Alex Sacerdote - How to Invest Through Technology Cycles — structured breakdown

SEGMENTS

3:27

The Anthropic Thesis and the Foundational Model Oligopoly

Sacerdote recounts Whale Rock's firm-wide AI deep dive after ChatGPT's November 2022 launch, the chips-and-infrastructure-first strategy, and how the foundational model layer narrowed from ~60 contenders to a three-horse oligopoly; the coding explosion (Anthropic employees spending $100/day on tokens) drove the investment at the $180B round.

14:01

Breaking Into Private Markets: Anthropic and Stripe

How a legacy public-markets investor wins private allocations: relationship-building with Anthropic's management and a 90-page diligence deck built with Claude Code, plus the origin story of Whale Rock's first private — a $100M Stripe block bought from a VC seller in April 2020, underwritten via Adyen comparables.

19:36

The S-Curve Framework

Sacerdote lays out the Whale Rock framework — S-curve, competitive advantage, underappreciated earnings power — explaining how removing barriers to adoption (iPhone price, Tesla range) triggers the demand tornado, how to size how tall an S-curve is (the AWS TAM work), and why exponential growth stops around 30-40% penetration.

25:27

Timing the Inflection: Intuition and Pattern Recognition

On knowing when to buy: Andy Grove's point that you can't trust data at strategic inflection points, visual/anecdotal clues (Gartner IT Symposium packed rooms for Splunk, VMware, AWS), Peter Lynch's 'white out the chart,' why it's okay to miss the first 100%, and the radio-vs-dishwasher models for adoption pace.

30:59

Digital Moats and the Model Layer's Competitive Advantages

Why digital competitive advantages (network effects, industry standard, scale, platform, critical IP, brand) rival offline ones, the 2013 'Coke and no Pepsi' AWS pitch, and the case that Anthropic and OpenAI now show critical IP, enterprise brand, escape-velocity scale, and possibly recursive improvement.

38:11

AI's Impact on Enterprise Software

Whale Rock went from 40-50% software to net short: incumbents' AI products weren't good, AI spend crowds out budgets, pricing power and seats are at risk, and AI-native challengers loom — though a half-baked counter-thesis says agents operating inside tools like Slack and CRM could entrench them; Sacerdote proposes a new 'rule of 40' (% of sales that are AI + market share).

44:41

The Decommoditization of Hardware

AI workloads growing 10x/year are pushing every data-center component to physical limits, reversing 40 years of commoditization: Celestica (Google TPU servers, 50-60% cloud Ethernet switch share), Corning fiber, 40-layer PCBs and Elite Materials, power supplies — with DRAM, NAND, and PCB markets already ~30% short.

55:48

Risks to the AI Trade

What worries Sacerdote: broad public and governmental negativity toward AI (Maine banning data centers), the risk that model improvement hits a wall — letting open source catch up and triggering a race to the bottom for the model stocks — and the possibility of a major player faltering and stranding compute.

58:09

The Application Layer

Why Whale Rock has done little in AI applications: the app layer always comes later, the boundary between foundation model and application is unclear and dangerous, though great application companies will be built — he's watching Bret Taylor's Sierra as the bellwether.

1:00:29

The WhaleRock Learning Machine, Fund Products, and His Father

AI is helping research as 'a great reporter' but can't pick stocks; the Philip Fisher scuttlebutt process and 2,500-3,000 management meetings a year remain the engine; Whale Rock's product evolution culminates in the new MegaCap Tech Fund (best 12-13 of the top 30 global market caps); he closes with a tribute to his father, his Goldman-veteran chairman and mentor.

NOTABLE QUOTES

“The world doesn't think exponentially, and they're so focused on the next year, the next quarter.”

Alex Sacerdote · 19:36 — The one-line thesis for why the S-curve framework keeps finding the best companies at single-digit PEs.

“It's okay to miss the first 100 percent.”

Alex Sacerdote · 25:27 — Counterintuitive timing advice — when the top of the S-curve is half a trillion away, being late beats being wrong.

“And we talk about S-curve, we call this an L-curve, just straight up.”

Alex Sacerdote · 8:37 — His coinage for AI's adoption curve — the central image of the episode's bull case.

“Not only are you creating tremendous unit growth, but we call it the decommoditization of the hardware industry.”

Alex Sacerdote · 44:41 — Names the hardware renaissance thesis that drives Whale Rock's Celestica, Corning, and supply-chain positions.

“It's like, why does a casino teach people how to play blackjack? It's really hard to do.”

Alex Sacerdote · 54:09 — His answer to why he shares his 'secret' framework publicly — the edge is in execution, not the idea.

“So the AI can be a great reporter. It can't pick into the future.”

Alex Sacerdote · 1:00:29 — A leading AI bull's crisp limit on what AI does to his own profession.

PREDICTIONS ON THE RECORD

Dated and falsifiable — full record at /api/episode/<id>.

Alex Sacerdote · next four years · 8:37

Real AI usage will climb the S-curve from ~10 basis points of knowledge workers to 1-3%, then 5%, then 15% over the next four years.

