Advanced semiconductor manufacturing is currently dominated by two companies, TSMC and ASML, with Samsung and Intel operating at a smaller scale. TSMC (Taiwan Semiconductor Manufacturing Company) operates numerous fabs and R&D facilities globally including Taiwan, the USA and Japan and is currently the largest contract chipmaker. ASML, a Dutch company headquartered in Veldhoven, Netherlands, builds the extreme ultraviolet (EUV) lithography systems needed to fabricate the world’s most advanced chips.
For many years, these companies and others have pushed the boundaries of chipmaking with ever-decreasing sizes of chip transistors, the building block of modern computing. Around six years ago,
However, the current model of advanced semiconductor manufacturing may be approaching a technical breakpoint. Advanced chips are defined by how small their transistor features are, and lithography is the most critical and expensive step in producing them. Today’s leading-edge chips rely on EUV lithography, a technology so complex that only ASML can currently build the machines. As features shrink further, manufacturers are forced to use multi-patterning, which increases cost, complexity, and defect risk.
Despite technical progress such as High-NA EUV, which is used for fabrication around the 2 nm region, the economics are worsening. These future leading-edge fabs are expected to cost around 10 - 20x more than current fabs, concentrating advanced chip production into the hands of a few massive players and locking out smaller companies and new entrants.
However, a new US-based company calling themselves Substrate has vowed to shake things up with their proposed X-ray lithography. X-rays have much shorter wavelengths than EUV, theoretically allowing much smaller features to be printed in a single exposure. While X-ray lithography has existed in theory for decades, it was previously impractical due to the need for massive synchrotrons and the difficulty of controlling X-rays.
Substrate claims recent advances in compact particle accelerators, optics, and materials now make X-ray lithography feasible inside a factory-scale tool. Their system reportedly uses a compact accelerator to generate X-rays within the lithography machine itself. The company claims to have demonstrated printing of 12 nm features suitable for sub-2 nm transistors, and single patterning for all layers, avoiding costly multi-patterning. They also claim feature consistency across a wafer with an accuracy of around 0,25 nm.
Unlike ASML, Substrate does not want to sell its machines. Rather, it intends to build its own fabs in the USA and operate as a foundry. This will put it in direct competition with the likes of TSMC and Samsung. However, success is far from guaranteed. TSMC’s dominance is built on decades of process knowledge, massive scale, and learning across dozens of factories. Moving from a lab demo to high-volume, reliable manufacturing could take a decade or more, just as EUV did.
However, if X-ray lithography succeeds, the consequences could be enormous; dramatically lower costs for advanced chip manufacturing, and increased compute availability for AI and future technologies. There could also be major implications to worldwide economic power.
Advanced semiconductor manufacturing has reached a point where technical progress is increasingly constrained by economic reality. Substrate’s proposed use of X-ray lithography represents a bold attempt to reset these economics.
Whether this approach succeeds remains uncertain. The challenges of materials, yield, throughput, and manufacturing discipline are immense, and history shows that success at scale requires decades of accumulated knowledge. Yet if X-ray lithography can deliver even a portion of its promised cost and simplicity advantages, it could reshape the current semiconductor manufacturing climate.
| Tel: | +27 11 543 5800 |
| Email: | [email protected] |
| www: | www.technews.co.za |
| Articles: | More information and articles about Technews Publishing |
© Technews Publishing (Pty) Ltd | All Rights Reserved
printer friendly version