"When software begins to write software, innovation becomes exponential." - [Ankur Crawford (quoted by Barry Ritholtz)] [00:39:15]
"In this world of AI, I think there's a lot of people who don't really understand what is happening under the covers, and that's dangerous." - [Ankur Crawford] [00:07:21]
Disclaimer: Orignal content owned by or sourced from third parties. It does not represent the views of 'Nuggets' platform or it's team. AI is used extensively across this platform including for summaries. Accuracy is not guaranteed, there can be mistakes. Any info or content on this platform is not a financial, legal, or investment advice. Do your own research. Refer for complete disclosures:- Terms of Use · Full Disclaimer
"The fundamental thing we look for is not necessarily growth, it is change. And the change begets the growth." - [Ankur Crawford] [00:28:51]
"If we put $3 trillion into the ground next year... I would say that that is over capacity, but we can't do it because there is almost a natural limiter to the growth of this market." - [Ankur Crawford] [00:54:45]
"The Kimi model is a little bit like the bird on the giraffe... they need the giraffe, which is our LLMs, in order to survive." - [Ankur Crawford] [00:58:12]
"A lot of the AI doomers who don't want the data center built in their backyard are ignoring the fact that there are many different aspects of AI that will be good for humanity... [like helping to solve] climate change." - [Ankur Crawford] [01:05:04]
Speakers & Credentials
Barry Ritholtz (Host): Host of the "Masters in Business" podcast on Bloomberg Radio, renowned financial commentator, and wealth manager.
Dr. Ankur Crawford (Guest): Co-Head of Portfolio Management for Large Cap Strategies at Alger and sole manager of the Alger Concentrated Equity ETF (CNEQ). She holds a BS in Mechanical Engineering and Material Science from UC Berkeley, and an MS and PhD in Material Science from Stanford University. Former Intel engineer and Intel PhD Fellowship awardee holding multiple patents.
1. Executive Summary
The current AI infrastructure build-out is not a purely speculative bubble analogous to the dot-com era; rather, it is capped by physical constraints (power, human capital, chip supply) that naturally protract the investment cycle and prevent immediate overcapacity.
Technological innovation has hit an inflection point of exponential velocity because software is now writing software, functionally driving the cost of software creation toward zero and shifting enormous capital value from IT services directly into hardware and networking layers.
Growth investing fundamentally requires identifying "change" rather than just screening for fast-growing revenues; this encompasses both high-unit volume compounders and mature companies undergoing dramatic "life cycle" pivots (e.g., legacy hardware components finding second lives in AI data centers).
Understanding the deep technical mechanics of semiconductors and hardware provides a vital alpha-generating edge in volatile markets, separating high-conviction structural investors from "loose holders" who panic during complex technological transitions.
Open-source AI models, specifically those emerging from China (like Kimi or DeepSeek), operate in a symbiotic, "distillation" relationship with western frontier models; they are highly efficient fast-followers but structurally rely on the massive capital expenditures of hyperscalers to generate the training data they refine.
The ultimate, long-term ROI of artificial intelligence will manifest not just in enterprise SaaS efficiency, but in democratizing complex societal bottlenecks, specifically driving the cost of personalized healthcare and universal education exponentially downward over the next decade.
[00:04:31] Cultural Nuances in Executive Management (Nebius, TSMC)
[00:06:54] The Alpha of Technical Engineering Backgrounds in Tech Investing
[00:15:58] The Transition from Intel to Alger & Learning to Invest
[00:26:18] Portfolio Construction: The Alger Concentrated Equity ETF (CNEQ)
[00:28:33] The Philosophy of Growth Investing: Change as the Catalyst
[00:39:15] The Exponential Era: Software Writing Software
[00:48:04] Deconstructing the AI Bubble Thesis & Supply Chain Limiters
[00:58:12] Open Source Models, Distillation, and the US/China AI Dynamic
[01:04:02] Overlooked Net-Positive Impacts of AI on Humanity
[01:06:03] Mentors, Reading Recommendations, and Advice for Graduates
3. Detailed Thematic Summary
The Intersection of Hard Engineering and Financial Markets
Crawford leverages her deep material science background (having spent 3.5 years in a Stanford basement managing a single deposition tool) to navigate tech volatility [00:11:10]. She notes that investors who lack a conceptual understanding of deposition, etching, or how a chip is fabricated often become "loose holders" easily shaken out of positions during market turbulence [00:07:21].
