All podcasts / Invest Like the Best / Summary

Alex Sacerdote - How to Invest Through Technology Cycles

2026-06-09 - 71 min - source - Read full transcript

Patrick O'Shaughnessy (host)Alex Sacerdote

Key insights

Technologies sit dormant for years before an S-curve inflects, and the inflection point is triggered by removing specific adoption barriers, not by the passage of time.

Sacerdote traces this across smartphones (out 10 years before the iPhone), the internet (20 years before Netscape), and Tesla (public 15 years before its 2019 breakout). In each case a concrete barrier - price, usability, network coverage, range - was removed, producing a 'tornado of demand' once the last blocker fell.

technology-s-curves

It is fine, even strategically sound, to be late to an S-curve if the total addressable market is large enough, because the growth runway extends well past the first movers.

Sacerdote cites Peter Lynch's advice to 'white out the chart' and focus on the future rather than the past. If a market's ceiling is in the hundreds of billions, missing the first 100% of gains still leaves years of compounding ahead - which is part of why Whale Rock felt comfortable initiating Anthropic in August 2025, well after ChatGPT's 2022 launch.

technology-s-curves

Whale Rock's edge in accessing hot private rounds comes from being a known, patient public-market buyer that founders and VCs prefer to typical flip-oriented private investors.

For Stripe, the firm built conviction for years through deep diligence on public comparable Adyen (talking to 200 Adyen customers) before meeting the Collison brothers in 2019. VCs who need to sell prefer handing shares to a firm that will hold in the public markets afterward, as Whale Rock did with both Stripe and Nubank.

private-market-investing

Anthropic's moat, in Sacerdote's framing, rests on three compounding legs: proprietary code-model quality, an enterprise brand that has become the default answer when CIOs are asked about AI, and a recursive improvement loop where its own coding tools now accelerate its own model development.

He notes Anthropic's roughly 10x sales growth and successful capital raises gave it 'escape velocity' against much better-capitalized incumbents like Google, and that using Claude Code internally to improve Claude itself is a distinct compounding advantage separate from raw scale.

ai-competitive-moats

Current AI usage is still at the earliest, 'tinkerer' stage of adoption - roughly 10 basis points of the world's knowledge workers, by Sacerdote's citation of Sundar Pichai - despite already showing up as a straight-up 'L-curve' in demand.

Anthropic has an estimated 14-15 million daily active users, a small fraction of whom are using AI agentically. Sacerdote expects the penetration rate to move from 10 basis points to the single digits and then into the teens over the next four years as enterprise adoption accelerates from its current sub-1% penetration.

technology-s-curves

Enterprise software incumbents face pressure even before any single AI-native competitor displaces them, because CIOs are reallocating budget toward AI model tokens and freezing software headcount and price increases.

Whale Rock cut its software exposure from 40-50% of the portfolio five years ago to net short entering this year after concluding most incumbents' AI features were not moving the needle or generating pricing power, while budget dollars shifted toward direct spend on frontier-model tokens.

ai-competitive-moats

A 'new rule of 40' - percent of revenue from AI plus market share within that AI category - is a better lens than legacy revenue or customer count for judging whether a software company is actually winning the AI transition.

Sacerdote contrasts this with companies like Salesforce, which has roughly $40B in total sales but only a low single-digit percent currently attributable to AI, meaning the metric that matters (AI share) is still small relative to the legacy base that dominates the headline numbers.

ai-competitive-moats

The AI buildout has 'decommoditized' hardware categories that were pure commodities for roughly 40 years, because workloads are growing about 10x annually versus the historical 25-40% and pushing components to physical limits.

Examples include high-bandwidth memory (now stacking 10 chips versus prior simple designs), printed circuit boards (jumping from 10 to 40 layers), and networking components upgrading annually instead of on a prior seven-year cycle - turning former low-margin commodity suppliers like Celestica and Elite Materials into structurally higher-margin, capacity-constrained businesses.

semiconductor-supply-chain

Physical, in-person demand signals - like packed conference ballrooms or standing-room-only briefing sessions - have historically preceded the financial data showing an S-curve inflection, and Whale Rock treats them as leading indicators.

