When the US enacted the CHIPS Act four years ago, the first goal was to revitalize domestic semiconductor manufacturing. Out of the $59 billion budget, two-thirds went to grants and loan guarantees for foundries. Intel and Taiwan Semiconductor Manufacturing Company (TSMC) built plants in Arizona. Samsung invested in Texas. Micron invested in New York.
But new CHIPS Act funding pivots away from treating success as a simple count of chip production. It funds a broad range of artificial intelligence computing capabilities: advanced packaging, interconnects, memory, photonics, materials and secure electronics.
The pivot is welcome. The US will never be entirely self-sufficient in chips. Instead, it needs to build US capability at critical bottlenecks, and avoid asking US taxpayers to finance every layer of a global supply chain. If it is to meet the challenge from China, Washington needs, above all, to work with trusted democratic allies – seeing them as assets, not as dangerous competitors.
The clearest evidence of the strategic shift comes with the announcement over the summer of $874 million targeting the connective tissue of the AI compute stack. GlobalFoundries could receive up to $300 million to develop optical data transfer equipment — an answer to the growing data movement bottleneck in large AI models. Kepler could receive up to $245 million for next-generation AI memory, and Multibeam up to $140 million for equipment to stack and connect multiple chips.
The remaining proposed awards support low-power computing (Extropic), specialist materials (Aeluma), and software to spot counterfeit or malicious components (OBSIDIA).
These grants represent a useful reminder that AI leadership no longer depends solely on making ever smaller transistors. Key solutions require tying together processors, high-bandwidth memory, optics, and specialized chips. Advanced packaging allows multiple chiplets to be stacked and connected at high density; co-packaged optics reduces the energy and latency cost of moving data between AI processors; advanced memory addresses one of the principal limits on AI training and inference.
The new US chip subsidies aim to create enduring leverage in AI compute, rather than attempting to recreate, at taxpayer expense, Taiwan’s mature foundries. At $874 million, the funding remains modest in an industry used to speaking in billions of dollars.
But it is significant because small and mid-sized companies will receive the most money at points where public support can unlock private capital. The funding avoids permanent state ownership or unlimited operating subsidies. It reduces first-mover risk at strategically vital bottlenecks, while firms, customers, and investors carry the main commercial burden.
The US has historically excelled at designing chips – but lacked the manufacturing expertise found in Taiwan, Japan, South Korea, and parts of Southeast Asia. Cooperation is essential. TSMC’s planned $165 billion US expansion creates leading-edge US capacity. It brings Taiwanese expertise, supplier relationships, and private capital, strengthening the US–Taiwan industrial partnership. The US gains capacity. Taiwan retains its central place in chipmaking.
Germany’s Bosch investment in the production of silicon-carbide chips in California offers a parallel example. It strengthens US supply security in power electronics without requiring Washington to invent a national competitor from scratch.
Another good example is the US-JOINT, which connects 12 US and Japanese equipment and materials firms to develop next-generation packaging technologies. These include cleanrooms and evaluation lines for patterning, bonding, molding, plating, and analysis: the “backend” capabilities needed to turn good US-designed chips into complete AI systems. Rather than building a publicly owned domestic champion, Washington is helping build an allied R&D platform.
The $874 million CHIPS commitments are so far only letters of intent, not final awards; they remain subject to due diligence and negotiation. Success will depend on disciplined execution. Washington should continue to welcome foreign investment with trusted allies. US strategic advantage will not come from trying to out-subsidize China or recreate Taiwan in Arizona. It will come from owning the technologies that define next-generation AI.
Christopher Cytera CEng MIET is a senior fellow with the Tech Policy Program at the Center for European Policy Analysis and a technology business executive with over 30 years of experience in semiconductors, electronics, communications, video, and imaging.
Bandwidth is CEPA’s online journal dedicated to advancing transatlantic cooperation on tech policy. All opinions expressed on Bandwidth are those of the author alone and may not represent those of the institutions they represent or the Center for European Policy Analysis. CEPA maintains a strict intellectual independence policy across all its projects and publications.
Tech 2030
A Roadmap for Europe-US Tech Cooperation