Cultural norms deeply infect management guidance. Crawford cites TSMC's historical refusal to raise leading-edge chip prices despite possessing a virtual monopoly, attributing this to founder Morris Chang's philosophy of pure customer service—a stance Western analysts struggled to model until the business reality finally forced price hikes [00:05:56].
Similarly, Crawford identified an opportunity in Nebius by understanding that its Russian CEO's cultural disposition was to highlight negatives and express extreme humility, which American investors incorrectly interpreted as fundamental business weakness [00:04:31].
Forecasting technological walls provides a multi-year edge. In 2012/2013, Crawford authored a presentation on the implications of the end of Moore’s Law, sharing it with portfolio companies to test her thesis eight years ahead of the market's realization of the physics constraints [00:08:37].
Redefining Growth Investing: The "Change" Framework
At Alger, "growth" is not defined as static top-line expansion, but as structural change. Crawford breaks this into two pillars: High Unit Volume Growth (traditional share-takers) and Life Cycle Change (mature businesses hitting a second S-curve) [00:28:51].
The Life Cycle Change thesis is heavily applicable today. Crawford points to legacy Hard Disk Drive (HDD) manufacturers. Previously viewed as terminal value traps trading at single-digit multiples, the explosion of AI data storage needs resulted in severe shortages, granting HDDs immense pricing power and driving earnings up 3x to 5x [00:32:49].
Not all mature companies can execute a pivot. Crawford rejected a fintech turnaround pitched around a new management team because the core market growth was permanently capped at 4-5%; conversely, Microsoft under Satya Nadella succeeded because the pivot to Azure Cloud structurally unlocked a return to mid-teens revenue growth [00:34:35].
The Alger Concentrated Equity ETF (CNEQ) applies this philosophy across a tight portfolio of 20 to 30 companies. Sizing is strictly dictated by risk/reward; currently, Nvidia holds a dominant 13.5% weighting, while nascent turnarounds like Figure Technologies are held at smaller weights pending traction validation [00:26:34].
The Economics of AI: Capex, ROI, and Supply Constraints
Crawford asserts the AI capex cycle is fundamentally misunderstood by bears who fixate on immediate ROI. While $650 billion is currently being spent by hyperscalers, Neocloud providers report having four times more demand for compute than they have capacity to serve [00:54:10].
A true speculative overbuild is currently impossible due to real-world physics and logistics. If hyperscalers could inject $3 trillion into data centers next year, the market would over-supply. However, severe bottlenecks in power generation, chip fabrication, and even blue-collar trades (plumbers/electricians) act as a "natural limiter," structurally elongating the investment duration and preventing a boom-bust bubble [00:54:45].
Enterprise integration is messy. Companies engaged in "token maxing"—incentivizing employees to burn through LLM tokens without tracking business output—blew through entire annual AI budgets in a single quarter, resulting in layoffs (like those at Uber) disguised as AI-driven efficiency gains [00:50:03].
The economic value transfer is massive. As software begins to write software, the cost to create software approaches zero. In an IT market that grew from $5.5 trillion to $6 trillion (where 50% was previously software and services), Crawford predicts a massive value migration directly into the hardware and networking sectors [00:43:41].
The Geopolitics of Open-Source Models and Societal Impact
Crawford dismisses the idea that efficient Chinese open-source models (like Kimi or DeepSeek) negate the need for massive US hyperscaler capex. She uses a "Giraffe and Bird" model: the open-source models (birds) are highly effective but survive by "distilling" the training data generated by the massive frontier models (giraffes) [00:58:12].
If frontier models choose to keep the 'N' and 'N-1' models entirely closed—releasing only the 'N-2' models to the public—they can effectively throttle the distillation pipeline that feeds open-source fast-followers [00:58:45].
Comparing AI to the dot-com bubble is structurally flawed. In 2000, visionary ideas (like Pets.com) failed because the required infrastructure (dial-up modems) was insufficient. Today, ubiquitous intelligence already possesses the network (broadband); it is solely waiting on silicon delivery [01:01:56].
