
    "$type": "blob",
    "ref": {
      "$link": "bafkreiaaglhn7t6rfgywmmqphxernvyvlids5dsusnviz42lgzpeptpnvu"
    },
    "mimeType": "image/png",
    "size": 351388
  },
  "description": "ASML just found a way to push its EUV light source to 1,000 watts — what does that mean for machine pricing, the High-NA timeline, and the quiet revolution happening in DRAM fabs?",
  "path": "/from-photons-to-pricing-power-asmls-1-000-watt-chain-reaction/",
  "publishedAt": "2026-02-25T11:50:12.000Z",
  "site": "https://www.jasonandjarvis.org",
  "tags": [
    "previous 4Q25 Earnings Review",
    "SK Hynix's first \"production-grade\" High-NA EUV installation",
    "earlier analysis on how a Memory Capex surge could rewrite ASML's 2026 outlook"
  ],
  "textContent": "## The 1,000-Watt Chain Reaction: How ASML's Light Source Breakthrough Is Redefining the Economics of EUV\n\nOn February 23, 2026, Reuters reported what appeared to be a deeply technical piece of news — yet one with potentially far-reaching implications for the semiconductor industry: ASML's researchers have found a way to push EUV light source power to 1,000 watts.\n\nThe number alone won't make headlines outside the industry. But for investors who understand ASML's business model, the implications run far deeper than the literal breakthrough suggests. It strikes at the core mechanism behind EUV lithography pricing power, shifts the strategic balance between Low-NA and High-NA, and could accelerate a long-brewing wave of DUV displacement.\n\nTo understand why those 1,000 watts matter so much, you need to trace a chain that runs through ASML's entire EUV business.\n\n## Light Source Power: The First Principle of EUV Economics\n\nThe basic principle of EUV lithography isn't complicated: molten tin droplets are fired into a chamber at roughly 50,000 droplets per second (the new approach doubles this to approximately 100,000), a high-power CO₂ laser heats them into plasma hotter than the surface of the sun, the plasma emits extreme ultraviolet light at a wavelength of 13.5 nanometers, and a precision optical system supplied by Carl Zeiss guides that light onto the wafer surface to \"print\" chip patterns.\n\nA stronger source means shorter exposure times. Shorter exposures mean more wafers processed per hour — that is, higher WpH (Wafers per Hour). The historical data reveals a strikingly linear relationship: roughly every 1 watt of source power corresponds to 0.33 WpH (at a dose of 30 mJ/cm²).\n\nASML's current flagship Low-NA platform, the NXE:3800E, operates at around 600 watts, delivering approximately 220 WpH. The existing roadmap targets 250 WpH for the next-generation NXE:4000F and about 280 WpH for the NXE:4200G. But the 1,000-watt breakthrough means that a figure once regarded as a \"long-term R&D aspiration\" — roughly 330 WpH — now has a concrete technical path to realization.\n\nEUV Light Source Power vs. Wafer Throughput (WpH) Roadmap\n\nsource: Bernstein\n\nWhat's worth noting is that this wasn't a fleeting lab demonstration. Michael Purvis, ASML's EUV light source technology lead, emphasized in an interview that this is a system capable of delivering a stable 1,000 watts in a customer environment. ASML already sees a clear path to 1,500 watts and has stated there is \"no fundamental reason preventing them from reaching 2,000 watts.\" At the 2024 Capital Markets Day, the company had listed 1,000 watts as a long-term goal but candidly admitted it hadn't yet entered R&D. The speed of this transition — from \"not yet touched\" to \"we've found a way\" — is itself worth paying attention to.\n\nIn fact, during the 4Q25 NDR (Non-Deal Roadshow) in early February this year, ASML had already confirmed to institutional investors that the 1,000-watt light source power had been achieved in factory testing.\n\n## ASML's Pricing Flywheel: Why Faster Machines Mean Pricier Machines\n\nIf an ordinary equipment manufacturer increased its machine throughput by 50%, the market would typically expect prices to hold steady — or even fall. ASML is not an ordinary equipment manufacturer.