TSMC's Chairman Shrugs Off Musk's Chip Plant Ambitions

The Confidence of a Market Leader

Che-Chia Wei stood before shareholders at TSMC's annual general assembly in early June and delivered a message that distilled decades of semiconductor manufacturing expertise into two words: "good luck." The chairman and CEO was responding to a question about Elon Musk's ambitions to construct what the SpaceX founder has termed a "terafab," a chip production facility targeting one million wafers per month. Wei's tone was polite, even gracious, but the subtext was unmistakable. Building cutting-edge fabs from scratch is not a venture for the impatient or the undercapitalized, and TSMC has no intention of losing sleep over new entrants, however well-funded or famous.

At DailyTechWire, we've tracked the foundry wars across Asia for years, watching as Samsung poured billions into EUV lithography and Intel pledged to reclaim process leadership by mid-decade. What Wei's remarks underscore is not complacency but a sober assessment of the barriers to entry in advanced node manufacturing. Those barriers are rising, not falling, and they extend far beyond capital expenditure into talent pipelines, supply-chain orchestration, and the tacit knowledge that comes from running hundreds of thousands of wafers through cleanrooms every week.

The Terafab Proposition

Musk's concept of a terafab, while still largely aspirational, would represent a step change in scale. One million wafers per month is roughly equivalent to the combined output of several large existing fabs. For context, TSMC's largest individual facilities produce in the range of 100,000 to 150,000 wafer starts per month at mature capacity. Scaling to a million would require not only vast physical infrastructure but also access to advanced process equipment, stable yields at leading nodes, and a customer base willing to commit years of volume to justify the investment.

According to Wei, the timeline for such a project would be measured in years, if not decades. Advanced semiconductor fabs typically require three to five years from groundbreaking to volume production, and that assumes access to proven process technology, established supplier relationships, and experienced engineering teams. Musk's ventures in automotive and aerospace have demonstrated an appetite for vertical integration and rapid iteration, but chipmaking is a different discipline. Yield improvement at the 3-nanometer or 2-nanometer nodes is a grinding, empirical process, and there are few shortcuts.

The broader question is whether Musk intends to develop proprietary process technology or license it from an existing foundry. If the former, the challenge multiplies. If the latter, he would likely be negotiating with the very companies Wei dismissed as non-threatening, including Intel Foundry Services or Samsung's contract manufacturing arm. Either path leads back to the same chokepoints: EUV scanners from ASML, advanced packaging know-how, and the ability to manage defect density at the atomic scale.

TSMC's Competitive Posture

Wei's comments also addressed the broader competitive landscape. He stated plainly that TSMC is not concerned about rivalry from Intel, Samsung, or any other player in the foundry market. That assertion rests on a foundation of market share, customer lock-in, and process leadership. As of early 2025, TSMC held approximately 60% of the global foundry market by revenue, with dominance at the leading edge even more pronounced. Apple, Nvidia, AMD, Qualcomm, and MediaTek all rely on TSMC for their most advanced chips, and switching costs are prohibitive in the short to medium term.

Intel's foundry ambitions, while serious, have been hampered by execution missteps and delayed node transitions. Samsung has made gains in certain segments, particularly memory-adjacent logic, but has struggled with yield stability at the most advanced nodes. For TSMC, the competitive moat is not just technology but also the compound effect of customer co-development, where chip designers and process engineers work in tandem to optimize performance, power, and area. That co-development cycle takes years to establish and is difficult to replicate.

Yet Wei's confidence should not be mistaken for invulnerability. The geopolitical landscape is shifting beneath the foundry industry. Export controls on advanced chipmaking equipment, restrictions on technology transfer, and pressure to diversify manufacturing footprints are all forces that could erode TSMC's centrality. The company is building new fabs in Arizona, Japan, and Germany, but these facilities will lag the leading-edge output in Taiwan for the foreseeable future. If Musk or another well-capitalized entrant can navigate the regulatory and technical maze, the competitive dynamics could shift faster than incumbents expect.

The Economics of Scale in Semiconductors

One of the underappreciated aspects of Musk's terafab concept is the economics. Semiconductor manufacturing exhibits both economies of scale and diseconomies of complexity. Larger fabs can spread fixed costs across more wafers, improving unit economics. But as process nodes shrink and feature sizes approach physical limits, the cost per transistor advantage diminishes. At 3-nanometer and below, the capital intensity of each incremental node is so high that only a handful of companies can afford to stay in the race.

