It’s a four-day week, but not a quiet one. With technology earnings still ahead, companies are using the gap to frame expectations, sharpen narratives and quietly reset the table.

Last week provided the overture. Taiwan Semiconductor’s latest earnings showed some important technological developments that could reframe the debate over power demand in AI, which we discuss below.

Check out the website, password free! https://epistrophy.beehiiv.com 

As always, I’m focused on three things:
1) Technology-driven change;
2) the latest in innovation and startup trends, and;
3) stock fraud.

Companies Discussed

In This Note:

1. Companies Discussed

2. Mind The Gate

3. Tweet O’ The Week

4. Epistrophy In The News

5. 📆 of Epistrophy Events

6. Availability This Week

Mind The Gate

Taiwan Semi’s 2 nm and Gate-All-Around Announcement Meets The AI Moment

Man has long wondered: “How many angels can dance on the head of a pin.” Taiwan Semiconductor (TSM: NYSE) is in a position to know, one-million times over. 

Buried amidst the excitement of a blowout fourth quarter and a rousing ~$54 billion capital expenditure announcement, was news of serious progress towards the latest holy grail in tiny wafer size - two nanometers, which is about one-millionth the size of a pinhead. 

The importance of Taiwan Semiconductor’s entry into high-volume 2-nanometer manufacturing lies in what the node represents rather than the number itself – in this case not just peak performance, but sustained efficiency. The most valuable chips are those that can run continuously within realistic power, cooling and cost constraints. Two nanometers was engineered for that environment.

AI workloads have turned electricity into the scarcest resource in computing. Training and inference no longer fail because transistors switch too slowly. They fail because power delivery, heat dissipation and operating expenses hit their limits. Process technology that reduces wasted energy has become more valuable than raw speed. Taiwan Semiconductor built its next node around that premise.

You’ve seen the surge in the supply side of the equation: surging share prices for power companies, particularly nuclear science projects like Centrus Energy (LEU:NYSE),  Oklo (OKLO:NYSE) and Talen Energy (TLN:NYSE). Two-nanometer can be seen as the opposite reaction, a reduction of demand.  

The move from 3 nanometers to 2 nanometers introduces a fundamental change in transistor architecture. Taiwan Semiconductor’s 2-nanometer process is its first to rely on gate-all-around nanosheet transistors rather than FinFETs. The geometry alters how the channel is controlled, tightening electrostatics, suppressing leakage and allowing designers to fine-tune the balance between performance and power. That control is central to modern AI systems, where marginal efficiency gains compound across massive deployments. 

Gate-all-around does not simply extend Moore’s Law. It reshapes the cost curve of computation. Higher transistor density paired with lower power consumption increases usable performance per watt, which in turn lowers system-level costs for data centers. No other foundry is delivering this combination at scale.

Samsung Electronics (005930: KRX) has laid out a plausible path to 2-nanometer production, though proof at sustained volume still lies ahead. Intel (INTC: NASDAQ)’s 18-angstrom announcement last week deserves closer attention. The company paired gate-all-around transistors with a revamped power-delivery architecture and, more importantly, tied those technical claims to concrete manufacturing milestones. If Intel can translate internal progress into consistent external output, the competitive landscape at the leading edge begins to look less settled. In advanced logic, credibility ultimately accumulates through shipped wafers, repeatable yields, integrated packaging and customers willing to commit real products—not roadmaps—to the process.

Meanwhile Taiwan Semiconductor is already operating at volume. TSM’s credibility in the market place surely means orders will follow. 

This dynamic underpins Taiwan Semiconductor’s pricing power. The company does not need aggressive tactics to extract value. Scarcity and qualification do that work quietly. Customers building AI accelerators or flagship processors have limited alternatives if efficiency is the priority. Wafer pricing follows from that reality, even as capital intensity rises.

Capital expenditures have become the mechanism through which this control is enforced.

Taiwan Semiconductor plans to ramp capital expenditures to $52 billion to $56 billion in 2026, building on already elevated investment levels. Most of that capital is directed toward advanced logic at 3 nanometers, 2 nanometers and the A16 node that follows. 