\n\nASML operates on a **value-based pricing model** : a machine's selling price is directly tied to the value it creates for the customer — specifically, how many wafers it can process per hour. Under this logic, EUV machine ASP (Average Selling Price) and WpH exhibit a near-perfect linear relationship, with a historical regression equation of **y = 0.9x + 17.3** (EUR Mn) — a slope of roughly €0.9M per additional WpH.\n\nEUV Blended ASP vs. Throughput: The Linear Relationship (Historical and Projected)\n\nsource: Bernstein\n\nWhat does this imply? Taking the current NXE:3800E's ~220 WpH and ~€220M ASP as the baseline, if a future platform equipped with a 1,000-watt source (hypothetically, the NXE:4400H) achieves ~330 WpH, then the same pricing relationship would put its ASP at approximately **€350M** — roughly **60% above** the 2025 blended ASP.\n\nOn the cost side, the engineering improvements needed to reach 1,000 watts — doubling the tin droplet rate and adopting a dual-laser pre-pulse scheme — will introduce some incremental costs, but the vast majority of core components (especially the optical system) remain unchanged. This means ASP growth will continue to outpace cost increases, with gross margins poised for further expansion. In my previous 4Q25 Earnings Review, I noted that the 2026 gross margin guidance of 51–53% came in below expectations, partly because the product mix was still weighted toward the older, slower 3600D. But as newer, higher-throughput platforms ramp, margin expansion is simply a matter of time.\n\nOver the past six years, EUV blended ASP has climbed from approximately €100M in 2018 to around €220M in 2025.\n\nHistorical Co-evolution of EUV Blended ASP and Throughput\n\nsource: Bernstein\n\nThis relentless upward price curve is the most elegant aspect of ASML's business model: **customers don't complain about more expensive machines because the manufacturing cost per wafer is actually falling.** For TSMC or Samsung, a single €350M machine processing 100 more wafers per hour is far more economical than purchasing a second €220M unit. ASML shares the productivity gains with its customers and itself — a textbook win-win flywheel.\n\n## Implication One: When Low-NA Becomes \"Too Good\" — High-NA's Timeline May Slip\n\nThis flywheel creates a fascinating strategic tension.\n\nASML currently advances two product lines in parallel: the Low-NA (numerical aperture 0.33) NXE series and the High-NA (numerical aperture 0.55) EXE:5000 series. High-NA's core selling point is superior resolution — it can print features 1.7x smaller, boosting density by up to 2.9x. But it also comes with clear drawbacks: a unit price of €360M (vs. €200–250M for Low-NA), current throughput of only ~175 WpH, and an exposure field half the size of Low-NA's, placing higher demands on yield and process integration.\n\nAs I discussed in my piece on SK Hynix's first \"production-grade\" High-NA EUV installation, High-NA's path to volume production unfolds in three stages: R&D validation, limited-layer adoption, and full deployment. Even though Intel announced acceptance of its EXE:5200B system in January 2026, ASML CEO Christophe Fouquet noted on the 4Q25 earnings call that \"whether DRAM or Logic will be the first to bring High-NA into volume production remains a neck-and-neck race.\"\n\nNow, if Low-NA EUV source power reaches 1,000 watts and throughput hits 330 WpH, a very practical question surfaces: **do customers still need to migrate to High-NA as urgently?**\n\nHigh-NA Throughput Roadmap\n\nsource: Bernstein\n\nConsider the comparison: by 2029, the latest EXE:5000D on the High-NA roadmap targets approximately 195 WpH. Meanwhile, if Low-NA equipped with the 1,000-watt source is already achieving 330 WpH in the same timeframe, the latter's single-exposure throughput is nearly 1.7x that of the former. High-NA's resolution advantage cannot be replicated by Low-NA — that much is true. But for layers where multi-patterning can compensate for the resolution gap, **a much faster Low-NA combined with multi-patterning may offer better cost-effectiveness than a slower but finer High-NA**.