Ben Thompson on Big Tech, China, and the AI Boom Running Out of Money - [Invest Like the Best, EP.487] – Invest Like the Best with Patrick O'Shaughnessy
Ben Thompson on Big Tech, China, and the AI Boom Running Out of Money - [Invest Like the Best, EP.487]
Bullish and bearish opinions expressed in this episode, paired with supporting transcript quotes. The quote confirms what was said—not whether the opinion is correct.
Bullish
$AMZN— Amazon has the most compelling setup among big tech due to their model of building for themselves first (AWS, logistics, Graviton, Trainium) then selling externally, giving them scale to iterate products. Their core retail business is also more insulated from AI disruption than other digital companies.
$META— Meta has the most interesting setup among frontier AI companies due to their massive advertising business providing cash flow, founder energy driving frontier investment, and unique zero-cost content advantage. Their ad business will see billions in incremental gains from AI improving ad matching and creative generation, with validation through their massive liquid marketplace.
$AAPL— Apple is well-positioned despite sitting out AI race due to ecosystem control giving them supplier leverage, potential for on-device AI using customer electricity (no inference costs), and core phone business being insulated by physical goods moat. Their deterministic product culture may actually be an advantage by focusing on what they do best.
$INTC— Intel will be saved by acute compute shortages forcing big tech companies to diversify away from TSMC despite the pain of working with Intel. TSMC's conservative capacity expansion created the scarcity that will economically incentivize customers to bring Intel up to speed, solving Intel's fundamental customer problem.
$POWER— Power infrastructure is the lasting benefit from AI investment bubble, similar to fiber from dot-com era. The US has brought more power online than expected, and energy abundance would be transformative. Power is the constraint that will outlast GPU cycles and data centers.
$OPENAI— OpenAI and Anthropic have the biggest upside among frontier AI companies due to religious-level belief in their mission, existential need to make business work, and being on the actual frontier. The power of conviction and necessity drives outsized outcomes.
$BRK.B— Berkshire Hathaway's railroad (BNSF) demonstrates the power of absolute profits over percentage profits, generating more free cash flow in one year than See's Candies did in its lifetime. This capital is now flowing into Google, symbolizing the shift from high-margin low-volume to lower-margin high-volume businesses.
Bearish
$MSFT— Microsoft is pursuing an IBM-like middleware strategy that may work tactically but faces existential threat from AI potentially eliminating need for their systems of record and user interface products. Their E7 pricing model breaks their bundled value proposition and forces customers to question individual product value.
$NVDA— NVIDIA faces margin pressure hidden through circular financing and risk-taking with neo-clouds. Hyperscalers (Google, Amazon) building their own chips and selling them externally as commodities are NVIDIA's ultimate threat, with lower cost of capital and scale advantages. Power coming online faster than expected gives competitors more time to catch up.
$TSMC— TSMC's conservative capacity expansion in 2023-2025 offloaded risk onto big tech customers who are now experiencing massive foregone revenue. This conservatism ironically creates the acute shortages that will force customers to diversify to Intel and Samsung, similar to how memory makers created targets on their backs.
$DATACENTER— Massive timing mismatch in AI infrastructure investment creates risk of oversupply. Current compute shortage is from insufficient 2023-2024 investment, while today's massive spending won't manifest until 2028-2029. Commodity market dynamics with fixed costs will drive behavior regardless of paper losses, similar to shipping and memory boom-bust cycles.
$XAI— xAI's case is weakest among frontier AI companies because their highly differentiated data center and space advantage means they may not need their own model. They can rent capacity to others (like Anthropic) without wasting billions on model development.