The economics are unforgiving, and has created an impenetrable competitive moat (if only the Taiwan Straits were as daunting.) Each successive node requires more expensive tools, more process steps and longer yield ramps. The capital required to add incremental capacity at 2 nanometers exceeds that of earlier generations by a wide margin, and the slope steepens beyond that. This capital burden is reshaping competition. The industry is consolidating around companies that can fund sustained investment without compromising margins.

Taiwan Semiconductor’s financial profile allows it to absorb that burden. Gross margins are surging, above 60% and guided to stay that way. Free cash flow remains substantial even after record spending. These outcomes reflect pricing discipline and utilization control rather than cyclical luck.

The company’s global expansion fits the same logic.

Taiwan remains the core. That is where the most advanced nodes are developed, refined and scaled. Deep expertise, dense talent pools and decades of process knowledge anchor two-nanometer production in Hsinchu and Kaohsiung. Concentration at the leading edge reduces execution risk.

Outside Taiwan, the strategy shifts toward access and resilience.

In the United States, Taiwan Semiconductor is assembling a manufacturing cluster rather than a symbolic presence. Construction is underway on a third fab in Arizona, plans for a fourth are advancing and advanced packaging capacity is being added nearby. Yields and defect densities have steadily improved, approaching the standards achieved in Taiwan. That progress turns U.S. capacity into a functional extension of the company’s network rather than a political gesture.

Japan serves a complementary role. The Kumamoto fab has entered volume production with strong yields, and a second facility is under construction. These fabs emphasize specialty and mature nodes, supplying automotive and industrial customers that prioritize reliability and supply continuity. Japan strengthens the supply chain without drawing advanced logic away from Taiwan.

Europe completes the triangle. In Dresden, Taiwan Semiconductor is building a specialty fab supported by government incentives. The focus is not technological leadership but assured access. Governments are underwriting stability rather than innovation.

Taken together, the footprint reveals a coherent design. Advanced manufacturing remains concentrated where execution advantages are strongest. Specialty capacity is distributed where customers and governments demand presence.

Taiwan Semi had little to prove before the advent of this AI age. But its rapid introduction and production-success of its 3 nm wafers — launched in 2023 and already 23% of sales proved that the companies technological innovations can quickly get to market.

Semiconductor demand once oscillated between shortage and surplus. AI-driven workloads behave differently. They are rooted in software economics, infrastructure buildouts and long-term service commitments rather than consumer cycles. That distinction helps explain why utilization remains high even as non-AI markets recover unevenly.

The implications extend well beyond Taiwan Semiconductor.

For customers, two nanometers is becoming the baseline for flagship products, raising design costs and narrowing supplier options. For competitors, the capital threshold for relevance continues to rise faster than revenue. Technical credibility alone no longer closes the gap. Manufacturing endurance does.

Governments face a similar reckoning. Semiconductor sovereignty is expensive, slow and dependent on accumulated expertise. Incentives can accelerate investment, but they cannot substitute for decades of operational learning. Taiwan Semiconductor’s expansion illustrates the limits of policy-driven replication.

Looking ahead, utilization discipline will matter more than node announcements. Capacity decisions are being made years in advance, coordinated closely with customers and their customers. Once built, these fabs must remain full. The willingness to commit tens of billions of dollars reflects confidence that demand extends well beyond a single cycle.

Two nanometers will pressure margins early in the ramp, as every major node transition has before it. Competition from Intel is notable, but Taiwan Semiconductor has already enclosed the gate.

Tweet O’ The Week

Epistrophy In The News

On Schwab Network we had many laughs (watch it!) talking about Aussie finance slang and, more interestingly, Taiwan Semiconductor (TSM) emergence as arguably the "most important participant" in the AI trade.

📆 of Epistrophy Events

Availability This Week

I’ll be in San Francisco office at the Ferry Building for this four day week. Call! Come visit! Earnings are perking up and it’s a good time to drill down!

Written reports are available to clients, with video summaries on YouTube, and of course our popular summaries of the summaries on Instagram, TikTok, and YouTube Shorts.