High NA EUV Goes Mainstream, China's DUV Break & Intel's Third Price Hike

Jamie Cole (0:00)

The Hardware & Chips Daily News.

Uh, I'm Jamie Cole. Thanks for joining me. Today, ASML's $400 million CapEx bet, Samsung and TSMC signal High NA turning point.

Samsung and TSMC have both committed to ASML's High NA EUV machines, and that changes the calculus for the entire advanced semiconductor supply chain. These aren't tentative pilot agreements. Samsung is targeting memory production using High NA EUV by 2028 TSMC is aiming for advanced logic chips by 2030 That's a coordinated multi-year CapEx cycle built around machines that cost $400 million each. The signal here is that the ROI case is now settled, at least for the two foundries that matter most. Part of what makes this significant is a parallel shift in photomask standards. The industry currently runs on 6-inch photomasks. The transition to 12-inch unlocks roughly 40% better throughput. That's not a minor efficiency gain.

It's the kind of structural improvement that compounds across every wafer run, and it requires the entire supply chain to move in step.

Samsung and TSMC committing together makes that coordination credible. ASML is the only manufacturer of EUV machines on the planet. That position makes every adoption timeline a direct revenue event for one company. The open question is whether ASML can expand capacity fast enough. Its current target is EUV accounting for 30% of revenue by 2027, but a sustained two-foundry adoption wave could push beyond that.

While the West has been focused on EVF export restrictions, China has quietly shifted the terrain on EUV.

Shanghai Aishengna is now shipping domestic immersion EUV lithography machines to Chinese foundries. That matters because the export control strategy was built on a single assumption that ASML held a monopoly across both EVF and DUV.

That assumption no longer holds cleanly. SMCIC and other Chinese foundries can now scale mid-tier production using multi-pattern techniques without Western equipment. The important distinction is that this isn't an EVEV replacement.

Multi-pattern at advanced nodes carries real yield in cost penalties. Whether state subsidies can absorb those economics at commercial scale is genuinely unresolved.

But the geopolitical frame has shifted. Containment via equipment denial is harder when domestic alternatives exist, even imperfect ones. Here's the thing. Intel is taking a different kind of risk on the CPU side. A third price hike is coming on October 5 That follows a 10% increase in the first quarter and a 15-17% increase in July. Three hikes in nine months, against a backdrop of volumes already down around 8% year on year. The tension is straightforward. Intel is defending margins. But every time it raises prices while AMD holds flat, the competitive gap widens for buyers who have flexibility. Industrial OEMs may not have that flexibility in the short term, but they're paying attention. The small core end-of-life situation sharpens that pressure. Intel is now applying a gross margin threshold test to small core product survival. For industrial buyers, this means the evaluation clock has started. Qualification cycles for embedded platforms can run 12 months or longer. If the formal EOL window compresses, some buyers will simply run out of time. Qualcomm and MediaTek are ready for that displacement. Qualcomm's Dragonwing Q-2390 and MediaTek's Iq-2390 are already shipping. The thermal advantage of integrated SoC designs over Intel Atom derivatives is meaningful in sealed, fanless industrial enclosures. That's not a future pitch. The silicon exists now. Consider this. AMD raised its AI total addressable market forecast from $2 trillion to $3 trillion by 2030 The stock moved up roughly 6% on the announcement. APE of around 130 times earnings is pricing in significant execution on Instinct, EPYC, and the Helios platform. The market liked the number. The risk is that a larger TAM forecast also raises the bar on what a miss looks like. Separately, Egypt launched a sovereign AI data center partnership with Nvidia, Vodafone Business, Elsewedy Electric, and Cassava Technologies. The first phase is $200 million, scaling toward 200 NW and $1 billion total. The GPU as a service model keeps data in country. That's the template other emerging markets are watching closely.

The near-term signal is worth tracking. Whether ASML's capacity expansion keeps pace with the Samsung and TSMC adoption timelines. Whether China's domestic DUV machine deliver commercially viable yields at 14-7 manometers. And whether Intel's October price hike accelerates the industrial arm migration faster than anyone is currently modeling. The high NA cycle is real, the containment logic around China is being stress-tested, and Intel's pricing decisions are quietly reshaping the industrial computing landscape. Those are the threads to watch. Thanks for listening. This podcast was built using AI technology.

A YesWe production.