Sacerdote cites Splunk and VMware sessions at past Gartner IT Symposiums and AWS's grand-ballroom crowds staying full across consecutive time slots as visible signs of enterprise demand exploding before it showed up in quarterly numbers.

technology-s-curves

Whale Rock's research process - thousands of annual face-to-face management meetings, Philip Fisher's scuttlebutt method, and a three-way 'tripod' conviction check - has not been meaningfully automated by AI despite two decades of the firm building institutional knowledge.

Sacerdote says AI currently helps the team get up to speed faster on new areas and write better first-draft notes, but the judgment-heavy work - meeting management teams, developing supplier and customer relationships, and writing the interpretive 'what does this mean' layer on top of AI-generated reporting - still requires humans.

investment-research-process

There is underappreciated alpha in mega-cap tech stocks because repricing a giant winner requires far more of the market to change its mind than repricing a small cap does.

Sacerdote argues it takes roughly 100 diversified generalist portfolio managers to collectively conclude a company like Google is a winner rather than a loser, versus one investor's conviction moving a small cap - meaning institutional underweighting of mega-cap tech (driven by a belief that 'there's no alpha in large cap') persists longer than it should.

private-market-investing

The infrastructure and chip layer is a structurally lower-risk way to invest in the AI buildout than picking model-layer or application-layer winners, because hardware suppliers benefit regardless of which foundation model ultimately wins.

Sacerdote notes chip and component companies 'don't care who wins' the model race and are already roughly 30% short of demand across DRAM, NAND, and PCB supply - a demand overhang he expects to persist even if a foundational-model player falters, since compute demand elsewhere would absorb the freed-up capacity.

semiconductor-supply-chain

Books referenced

Common Stocks and Uncommon Profits - Philip Fisher - Sacerdote says Whale Rock's research process is built directly on Fisher's scuttlebutt method - getting out to talk to suppliers, customers, and competitors to build conviction.
The Tao Jones Averages: A Guide to Whole-Brain Investing - Bennett W. Goodspeed - Cited as shaping Sacerdote's view that spotting an S-curve inflection early requires right-brain, visual, intuitive pattern recognition, not just data - he gives the example of seeing a kid playing an advanced video game on a phone in China as an early signal for mobile gaming.