\n\nThis isn't necessarily bad news for ASML. Put conservatively, it means Low-NA's lifecycle and revenue contribution could be longer and stronger than the market previously anticipated. And let's not forget — the EUV light source is **shared** between Low-NA and High-NA platforms. The 1,000-watt breakthrough will equally boost High-NA's future throughput. According to Bernstein's analysis, High-NA may deliver higher gross margins than Low-NA once it reaches maturity, thanks to a shorter learning curve and extensive reuse of common components. It's just that this \"maturity\" may arrive a bit later than previously envisioned.\n\n## Implication Two: DRAM — Fewer Layers, Greater Leverage\n\nIf the implications for High-NA involve a subtle strategic calculus, the impact on the DRAM market is far more direct — and more bullish.\n\nLeading-edge logic chips (such as the most advanced 2nm node) currently use around 20–25 EUV mask layers, while DRAM's EUV layer count is significantly lower — from 2 layers at the 1a node to approximately 6 at the current 1c node (which both Samsung and SK Hynix have reached). According to ASML's projections at the 2024 CMD, DRAM EUV layers could increase to roughly 10 by the 0a node in 2030, with SPIE papers even hinting at 13–16 layers.\n\n**But this is precisely the point: because DRAM uses relatively few EUV layers, throughput improvements on each layer have a disproportionately large marginal impact on overall manufacturing efficiency.**\n\nLet me walk through the economics with concrete numbers. According to a BNP Paribas report published in February this year titled \"DRAM EUV Paradox\": upgrading from the 1a to 1c node, while increasing EUV exposure steps from 4 to 6, **actually reduces total production steps from 162 to 138** — because each additional EUV layer replaces multiple DUV multi-patterning steps. A 100 kwspm (100,000 wafer starts per month) fab upgrading from 1a to 1c can add approximately 17,000 wafers per month in capacity, translating to roughly $480M in incremental revenue at equivalent die pricing. The required incremental equipment investment is about $1.2B — which, in the current tight DRAM supply-demand environment, can be recouped within a matter of months.\n\nDRAM EUV vs. DUV Exposure Count by Node — Total Exposures Decline as EUV Share Rises\n\nsource: BNP Paribas Exane\n\nThe chart above clearly illustrates the visual logic of this \"paradox\": the orange portion (DUV exposures) shrinks with each successive node, the purple portion (EUV exposures) grows steadily, yet the total height of each bar — the total exposures per wafer — actually declines. From 53 total exposures at the 1a node down to 38 in the most aggressive scenario at 0a, a reduction of nearly 30%. This is the structural efficiency gain from EUV single-exposure replacing DUV multi-patterning (e.g., SAQP requires 4 exposures where EUV needs just 1).\n\nBNP provided a line-by-line breakdown of this economic equation:\n\nFull Economics of DRAM Node Upgrades — More EUV Layers = Fewer Total Steps = Higher Throughput and Revenue\n\nsource: BNP Paribas Exane\n\nThe key takeaway from this table: upgrading from 1a to 1c increases EUV system count from 3 to 8 and EUV tool value from $900M to $2,400M — but ArFi tool count falls from 35 to 32, and total lithography system value rises only from $4,400M to $5,600M, an increment of just $1,200M. Meanwhile, monthly capacity jumps from 100k to 117k wafers/month and monthly revenue from $1,400M to $1,880M. **Annualized incremental revenue of roughly $5.8B against incremental tool costs of just $1.2B — the returns are extraordinary.**\n\nIf, on top of this, per-tool EUV throughput increases by another 50% — from 220 WpH to 330 WpH — then the same number of EUV tools can process approximately 50% more wafers. **For DRAM, where EUV layer counts are inherently low, EUV ceases to be a production line bottleneck and becomes a lever for acceleration.** This significantly strengthens the economic incentive for DRAM makers to adopt more EUV layers — because the \"cost\" of upgrading (more EUV tools consuming line time) gets substantially diluted by the throughput gains.