“But you're going to go from 10 bips to 1 to 2 or 3% to 5% to 15% in the next four years.”

Marc Andreessen · next four years · 8:37

There will not be enough compute over the next four years.

“Mark Andreessen said in the next four years, one thing he's sure of is there's not going to be enough compute.”

Alex Sacerdote · unspecified · 44:41

Scale-up networking (connecting every GPU within the rack) will eventually move from copper to fiber, which will 2-3x Corning's opportunity.

“Eventually, that'll be done over fiber. When that happens, that 2-3x is Corning's opportunity.”

WHO CLAIMED WHAT

Every position taken in the episode, attributed to its holder, with conviction (0–1) and the verbatim receipt.

Alex Sacerdote — 29 claims

assertionconviction 0.83:27

The foundational model layer has resolved into a three-horse race and somewhat of an oligopoly (Anthropic, OpenAI, Google), much as three clouds came to underpin the SaaS world.

“So it really started to look like a three-horse race and somewhat of an oligopoly.”

assertionconviction 0.73:27

Open-source models lack the compute to leapfrog the leading edge — they can come close to the frontier but then falter.

“The open source guys, they don't have as much compute, so they can come close to the leading edge, but they can't leapfrog it and then they kind of falter.”

opinionconviction 0.753:27

Coding alone is a half-trillion-dollar market for AI, based on ~20 million coders worldwide spending $20,000-30,000 a year on tokens.

“And if you think about how many coders there are in the world, $20 million, you've got a half a trillion dollar market just from coding alone.”

assertionconviction 0.88:37

Contrary to the commodity thesis, AI models have tremendous differentiation — different training methods and skills constitute critical IP and a great competitive advantage.

“But in the AI models, everyone thought it would be pure commodity, but there's tremendous differentiation within.”

assertionconviction 0.758:37

Anthropic is building an entire ecosystem of products (SDK, orchestration, the 'harness') around its API, repeating the lock-in playbook AWS ran from 2013.

“They're building a whole monopoly or a whole ecosystem of products around the API.”

assertionconviction 0.758:37

The AI infrastructure S-curve is only ~10% penetrated and the enterprise application AI market is less than 1% penetrated.

“And so the enterprise AI or enterprise application AI market is less than 1% penetrated.”

opinionconviction 0.919:36

AI is the biggest S-curve Whale Rock has seen, bigger than Internet 1.0, mobile, cloud, and e-commerce.

“We've been lucky that we've had Internet 1.0, mobile, cloud, e-commerce, and now AI, which we can confidently say is the biggest.”

opinionconviction 0.8519:36

If you understand the S-curve, the moats, and the modeling, you genuinely can predict earnings two to four years out — which most of the market refuses to believe.

“But if you follow and understand the S-curve, you know the motes and you know how to model, you really can predict these great things.”

assertionconviction 0.819:36

Technology adoption inflects vertically only when the barriers to adoption are removed — that removal triggers the tornado of demand.

“And then you rocket when those barriers are removed. That's the tornado of demand that everybody in the world knows they need this right away.”

assertionconviction 0.719:36

The electric vehicle S-curve, expected to reach 40-50% of cars, unusually stalled at 10-15% penetration — S-curves are dynamic and must be re-underwritten.

“With the electric vehicle S-curve, you have to pay attention to because we thought probably maybe 40 to 50 percent of the cars will go electric, but it did hit a big wall at 10 or 15 percent.”

assertionconviction 0.7519:36

Exponential growth — and the big stock gains — generally end around 30-40% penetration, when the sell side catches up and big beats stop.

“And generally, when something gets to sort of 30, 40 percent penetrated, then you stop having exponential growth, which means the cell side catches up and there's no longer big beats.”

opinionconviction 0.825:27

It's okay to be late to an S-curve — missing the first one to three years (even the first 100%) is fine when the top of the curve is half a trillion dollars away.

“It's okay to miss the first one, two, three years in a lot of cases, because if the top of the S curve is half a trillion, the growth can go on for a long time.”

assertionconviction 0.7528:41

AI adoption will be faster than cloud or SaaS because there is no integration barrier — you just open the browser and it's there, producing a 'backwards L curve.'

“But what's amazing about AI is you just, at least with consumers or even business, you just open up the browser and it's there.”

opinionconviction 0.834:18

Anthropic has reached escape velocity — critical IP in code, the leading enterprise brand among CIOs, 10x sales growth, and proven fundraising ability.

“Certainly, with Anthropic, with their 10X sales growth and their fundraising ability, it looks like they've reached escape velocity.”

opinionconviction 0.5534:18

Anthropic feeding its leading code capability back into its own model is recursive improvement — its innovation pace is accelerating and it may reach a lift-off stage.

“If you look at the pace of their innovation, it's accelerating. Maybe they can have this lift-off stage.”

assertionconviction 0.8538:11

Software incumbents' AI products were not good — they weren't moving the needle and nobody could charge for them — so Whale Rock sold almost all of its software.