A million-wafer-per-month fab would require upfront investment in the tens of billions of dollars, likely exceeding $50 billion when factoring in land, cleanroom construction, process tools, and ramp costs. Depreciation schedules for leading-edge equipment are typically three to five years, meaning the facility would need to maintain near-full utilization and command premium pricing to achieve acceptable returns. That, in turn, requires either captive demand at unprecedented scale or the ability to attract a broad base of customers away from TSMC, Samsung, and Intel, a scenario that seems implausible without a significant process or cost advantage.

Musk's track record in manufacturing suggests he would pursue vertical integration, using the terafab to supply chips for Tesla's full self-driving systems, SpaceX's Starlink satellites, and potentially other ventures. If the facility were designed around a narrow set of custom chips optimized for specific workloads, the economics might pencil out differently than a traditional merchant foundry. But even then, the technical risk remains formidable. Custom chip design is only as valuable as the yield and performance of the underlying process, and achieving competitive metrics at advanced nodes is a multi-year learning curve.

Talent and Tacit Knowledge

Beyond capital and equipment, the foundry business hinges on talent. TSMC employs tens of thousands of engineers, many with decades of experience in process integration, lithography, etch, deposition, and metrology. The company's ability to ramp new nodes faster than competitors is partly a function of institutional memory: the accumulated knowledge of what works, what fails, and how to troubleshoot when yields dip unexpectedly.

Building a terafab from scratch would require assembling a workforce of comparable scale and expertise. Musk has demonstrated an ability to attract top engineering talent in automotive and aerospace, but semiconductor process engineering is a specialized field with a limited talent pool. Much of that pool is currently concentrated in Taiwan, South Korea, and pockets of the United States and Europe. Recruiting at the necessary scale would likely trigger wage inflation and poaching wars, raising costs for all players.

There is also the question of tacit knowledge, the kind that does not appear in patents or technical papers. Yield optimization at advanced nodes often involves subtle adjustments to process recipes, equipment settings, and fab environmental controls. These adjustments are discovered through trial and error, and the learning curve is steep. Even with access to the same equipment as TSMC, a new entrant would face years of yield ramp before reaching parity.

Strategic Implications for the Foundry Landscape

If Musk's terafab does materialize, it would represent a new category of competitor: a vertically integrated tech conglomerate with captive demand and the financial resources to sustain losses during the ramp phase. This model is not unprecedented; Samsung operates foundries primarily to support its own memory and system chip businesses, and Intel's foundry services are partly intended to leverage capacity built for its own products. But Musk's ventures span a wider range of end markets, from electric vehicles to satellite internet to neural interfaces, each with distinct chip requirements.

For TSMC, the near-term threat is minimal. The company's pipeline is full through at least 2027, with customers already committed to multi-billion-dollar prepayments for capacity at 2-nanometer and beyond. The medium-term risk is more nuanced. If a well-funded competitor can achieve competitive yields at a trailing node, say 5-nanometer or 7-nanometer, and offer lower prices or tighter integration with system design, it could capture share in mid-tier segments. That would not dethrone TSMC at the leading edge, but it could compress margins and force the company to defend territory it has long taken for granted.

The other wildcard is geopolitics. If tensions over Taiwan escalate, customers may prioritize supply-chain resilience over pure performance, favoring fabs in the United States, Europe, or Japan even if they lag in process technology. Musk's terafab, if located in the United States, could benefit from this dynamic, especially if it receives government subsidies under the CHIPS Act or similar programs. Wei's confidence may be well-founded today, but the rules of the game are in flux.

The Long Game

Wei's "good luck" remark captures a broader truth about the semiconductor industry: it rewards patience, precision, and the willingness to invest for decades, not quarters. TSMC has spent more than thirty years building its position, refining its processes, and deepening its customer relationships. That accumulated advantage is not easily replicated, even by someone with Musk's resources and ambition.

Yet the industry has also seen upheavals before. Intel dominated logic manufacturing for decades before stumbling on 10-nanometer. GlobalFoundries exited the leading-edge race entirely. Samsung's foundry business was written off as a laggard before it won Qualcomm's flagship Snapdragon business. The lesson is that incumbency is powerful but not permanent.

For now, TSMC can afford to be sanguine. Its order book is full, its technology roadmap is on track, and its competitors are struggling with their own execution challenges. But the foundry landscape is entering a period of heightened uncertainty, driven by geopolitical fragmentation, surging capital costs, and the emergence of new players willing to challenge the old order. Whether Musk's terafab joins that list of challengers or remains a footnote in the annals of ambitious announcements will depend on factors far beyond a shareholder meeting soundbite.

The chip wars are far from over, and the next decade will test whether scale, expertise, and incumbency are enough to hold the line, or whether new models of vertical integration and risk-taking can redraw the map. Wei's confidence is earned, but in an industry where a single process misstep can cost billions, no position is ever truly secure.