\n\nConnecting this to my earlier analysis on how a Memory Capex surge could rewrite ASML's 2026 outlook — SK Hynix plans to double its EUV tool count by 2027, adding roughly 20 units; Micron announced FY2026 capex of approximately $18B — DRAM customers' demand for EUV is entering a self-reinforcing positive cycle. ASML CEO Fouquet captured it precisely on the 4Q25 earnings call: \"The more DRAM customers use EUV, the more they love it.\"\n\n## Implication Three: The Endgame for DUV Multi-Patterning — EUV Is Closing In on the Cost Crossover\n\nWithin ASML's product portfolio, there's a long-running theme the market has largely overlooked but one with profound significance: when will EUV achieve full economic parity with DUV multi-patterning?\n\nEUV vs. ArFi (DUV Immersion) Throughput Evolution\n\nsource: Bernstein\n\nThis chart presents two starkly different throughput trajectories. ArFi (DUV immersion) throughput is approaching its physical ceiling — the surface tension of the water layer imposes a hard limit on wafer stage speed, capping ArFi at roughly 330 WpH. EUV, by contrast, has climbed from approximately 105 WpH in 2018 all the way up, and with the tailwind of the 1,000-watt source, is on track to **match ArFi's throughput around 2029–2030**.\n\nThe significance of this crossover extends far beyond the symbolic. Since DUV multi-patterning (such as double patterning) requires two or even three exposures for the same layer while EUV needs only one, once EUV's single-exposure speed matches ArFi's, **EUV's effective throughput becomes 2–3x that of DUV multi-patterning**.\n\nPer-Layer Patterning Cost: Multi-Patterning vs. Single-Pass EUV\n\nsource: Bernstein\n\nFrom a cost perspective, this transition is unfolding in stages:\n\n  * EUV became cheaper than DUV triple patterning (logic LE LE LE) as early as 2022–2023 — this is already established fact.\n  * EUV is expected to become cheaper than DRAM double patterning (SA LE LE) around 2026–2027.\n  * And by 2028–2029, **EUV will become cheaper than logic double patterning (LE LE)** — a potential incremental market that Bernstein has not yet incorporated into its model.\n\n\n\nIn other words, light source power improvements don't merely reinforce ASML's pricing power in the existing EUV market. They're quietly opening an entirely new door: the gradual conversion of layers still relying on DUV multi-patterning into potential EUV demand.\n\n## A Light in the Distance\n\nFrom a few watts of faint luminescence in a 2010 laboratory to a stable 1,000 watts in a 2026 factory test, the evolution of EUV light sources is almost a microcosm of the semiconductor industry's \"the impossible is merely difficult\" ethos.\n\nEUV Light Source Power: A Historical Journey\n\nsource: Bernstein\n\nLooking back at ASML's narrative arc over the past eighteen months — from the 2Q25 earnings trough of \"unable to confirm 2026 growth,\" through 3Q25's floor-setting \"no lower than 2025,\" to the €13.2B in 4Q25 orders (nearly double consensus) that confirmed a \"multi-year upswing\" — the light source breakthrough adds yet another layer of structural support to an already ascending curve.\n\nA note of prudence is warranted, though. From technical breakthrough to volume deployment, several years of work remain. The platform corresponding to 330 WpH likely won't reach customer production lines until around 2029–2030. In the interim, the NXE:4000F and NXE:4200G will serve as transitional products, gradually releasing throughput dividends. And the market is unlikely to immediately incorporate a €350M ASP into near-term valuation models.\n\nBut for understanding ASML's core competitive advantage, the significance of 1,000 watts transcends any single quarter's order numbers. It answers a more fundamental question: **on the EUV lithography track, the laws of physics have not yet said \"no\" to ASML — and it doesn't look like they will anytime soon.**\n\nWhen the world's only commercial EUV manufacturer tells you it sees a path to 1,500 and even 2,000 watts, this is more than a technical statement. For ASML's customers, competitors, and investors alike, the message is the same: the frontrunner in this race is nowhere near the finish line — and the price of entry will only keep climbing.",
  "title": "From Photons to Pricing Power: ASML's 1,000-Watt Chain Reaction",
  "updatedAt": "2026-02-25T11:50:12.000Z"
}