“Pretty quickly, we realized their AI products were not very good. They weren't moving the needle. Nobody could charge for them. We basically sold almost all of our software.”

assertionconviction 0.7538:11

Even if AI doesn't disrupt software directly, it hurts incumbents through CIO priority lists, budget crowd-out from Anthropic token spend, lost pricing power, and seat losses from job cuts.

“So even if AI is not going to be disruptive, they're spending it on Anthropic tokens because there's faster ROI there.”

predictionconviction 0.540:11

Within one to five years, brand-new AI-native companies could go after each strong software incumbent, especially if the incumbents' data advantage gets obviated.

“But you can imagine a world where in one, two, three, four, five years, you could have a brand new AI native company going after each one of these very strong incumbents.”

opinionconviction 0.440:11

A counter-thesis (self-described as half-baked): AI agents operating inside incumbent platforms like Slack or CRM could make those platforms permanent fixtures rather than killing them.

“AI could make some of these software platforms more important because what's the first thing you do with Cloud? You plug it into Slack.”

opinionconviction 0.740:11

Software should be screened by a new 'rule of 40' for the AI era: the percent of sales that are AI plus market share in that AI category.

“For AI, we have a new rule of 40 What percent of your sales are AI?”

assertionconviction 0.8544:41

AI is driving the decommoditization of the hardware industry — workloads growing 10x a year push every server component to its physical limits, creating both unit growth and durable IP.

“Not only are you creating tremendous unit growth, but we call it the decommoditization of the hardware industry.”

assertionconviction 0.751:32

If the L-curve demand thesis is right, the DRAM, NAND, and PCB markets are already roughly 30% short of supply.

“The broader picture is the AI demand, if we're right with this L-curve, we're already short the DRAM market, the NAND market, the PCB market. We're already 30% short all these things as we are now.”

opinionconviction 0.854:09

Whale Rock's framework stays an edge despite being public because it's genuinely hard to execute — like a casino teaching blackjack — requiring decades of cycles, hardware coverage no one does, and conviction through bubble scares.

“It's like, why does a casino teach people how to play blackjack? It's really hard to do.”

opinionconviction 0.6555:48

Broad negativity about AI in the general population and parts of government — and the resulting regulation risk — is a real threat to the AI trade, though the genie is out of the bottle.

“One thing that bothers me is there's a lot of negativity in the general population about AI.”

assertionconviction 0.5555:48

If Anthropic or OpenAI hit a wall and stop improving, open-source models will catch up and it becomes a race to the bottom — bad for the model stocks, though the chip companies don't care who wins.

“In AI, if Anthropic sort of hits a wall and stops improving or OpenAI, then the open source models will catch up.”

assertionconviction 0.755:48

Jensen Huang really wants open source to take off — it was all he kept mentioning at his last GTC.

“Jensen really wants open source to take off. It's all he kept on mentioning at his last GTC.”

opinionconviction 0.758:09

The AI application layer is still unclear and a little dangerous to invest in — it's hard to tell where the foundational model ends and the application begins — though great application companies will eventually be built.

“And at the application layer, it's still kind of unclear and a little bit dangerous.”

opinionconviction 0.751:00:29

AI can be a great reporter for investment research but cannot pick stocks or supply the forward-looking wisdom — the analyst's insight on top is still the job.

“So the AI can be a great reporter. It can't pick into the future.”

opinionconviction 0.81:04:04

There is tremendous alpha in the largest-cap tech stocks because endowments and allocators are structurally underweight them and it takes 100 diversified PMs to move a mega cap.

“I think there's tremendous Alpha in the largest cap, because if you think about it, a small cap, it just takes one person to figure out it's good and move it up.”

Andrej Karpathy — 1 claim

assertionconviction 0.858:37voiced by Alex Sacerdote

AI coding flipped from writing ~20% of code (80% handwritten) to essentially all of it — he no longer writes a line of code except in English.

“And Karpathy said, last year's code tools could write 20 percent and 80 percent would be handwritten. That flipped when the latest model came out, and now he hasn't written a line of code except in English.”

Andy Grove — 1 claim

assertionconviction 0.7525:27voiced by Alex Sacerdote

At strategic inflection points you can't trust the data — they are about intuition and anecdotal evidence.

“Andy Grove says, when you have strategic inflection points, you can't trust the data.”

Jensen Huang — 1 claim

assertionconviction 0.7555:48voiced by Alex Sacerdote

If good enough is good enough, NVIDIA won't have a business — the franchise depends on people always wanting the best.

“But Jensen said this years ago when he was talking about his graphics chips, if good enough is good enough, I won't have a business.”

QUERY THIS EPISODE

agent access

# full structured record
curl 'https://aipod-two.vercel.app/api/episode/invest-like-the-best-2026-06-09-alex-sacerdote-how-to-invest-through-technology-cycles'

# or over MCP
aipod_episode { "id": "invest-like-the-best-2026-06-09-alex-sacerdote-how-to-invest-through-technology-cycles" }