Companies

Anthropic - Whale Rock's highest-conviction position; discussed in depth as the anchor for the firm's entire AI thesis, from the August 2025 investment at a $180B-then-toward-$350B valuation path to its enterprise moat and coding dominance.
OpenAI - Framed as the consumer-winning member of what Sacerdote sees as a three-horse foundational-model oligopoly alongside Anthropic and Google.
Google / Gemini - One of Whale Rock's largest positions; described as the third leg of the foundational-model oligopoly, complicated by its large legacy ad business attached to the AI bet.
Meta - Described as having faltered early in the foundational-model race and needing a 'total reboot' after initially not showing up strongly.
Amazon / AWS - Core historical case study for the firm's S-curve framework - Whale Rock pitched AWS publicly in 2013 as an underappreciated line item inside Amazon, calling it 'Coke with no Pepsi' at the time.
Stripe - Whale Rock's first major private investment (2017-2020), sized up initially through deep diligence on public comparable Adyen before meeting the Collison brothers directly.
Adyen - Public payments comparable (referred to in the transcript as 'Audion'/'Audien') that Whale Rock used to underwrite its Stripe position by comparing take rates and TPV.
Nubank - Cited alongside Stripe as a company Whale Rock held from private rounds into a long public-market hold.
Databricks - Named as one of the marquee private companies Whale Rock now has access to alongside Stripe, OpenAI, and Anthropic.
Nvidia - Chip-layer holding bought at roughly four times earnings in 2023, used as an example of buying underappreciated exponential earnings power early in an S-curve.
Tesla - Bought in 2019 at roughly five times earnings for the EV S-curve; also used to illustrate how removing price and range barriers (to $40,000 and 300 miles) triggered the adoption inflection.
Apple - Held as Whale Rock's largest position through the 0-50% smartphone penetration curve, then sold in 2012 once roughly half the US had a smartphone and exponential growth slowed.
Celestica - Contract manufacturer bought around eight times earnings after Whale Rock noticed it was the sole supplier of Google's TPU servers and held critical liquid-cooling and Ethernet switch expertise the market was pricing as commodity.
Corning - Fiber supplier example of a decommoditizing hardware layer; cited for supplying enough fiber for a single Microsoft data center to circle the world four and a half times.
Broadcom - Worked closely with Celestica's engineers on the open-source SONiC networking software layer underpinning cloud Ethernet switches.
Salesforce - Used as an example of how slowly AI revenue penetrates a huge legacy software base - roughly $40B in sales with only a small single-digit percent currently AI-attributable.
Sierra - Brett Taylor's AI applications company, watched closely (not invested) as a test case for whether application-layer AI companies can build a durable moat.
AppLovin - Past Whale Rock research win in ad-tech, cited as an example of the firm's deep-diligence process (analysts covered it as a private company before others caught on).
Qualcomm - Example of critical intellectual property as a moat - no phone could be made without paying Qualcomm.
ASML - Cited twice as an IP-moat example (no advanced chip without its lithography) and as a current holding levered to AI chip demand.
TSMC - Named as a chip-supply-chain holding levered to AI demand.
SK Hynix - Named (as 'Hynix') as a memory-chip holding extremely levered to AI demand.
Oracle - Cited both as a historical industry-standard moat example (relational databases) and as a current AI infrastructure player that canceled a big compute deal Meta then absorbed.
Bloomberg - Cited alongside Oracle as an example of becoming an unassailable industry standard.
Splunk - Gartner IT Symposium anecdote - standing-room-only briefing sessions were an early physical signal of enterprise demand before it showed up in the numbers.
VMware - Same Gartner conference pattern-recognition anecdote, cited from roughly 30 years ago when server virtualization first drew standing-room-only crowds.
Microsoft - Cited as building its own foundational models and as the owner of the data center example used to illustrate Corning's fiber scale.
Elite Materials - Holding that makes copper-clad laminate, a critical input for the higher-layer-count printed circuit boards AI servers require.
Whale Rock Capital Management - Sacerdote's own $17B+ technology-focused hedge fund, long-only, and hybrid manager, founded roughly 20 years ago.

Techniques and frameworks

S-curve, competitive advantage, underappreciated earnings power framework - Whale Rock's core investment framework - find the right point on a technology adoption S-curve, confirm a durable competitive advantage, and buy before the market appreciates how exponentially earnings can grow.
Scuttlebutt research method - Philip Fisher-derived approach of extensive face-to-face meetings with management teams, customers, suppliers, and competitors (Whale Rock does 2,000-3,000 meetings a year) rather than relying on models alone.
Rule of 40 for AI - Sacerdote's adapted version of the software Rule of 40: instead of growth rate plus margin, he looks at percent of revenue from AI plus market share within that AI category to judge whether a software company is actually winning the AI transition.
The tripod conviction check - Sacerdote's heuristic for high conviction - an idea he likes, that his analyst also likes, and that a respected outside investor also likes, forming three legs of validation.
Whale Rock Learning Machine - The firm's term for its compounding, 20-year institutional research process and knowledge base, run by a stable team with roughly 10 years average tenure.
Conference-floor demand signal - Using physical crowd size at industry events like the Gartner IT Symposium as a leading indicator of enterprise adoption inflections, ahead of financial data.

Summary

Alex Sacerdote, founder of the roughly $17 billion technology-focused firm Whale Rock Capital Management, walks Patrick O'Shaughnessy through the investment framework he has refined over 20 years: find the right point on a technology S-curve, confirm a durable competitive advantage, and buy before the market appreciates how exponentially earnings can compound. The conversation opens with Whale Rock's highest-conviction position, Anthropic, as the entry point into the entire AI stack, from chips to foundational models to applications.