The ultimate tail-end valuation of AI lies in deflationary societal goods. Crawford notes DNA sequencing costs fell from $1,000,000 to $100 per sample; applying AI to this data will eventually bend the cost curve of personalized medicine to the point of enabling viable universal healthcare globally [00:46:42]. Ritholtz further emphasizes AI's ability to unearth unexcavated pharmaceutical research, comparing it to finding the next Sildenafil or GLP-1 from existing molecules [00:44:43].
The Reference Vault
4. Data & Figures
Data Point
Value
Context
Timestamp
CNEQ Portfolio Size
20 to 30 Stocks
The highly concentrated actively managed ETF run by Crawford.
The Two Pillars of Growth: High Unit Volume vs. Life Cycle Change [00:28:51]
Crawford outlines Alger's overarching framework that redefines "growth" simply as "change." The first pillar is standard: High Unit Volume Growth, characterizing dominant companies aggressively taking market share and expanding top-line revenue. The second, more contrarian pillar is "Life Cycle Change." This occurs when a mature business that has saturated its market (tracking flat GDP growth) makes a dramatic strategic pivot—via M&A, regulatory shifts, or technological adoption (e.g., Microsoft pivoting to Azure). This framework allows a growth portfolio to capture value in companies mistakenly priced as terminal-value traps (like legacy hard drive manufacturers) that suddenly discover a second growth S-curve.
The Deflationary Nature of Zero-Cost Software Creation [00:43:06]
When software begins to autonomously write, decode, and innovate software, the traditional economic moat of human developer hours evaporates. Crawford’s framework suggests that as the marginal cost to create software approaches zero, legacy SaaS and IT service incumbents will face severe operating margin compression. Consequently, the value within the $6 trillion IT economy physically migrates away from the software layer and consolidates into the hardware and networking infrastructure layers that power the exponential code generation.
The Natural Limiter of Infrastructure Bubbles [00:54:45]
In response to fears of a $650 billion AI infrastructure bubble, Crawford posits the concept of a "natural limiter." Unlike software, which can scale infinitely and crash instantly, the AI build-out is anchored in the physical realm. The inability to source sufficient power grids, secure leading-edge silicon, or even hire enough human plumbers and electricians structurally caps the velocity of capital deployment. This friction is ironically a blessing for investors; it forces a protracted, durable duration cycle rather than a massive, one-time oversupply pop.
The Giraffe and the Bird (AI Distillation) [00:58:12]
To explain the rapid rise of highly efficient, open-source Chinese AI models, Crawford uses a biological symbiosis framework. Massive western frontier LLMs (the Giraffes) do the heavy lifting of raw computation and data ingestion. Open-source models (the Birds) ride on their backs, utilizing a technique called "distillation" to train their smaller models on the high-quality outputs generated by the frontier models. This means that while open-source models are hyper-efficient, they are fundamentally parasitic—they cannot replace the need for the massive Capex spending of the hyperscalers, because without the Giraffe, the Bird starves.
6. Anecdotes
The Job Interview Argument over NAND vs. HDD [00:15:58]
Context: Highlighting Alger's culture of non-consensus debate and Crawford's entry into finance.
Crawford arrived at her Alger interview soaked from the rain. She struck up a conversation with CEO Dan Chung about skiing, which pivoted into a fierce debate regarding semiconductor tech. Chung argued that all Hard Disk Drives were going to zero, entirely replaced by NAND flash memory. Crawford vehemently took the other side. As she walked out, despite having zero finance experience, Chung hired her on the spot, explicitly stating he needed analysts who were unafraid to challenge his viewpoints. (Ironically, Chung was eventually proven right two decades later).
TSMC's Cultural Resistance to Pricing Power [00:05:56]
Context: Explaining how cultural nuances blind investors to financial realities.
For years, Crawford urged Taiwan Semiconductor (TSMC) management to raise their prices, pointing out they were quickly becoming the absolute single supplier of leading-edge silicon globally. TSMC consistently rebuffed her, citing founder Morris Chang's deep-seated philosophy that their sole purpose was to serve the customer, not extort them. Crawford had to underwrite the stock knowing their philosophy was artificially depressing margins, until the sheer scale of the business finally forced the pricing power into the numbers.
The Folly of "Token Maxing" and Enterprise AI Adoption [00:50:03]
Context: Demonstrating the sloppy, mismanaged integration of AI at the enterprise level that currently distorts ROI calculations.
