Sharp Tech with Ben Thompson

Episode: (Preview) A Call to Action for TSMC’s AI Customers, Wall Street’s Netflix Anxiety, Q&A on Tech’s Cignetti, OpenAI, Starbucks

Date: January 23, 2026

Episode Overview

In this episode, hosts Andrew Sharp and Ben Thompson offer an in-depth discussion of TSMC’s critical role in the AI supply chain, focusing on the company’s unique position, challenges around capacity planning, customer relationships, and the underlying business and cultural forces shaping its pricing and capital expenditure decisions. They assess the downstream impact on major TSMC customers like Nvidia, Apple, and the entire AI ecosystem—highlighting how decisions made today will shape innovation for years to come. The episode also hints at broader themes (Netflix, Q&A) but remains tightly focused on the semiconductor industry, risk allocation, and what tech’s top companies must do to ensure future growth.

Key Discussion Points & Insights

1. Setting the Scene: In the Room, in the Midwest (00:00–02:27)

• Andrew and Ben open the show together in Wisconsin, with Ben poking fun at Andrew’s fashion choices for the frigid weather, setting a collegial, bantering tone.
• Transition: Andrew flags the episode will start with an in-depth look at TSMC—a rare topic for their podcast.

2. Why Doesn’t TSMC Charge Insane Margins? (02:27–03:45)

• Listener Sam questions why TSMC, effectively a single source for advanced semiconductors in the AI boom, doesn’t raise its margins even further given their leverage.
• Andrew observes: "Basically anyone who's doing anything in AI is ultimately going through TSMC." (02:37)

3. TSMC’s Early History and Customer-Centric Culture (03:45–08:14)

• Ben explains TSMC's origins: customer guarantee, capacity reliability, and strict IP protections were their founding advantages—contrasting with competitors like Intel and Samsung, both of which faced trust issues from potential customers.
• "There is a fundamental conflict of interest if you're Apple making your chips at Samsung and Samsung is competing with you." (06:42, Ben)
• "You have in general this very customer-centric, customer-first mindset because that was literally the only thing they could sell." (08:18, Ben)
• TSMC’s low-cost, longevity-focused fab operations became a hallmark approach.

4. Transition Challenges: From Customer-Centric to Premium Provider (08:14–15:33)

• TSMC historically ran fabs for decades, extracting value over time. As cutting-edge nodes (7nm, 5nm, 3nm) rose in cost, this model began to break down—requiring faster value capture and higher prices up front.
• Certain nodes, such as 28nm, became natural stopping points for many customers, especially in China.
• Ben: "What TSMC… needed to do… was become more like Intel." (15:34)
• The shift from a depreciation-driven, cost-plus mentality to a premium, margin-maximizing approach was culturally challenging. Ben theorizes executive changes were linked to a failure to adequately raise prices during critical launch windows.

5. The Capacity Constraint and Its Downstream Impact (16:13–24:00)

• Post-ChatGPT (late 2022), TSMC faced a pivotal decision: how much to invest in CapEx to meet future AI demand (2025–2029). Cautious planning led to insufficient capacity today, resulting in lost revenue for major AI players.
• Ben: "What does every single CEO say on their earnings call? …If we had more capacity, we would have sold more." (24:05)

6. TSMC as the “Brake” on AI Acceleration (24:00–26:07)

• Ben describes TSMC as “the brake” (play on “accelerationism”) on the AI boom, limiting how fast the AI ecosystem can grow due to real-world manufacturing bottlenecks.
• "TSMC is just not there. …They're the brake on a bubble on sort of AI, generally." (24:43, Ben)
• Andrew: "It's fascinating when you step back… and think about how much crazier the numbers could be if TSMC could serve the capacity in the middle of all of it." (25:16)

7. Risk Allocation: CapEx, Customers, and the Need for Competition (26:07–34:43)

• TSMC's cautious approach pushes risk downstream; AI companies incur “foregone revenue” due to unmet demand.
• TSMC is "doing what's right for TSMC because they don't have any competition." (26:33, Ben)
• Ben calls for Google, Microsoft, Amazon, OpenAI, et al to seriously develop alternatives at Intel or Samsung (not simply plead for more TSMC capacity)—otherwise they'll continue to bear the brunt of risk from a conservative supply approach.
• "What you need to do is you need to go out and you need to empower and enable an actual competitor for TSMC. That's how you get more TSMC volume." (30:04, Ben)
• Ideal scenario for AI customers: foundries overbuild, resulting in cheap, abundant chips.
• Current customer hesitance ("being chickens") might lead to a future bubble burst solely because of inadequate chip supply, not weak AI demand or technological limits.
• Ben: "All these folks who claim to be big capitalists… are not taking the risk of trying to get Intel going, to get a Samsung going." (31:36)

8. The Difficulties of Planning for the Decade Ahead (33:06–34:43)

• Silicon Valley’s traditional approach—assuming abundant hardware—makes CapEx investment planning for years out a “new muscle” tech companies haven't yet developed.
• Ben: "It's almost like a new muscle for Silicon Valley… Meaningful Capex investment…" (33:06)
• If no action is taken now, risk of missed opportunities (and profits) will be realized by 2028–2029, especially if geopolitical or demand shocks occur.

Memorable Quotes & Moments

• "TSMC has been, was relatively conservative. And the net result is… there just isn't nearly enough capacity at TSMC for all the demand. …All these companies are realizing that risk right now." — Ben (20:09, 24:00)
• "The way you get cheap chips is by there being too much capacity and they’re having to sell it at very low prices." — Ben (28:19)
• "If you're an AI company, you want the foundries taking risk. …All these companies need to sack up and stop playing." — Ben (29:20)
• "[The] clear-eyed recognition would be TSMC is just not going to go that direction unless... [there’s a credible alternative]." — Andrew (29:58)
• On CapEx risk: "It’s a big — that’s what I mean, being underwater, like you can really struggle." — Andrew (23:29)
• Ben’s “Ben Theory”/“conjecture” on TSMC’s missed opportunity to reprice and its impact on leadership transitions. (14:20–16:13)

Key Timestamps for Important Segments

• 02:27 – Introduction of TSMC topic and question about pricing power
• 03:45 – TSMC’s origins: customer-first culture, capacity guarantees, IP protection
• 08:14 – Challenges of transitioning from a budget foundry to a premium leader
• 13:06 – The exponential increase in fab costs and rationale for shifting business models
• 20:09 – TSMC’s cautious CapEx in response to post-ChatGPT AI boom; resulting global bottleneck
• 24:00 – TSMC as “the brake” on AI acceleration; the limits of current manufacturing
• 26:07 – Who bears the risk from TSMC’s conservatism? Foregone revenue for AI leaders
• 29:20 – The case for actively supporting Samsung and Intel to break TSMC’s monopoly
• 33:06 – Valley’s new muscle: long-term hardware investment planning and strategic risk

Tone & Style

The episode is friendly, sharp, and lightly irreverent—Ben’s deep, insightful analysis is consistently peppered with Andrew’s clarifying questions and wry asides. The hosts maintain an open, analytical lens, admitting when comments are “conjecture” or “Ben Theory,” and display their familiarity with industry leaders and dynamics. The language is accessible yet precise, inviting listeners to both the strategic and the operational sides of the global semiconductor drama.

For listeners or readers seeking to understand why the AI revolution’s pace is limited not just by software or ideas but by manufacturing realities, this episode offers a masterclass in deep tech business analysis—complete with clear calls to action for tech’s largest companies to shape their own destinies.