Episode Overview

Podcast: Asianometry
Host: Jon Y
Episode Title: Silicon Valley Thinks TSMC is Braking the AI Boom
Date: February 1, 2026

In this episode, Jon Y responds to a growing narrative in Silicon Valley: that TSMC’s strategic conservatism and perceived supply chain bottlenecks are throttling the progress of the AI boom. Using both industry analysis and personal insight, Jon unpacks whether TSMC should truly bear the blame for compute shortages, explores the larger realities of the semiconductor hardware ecosystem, discusses capital expenditure dynamics, and examines the profound gap between software and hardware perspectives in the tech industry.

Key Discussion Points & Insights

1. Silicon Valley’s Frustration with TSMC (00:02–05:30)

• Jon references Ben Thompson’s recent article in Stratechery, which criticizes TSMC for conservative investment and suggests this has cost American hyperscalers (cloud giants) “hundreds of billions in revenue.”
• Jon discloses his own professional relationship with Ben but stresses he’s hearing similar opinions widely in Silicon Valley—that TSMC is acting as a "brake" on AI’s growth.

"As if they're the reason why we don't have AGI yet. Because they didn't and still don't believe... a company that spent 41 billion on capital expenditure in 2025 with another 53 to 56 billion in 2026 planned is sitting on its hands doing nothing."
— Jon (02:00)

• He acknowledges that TSMC’s near-monopoly (90% AI chip market share) is unhealthy, agreeing competition is needed and is coming (Samsung is making advances).

2. The Realities of Semiconductor Supply Chains (05:31–18:00)

• Jon argues that the heart of the problem is the complex, inflexible nature of semiconductor supply chains, not solely TSMC’s investment strategy.
• Introduces the “Beer Game” (renamed the “Boba Game” for fun), a supply chain simulation developed by Jay Forrester in the 1950s, which demonstrates how small demand spikes can cause massive fluctuations up the chain.

"At each step, we have time delays for order processing, shipping or production… As the game progresses, the cards start showing unannounced spikes in demand… delays and shortages or overreactions and overproduction and everyone thinking someone else other than themselves messed up."
— Jon (07:45)

• The same amplifying “bullwhip effect” causes both shortages and massive gluts in the chip industry.
• Illustrates supply chain intricacies: Deposition tools are not interchangeable. Power, water, land, and labor are shortages in themselves.
• AI systems are highly memory-dependent (specialized DRAM, NAND), and memory is just as consolidated as foundry fabrication.

3. History Repeats: COVID, Capital Expenditure, Gluts & Pivots (18:01–31:30)

• Recalls the 2020-22 chip shortages: carmakers, trailing-edge chips, and TSMC's struggles with double booking & demand forecasting.
• TSMC responded by ramping up capital expenditure ($36 billion in 2022), but was burned when demand collapsed, leading to underutilized fabs and major financial drains.

"AFAB's break-even utilization rates are about 60 to 70%. So those N7 Taichung fabs were taking financial losses, potentially on the order of hundreds of millions, maybe even billions."
— Jon (27:00)

• Contrasts TSMC’s resilience and pivot to AI with Intel’s aggressive expansion and near-collapse, resulting in layoffs, CEO ouster, and eventual government intervention.

4. TSMC’s Investment Decisions & the "BOBA Game" Signal Delay (31:31–38:00)

• Ben Thompson’s critique: TSMC did not ramp up capex quickly enough post-ChatGPT, hence shortages now.
• Jon counters: It takes time for AI demand signals to flow through ("BOBA game")—TSMC had legitimate uncertainty in 2023–24 about AI’s business case.

"In the April 2023 earnings call... CC Wei says he noticed ChatGPT's growth, but repeats multiple times that he has no idea what AI's impact on TSMC will be."
— Jon (35:00)

• Technical bottlenecks further delayed scaling: in 2024, TSMC faced COAS capacity and chip packaging yield issues, notably affecting Nvidia; deployment was also hampered by server rack problems (overheating, leaks) on the customer side.
• Demand forecast uncertainty lingered until late 2024–2025, when the economic case for mass AI compute finally firmed up.

5. The Real Bottleneck: Power, Not Just Chips (38:01–43:00)

• References Ben’s claim: “It is chips not power behind the shortage of compute capacity.”
• Jon disagrees, highlighting power as the true supply constraint for coming AI data centers—the requirements have outpaced earlier projections.
• Evidence: TSMC Arizona CFO’s deleted post and high-profile examples (Elon Musk’s data centers using truck-mounted turbines; long lead times for new gas turbines).

"The power shortages are real and way more serious than the silicon ones. Elon is bringing in truck mounted gas turbines to its data centers, and new gas turbines aren't available until 2029."
— Jon (41:30)

6. The Hardware/Software Divide (43:00–End)

• Jon observes a widening disconnect between Silicon Valley’s software/AI community and those who actually build the hardware.
• Notes that many software-side professionals have no understanding of what’s involved in hardware supply chains, fabrication, or engineering.

"It's been a good 30 years since Silicon Valley was actually about making silicon...even the smartest in their domain...know virtually nothing. It is a hard silicon line. I feel like both sides know so little about the other."
— Jon (44:50)

• Final message: Patience is needed; both hardware and infrastructure are rapidly scaling, but supply chain physics can’t be willed away.

"I'm sorry that Claude code is a little slow for you right now, but the chips are coming. People are torturing themselves to make them, put them into racks and start up the data centers. Let's exercise a little patience."
— Jon (45:32)

Notable Quotes & Memorable Moments

• “Shortages are a fact of life in semiconductors. As are horrific gluts.” (04:30)
• “These are not fungible like AWS compute units.” (09:15)
• “TSMC tried to discern between double booked orders and real demand, which is not an uncommon experience for them. Customers lie about their own demand all the time, or at least we can say that they are eternally optimistic.” (22:23)
• “Having slack advanced node capacity means taking massive depreciation losses. Having such pricey non performing 7 nanometer fabs could have been crippling.” (29:53)
• “The optimal time for TSMC and the rest of the semiconductor industry to really scale CAPEX was 2025, whereupon the BOBA game kicked into effect. Some things just take time.” (38:45)
• “At least the semiconductor people are trying Semianalysis said in a report that the various legacy gas turbine makers will not greatly expand their factory footprints. They seem a bit grumpy that the turbine boys aren't AGI pilled.” (42:00)

Timeline of Important Segments

• 00:02–05:30 – Silicon Valley narratives on TSMC, Ben Thompson’s critique
• 05:31–10:00 – The “Boba Game” and semiconductor supply chain complexity
• 10:01–18:00 – Supply chain amplification, bullwhip effect, and real-world impacts
• 18:01–31:30 – Post-pandemic chip market swings, gluts following shortages, underutilization
• 31:31–38:00 – Timing of TSMC’s capex decisions, limitations on forecasting & technical bottlenecks
• 38:01–43:00 – Power infrastructure vs. chip bottlenecks; industry anecdotes
• 43:00–End – The hardware/software cultural divide in Silicon Valley and closing remarks

Takeaway

This episode deftly rebuts the simplistic view that TSMC’s decisions are singularly responsible for AI compute shortages. Instead, Jon Y demonstrates how hard, long, and interconnected the supply chains are, highlights the risks of overbuilding during uncertain cycles, and underscores that even rapid, massive investments can be stymied by both technical and infrastructure bottlenecks (especially power). He concludes with a broader reflection on how little many software professionals grasp about these hardware realities—urging mutual patience and understanding as the industry catches up.