# ASML, TSMC, Samsung and Intel Agree on AI Chip Roadmap Through 2033 > Source: > Published: 2026-09-08 06:34:47+00:00 *ASML, TSMC, Samsung and Intel just told the entire chip industry how AI processors will get made for the next seven years, and it starts with photomasks four times the size of today's.* Four companies control nearly every advanced chip on earth. On September 8th they laid out where that supply chain is headed. ASML and TSMC announced a joint initiative to move the industry from today's 6-inch photomasks, the templates that etch circuit patterns onto silicon wafers, to a new 12-inch format. Samsung joined the same day, expanding its existing partnership with the Dutch lithography maker. Intel is already running the machines the plan depends on. The numbers, laid out in a joint statement carried by GlobeNewswire, are specific. The group is targeting a 12-inch mask pilot line by 2031, with 12-inch High-NA EUV lithography systems entering advanced-node production by 2033. Samsung will bring ASML's High-NA EUV machines into high-volume manufacturing by 2028, starting with DRAM. TSMC follows in 2030. Each machine runs about $400 million, according to Bloomberg's reporting on the announcement. Four hundred million dollars. For one machine. Intel got there first. The company has already run more than a million wafers through its High-NA EUV tools, according to Tom's Hardware - more than the rest of the industry combined. It's also the company trailblazing the larger 6x12 photomasks the new initiative depends on. For years, that looked like a lonely, expensive gamble. TSMC and Samsung waited. Now the other two are committing to the same machines, on their own timelines but the same technology. [TSMC posts record revenue and hands Wall Street its verdict on the AI buildout](https://startupfortune.com/tsmc-posts-record-revenue-and-hands-wall-street-its-verdict-on-the-ai-buildout/) TSMC reported record second-quarter revenue of about $39.62 billion on Thursday, a print Wall Street is treating as a referendum on the entire AI infrastructure trade after chip stocks lost well over a trillion dollars in value since June. Sold-out CoWoS packaging capacity and ASML's second straight guidance raise suggest the demand behind the... - [TSMC record revenue AI spending](https://startupfortune.com/tsmc-posts-record-revenue-and-hands-wall-street-its-verdict-on-the-ai-buildout/) - [whether AI bubble is real](https://startupfortune.com/tsmc-posts-record-revenue-and-hands-wall-street-its-verdict-on-the-ai-buildout/) ## The Stencil Problem You can think of a photomask as the stencil a fab uses to print a chip's circuitry onto a wafer. The 6-inch mask standard dates back decades. It forces engineers to stitch together multiple exposures to build the enormous, transistor-dense dies that AI accelerators now need. Stitching adds cost. It adds defects too, and caps how big a single chip can get. A 12-inch mask, four times the area, removes a lot of that stitching requirement in one move. According to the ASML-TSMC announcement, that combination is expected to raise fab productivity and lower the cost per chip - which matters directly to a customer like Nvidia, buying tens of billions of dollars of wafers a year. ## Why TSMC and Samsung Are Suddenly In It also matters because nobody sells High-NA EUV machines except ASML. The company already holds a monopoly on the extreme ultraviolet lithography that makes any leading-edge chip possible. This initiative extends that position by years. TSMC, the world's largest contract chipmaker, spent much of the past two years hesitant about High-NA: trade outlets reported the company balked at the machine's price relative to the productivity gains on offer. That hesitation is over. TSMC is now locked into a 2030 adoption date, publicly, alongside Samsung and Intel. Samsung's move is the more interesting bet of the two latecomers. Samsung Electronics' announcement, carried on its own newsroom site, frames the 2028 adoption as starting in memory production rather than logic - a sign the company wants High-NA's precision for DRAM scaling before it commits the tool to its foundry business competing directly with TSMC. Two years ahead of TSMC, on the same $400 million machine. That's not a small gap in a business where node timing decides who wins the next generation of AI chip contracts. You won't see any of this on a chip that ships this year, or next. The pilot line isn't due until 2031. Production-ready 12-inch High-NA systems aren't expected until 2033. But every AI accelerator Nvidia, AMD or a hyperscaler's custom silicon team designs between now and then gets built against the assumption that this roadmap holds. Miss a date. The whole industry's cost curve for the next decade shifts with it. That's the real weight of this week's announcement. It confirms that ASML, TSMC, Samsung and Intel - the four companies that already decide what an advanced chip costs - are the same four deciding what it costs next. No government subsidy programme or rival lithography startup has come close to breaking that grip, and this initiative just extended it through 2033. **Also read:** [Sequoia's Konstantine Buhler Bets AI Will Claim 99.9% of Human Thinking](https://startupfortune.com/sequoias-konstantine-buhler-bets-ai-will-claim-999-of-human-thinking/) • [Firmus Kicks Off Roadshow for a Landmark Nvidia-Backed Australian AI IPO](https://startupfortune.com/firmus-kicks-off-roadshow-for-a-landmark-nvidia-backed-australian-ai-ipo/) • [DeepSeek Bets $2.6 Billion on Huawei Chips It May Wait a Year to Get](https://startupfortune.com/deepseek-bets-26-billion-on-huawei-chips-it-may-wait-a-year-to-get/) ## Join the discussion [Open in the community →](/community/) Almost there. Sign in and your reply posts straight away. [Arm CEO Rene Haas Says Chip Shortages Are Slowing Down AI Cancer Research](https://startupfortune.com/arm-ceo-rene-haas-says-chip-shortages-are-slowing-down-ai-cancer-research/) Arm CEO Rene Haas told the BBC that AI could eventually crack cancer by modeling how DNA markers drive tumors, but said the breakthrough is being held back by a shortage of chips needed to build AI data centers. He also predicted humanoid robots doing repair and manufacturing work within a decade and downplayed fears of mass AI job losses. - [chip shortage slowing down AI cancer research breakthroughs](https://startupfortune.com/arm-ceo-rene-haas-says-chip-shortages-are-slowing-down-ai-cancer-research/) - [how chip shortages are impacting AI medical research](https://startupfortune.com/arm-ceo-rene-haas-says-chip-shortages-are-slowing-down-ai-cancer-research/)