Sacerdote traces Whale Rock's Anthropic thesis back to a "massive deep dive" the firm ran immediately after ChatGPT launched in November 2022, when the team decided to buy chips and infrastructure first because whoever won the model layer, tremendous compute demand was guaranteed. Over the following years the firm watched roughly 60 foundational-model contenders collapse into what it now calls a three-horse oligopoly - Anthropic, OpenAI, and Google - won on differentiated IP quality, enterprise brand, and fundraising-fueled scale. He describes the firm's August 2025 investment at Anthropic's roughly $180 billion valuation, underwritten heavily by seeing coding spend explode ($100/day token usage among power users implying a half-trillion-dollar coding market alone on then seven-to-nine-month-old technology), and by Whale Rock's own 90-page diligence deck built partly with Claude Code.

The heart of the episode is Sacerdote's detailed anatomy of the S-curve: technologies sit dormant for years (smartphones existed a decade before the iPhone, the internet two decades before Netscape) until specific adoption barriers - price, usability, network coverage - are removed, triggering a "tornado of demand." He argues it's fine to be late to a curve if the total market is large enough, and describes leading indicators the firm watches for catching inflections early, including intuitive, visual pattern recognition (spotting an advanced mobile video game being played by a child in China) and physical crowd size at industry events like the Gartner IT Symposium, where standing-room-only sessions for Splunk, VMware, and AWS preceded the financial data showing enterprise demand taking off.

On competitive advantage, Sacerdote runs through the moats Whale Rock looks for - network effects, industry-standard status, scale advantages achieved in years rather than decades, critical intellectual property (Qualcomm, ASML), and brand - and argues digital-world moats are often stronger than their offline analogues. He is candid that Whale Rock has sharply cut enterprise software exposure (from 40-50% of the portfolio five years ago to net short entering this year) because AI features from software incumbents "were not moving the needle," and proposes a "new rule of 40" - percent of revenue from AI plus market share in that category - as a better lens than legacy revenue for evaluating software companies through the AI transition. He is comparatively cautious on the AI application layer overall, noting it's genuinely unclear where the foundational-model layer ends and applications begin, and that durable application-layer moats "usually come a little bit later" in a platform's life, citing Brett Taylor's Sierra as a company Whale Rock is watching but hasn't invested in.

A substantial middle section covers the semiconductor and data center supply chain, which Sacerdote frames as a "decommoditization" of hardware that had been stagnant for 40 years - AI workloads growing roughly 10x annually versus the historical 25-40% are pushing memory, printed circuit boards, and networking components to physical limits, turning former commodity suppliers like Celestica, Corning, and Elite Materials into structurally higher-margin, capacity-constrained businesses. He argues this layer is a lower-risk way to play AI than picking model or application winners, since chip suppliers "don't care who wins" and the industry is already roughly 30% short of DRAM, NAND, and PCB demand.

The conversation closes on process and people: Whale Rock's approach to accessing private rounds (built on being a known, patient public buyer that VCs and founders prefer to sell to, as with Stripe and Nubank), the firm's "scuttlebutt" research method drawn from Philip Fisher's Common Stocks and Uncommon Profits, its "tripod" conviction check, and Sacerdote's argument that mega-cap tech stocks carry underappreciated alpha because it takes far more of the market to reprice a giant winner than a small cap. He closes, per the show's standard final question, describing his father's mentorship at the firm's founding and the outpouring of letters he received after his father's death in 2011.

Notable Quotes

"We have an investment framework. It's S-curve, competitive advantage, and then underappreciated earnings power." - Alex Sacerdote

"It's okay to be late. 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." - Alex Sacerdote

"For the past 40 years, nothing has changed in the data center." - Alex Sacerdote

"It takes 100 people, 100 diversified PMs to realize Google's not a loser. It's a winner." - Alex Sacerdote

"And if I could be half the person that he is, I'd be completely winning." - Alex Sacerdote