Crawford highlights a recent corporate trend called "Token Maxing," where management created leaderboards rewarding software engineers simply for consuming the highest volume of AI compute tokens, regardless of actual business output. This resulted in companies burning through their entire annual AI budget within a single quarter without generating any measurable ROI. Subsequently, these same companies instituted massive layoffs, inaccurately blaming "AI efficiency" to mask poor management incentive structures.
7. References & Recommendations
Books & Literature
AI and the Declining Cost to Create: A research paper authored by Crawford's team at Alger three years ago detailing how AI drives software creation costs to zero and redistributes IT value [00:42:52].
Think Again by Adam Grant: Recommended by Crawford for its organizational psychology insights into the humility required to constantly question one's own investment thesis and remain intellectually nimble [01:07:09].
How to Make a Few More Billions by Brad Jacobs: Mentioned specifically for its opening chapters focusing on meditation, centering, and the non-transactional human elements of building massive businesses [01:08:53].
Principles by Ray Dalio: Brought up by Barry Ritholtz in the context of turning deep professional failures and mistakes into systemic frameworks for future success [01:23:40(Note: Timestamp approximate based on conversation flow near end of the episode)].
Companies & Entities
Alger: The asset management firm where Crawford operates, known for its focus on dynamic change and non-traditional hiring [00:15:58].
Nebius: A company with a Russian CEO used as an example of how cultural humility can be misinterpreted by Wall Street as fundamental weakness [00:04:31].
Taiwan Semiconductor (TSMC): The leading-edge foundry whose early reluctance to flex pricing power highlighted the impact of corporate philosophy on margins [00:05:56].
Intel & AMD: The legacy chipmakers discussed during Crawford's 2004 hiring interview, noting Intel's peak and AMD's structural positioning [00:16:47].
Figure Technologies: A smaller weighting in the CNEQ portfolio awaiting execution validation [00:28:00].
Microsoft: Cited as the pinnacle example of a "Life Cycle Change" company pivoting from mature stagnation to hyper-growth via Azure Cloud under Satya Nadella [00:35:56].
United Healthcare: Mentioned as a legacy incumbent that must figure out how to utilize AI to bend the massive cost of healthcare to survive [00:44:22].
Uber & Block (Square/Jack Dorsey): Used as examples of companies executing significant workforce reductions and blaming AI productivity, when it may actually mask overhiring [00:52:02].
OpenAI & Anthropic: The frontier model builders experiencing unprecedented, historically unmatched revenue velocity [01:00:37].
Pets.com / Chewy / Amazon: Used to contextualize the Dot-Com bubble—noting that early ideas were fundamentally correct but lacked the physical infrastructure (broadband) to execute, unlike AI today [01:01:56].
Medical / Healthcare Concepts
Sildenafil (Viagra) & GLP-1s: Referenced by Ritholtz as historical precedents of drugs finding massive, unexpected off-label uses—a discovery process he argues AI will rapidly accelerate by parsing unexcavated pharmacological data [00:45:12].
Kimi (by Moonshot AI): Discussed as a prominent Chinese open-source LLM that utilizes distillation techniques to efficiently mirror Western frontier models [00:58:12].
People
Dan Chung: CEO of Alger, a former lawyer turned investor, cited as Crawford's primary mentor and an incredibly non-traditional thinker [01:06:11].
Fred Alger: The founder of Alger, referenced as the originator of the firm's foundational "change" investing framework [00:29:35].
Satya Nadella: Microsoft CEO who orchestrated the pivot to cloud computing [00:35:56].
Morris Chang: Founder of TSMC, whose deep philosophy of pure customer service artificially depressed early pricing power [00:06:26].
Media & Podcasts
Macrovoices: A macroeconomic podcast Crawford streams regularly while running [01:10:11].
The Knowledge Project (Shane Parrish): Praised for its diversity of thought ranging from education reform to wellbeing [01:10:20].
The Circuit: A podcast focused explicitly on semiconductors and chips [01:10:52].
Aug 22, 2026
The Next China Shock Is Here | 21 Aug 2026 | The Ezra Klein Show
"China shock 1.0 academic literature shows is that even though these weren't the industries of the future, they were still employing a meaningful number of Americans... and it becomes a generalized downturn in those communities that was cl…
Global IT Spend Expansion
$5.5 Trillion to $6 Trillion
The macro total addressable market of IT spending shifting due to AI.