“The Bubble Theory is Completely Wrong”—Anthropic Investor and Semiconductor PM Ankur Crawford on Investment Philosophy in the AI Era

Alger Portfolio Manager (PM) Ankur Crawford appeared on episode 238 of The Compound and Friends Podcast. Having started her 22-year career as a semiconductor analyst, she currently manages several funds, including Alger Capital Appreciation, Alger Focus Equity, and Alger Spectra, and actually participated in Anthropic's $380 billion round. Her view, stating, “I have never felt such a clear investment opportunity in my career,” can be read as a frank rebuttal to market participants skeptical of AI investment.

1. Investment Philosophy—Two Lenses: “Discovering Change” and “Lifecycle Change”

1-1. The Essence of Growth is “Change Not Recognized by the Market”

“Alger has been a growth investor since the 1960s. The essence is that where change is happening, there are unrecognized opportunities,” Crawford said. She noted that there are two buckets for growth investing.One is typical high-growth companies with high sales growth. The other is “lifecycle change.”

“Lifecycle change is the stage where companies that have already gone through a growth phase and reached market saturation face the question of how to find their next growth,” she explained. She cited Apple, Microsoft, Nvidia, Western Digital, and Danaher as specific examples. The idea behind these is that “if you can spot the seeds of regrowth while they are being ignored at a low valuation, it can lead to significant returns.”

The statement, “Sometimes people ask, ‘Are you buying value stocks?’ That's not it. This is unrecognized growth,” succinctly illustrates this philosophy.

2. Rebuttal to the AI Bubble Theory—“You Misjudge Because You Think Linearly”

2-1. “If you thought agents were coming in 2028, they were already here”

“What I envisioned at the time of the report I wrote three years ago, ‘AI and the Falling Cost of Creation,’ is happening now. At the time, I thought agents would become mainstream in 2028, but the timeline has moved up significantly. That is because the growth of capability is exponential,” Crawford stated.

“Analysts and the market are trained to think linearly. If there are two points, they connect them with a straight line. However, we are now in an era of exponential growth. It is very difficult to grasp what exponential actually means and where the endpoint is,” she pointed out, explaining the structural reason why many market participants are underestimating AI investment.

2-2. The Bubble Theory is an “Excuse for Not Understanding”

“Every time I hear the bubble theory, I always ask where they think the bubble is,” Crawford said. “Many of the people who call it a bubble are PMs or CIOs from healthcare or consumer goods backgrounds, and it is difficult for them to understand what is happening now based on their own background. So, they say, ‘It must be a bubble.’ Things they don't understand look like a bubble,” is her frank analysis.

As a decisive difference from the bubble theory, she cited a comparison with the 2000 dot-com bubble. “The problem in 2000 was that huge amounts were invested in fiber optics, but no one had the content or technology to use that infrastructure. They had a dream, but the actual technology didn't follow. It's different now. The technology is already there. All that is needed is compute (computing resources).” Using an F1 car analogy, she expressed, “This time, it's not a dream; the car is finished. All that is needed is fuel.”

3. The Investment Theme of Compute Shortage—“Everyone is Short”

3-1. “Compute supply and demand will remain tight until at least 2026–2028”

“Right now, everyone is short on compute,” Crawford stated. “Because the moment you can ask someone to do something productive, people keep using it. The CTO of Uber said they had used up their 2026 compute budget by April. That is the reality,” she said, citing a specific example that shows the explosive growth in demand.

“When you try to replicate intelligence, it is an exponential problem. Each connection in a neural network increases exponentially in the next layer. The demand is larger than you can imagine, but it is very real,” she also stated. Based on announced capital expenditure (Capex), she indicated a forecast that the supply-demand balance will not be met until at least 2026–2028.

3-2. DRAM and HDD as “Overlooked Bottlenecks”

Amidst the compute shortage, she, as a former semiconductor analyst, is paying particular attention to the DRAM (memory) market.“Two or three years ago, I thought AI would exert a huge gravitational pull on DRAM. However, because the memory market was sluggish, I thought I was wrong. I thought I was wrong even though the market was giving the wrong signal,” she said, sharing her experience.

“Micron's stock price was $85 as of August 2025. Now it is $450. DRAM rose about 100% year-on-year. For a normal commodity, this is abnormal,” she said, with the figures retrospectively showing the correctness of her view. Regarding the HDD (hard disk) market, she stated, “All the data generated by AI agents will be saved. It is a market with only two companies, Seagate and Western Digital, but this industry, which no one wanted to enter five years ago, has now become an important investment target.”

3-3. Understanding the Investment Unit of "Tokens"

Crawford's point that "not all tokens are created equal" is crucial. Her prediction that "there is no reason to buy a token for checking the weather at the same price as a token used for drug discovery. The market will eventually change pricing based on use cases. A drug discovery token will have a completely different value than an educational token" suggests the long-term revenue structure of AI infrastructure investment.

4. The End of SaaS—"40% Operating Margins are Unsustainable"

4-1. Rebuttal to Jensen Huang's Remarks

"Jensen said that AI agents would complement SaaS companies, but honestly, I don't think that's true," Crawford stated. The implication that "there are reasons Jensen cannot say otherwise" comes from an understanding of his position, which forces him to avoid direct references to competitors.

"Salesforce has raised prices every year for the past 15 years, and no one said no. I don't think that will continue. The terminal value of SaaS companies will change. My sense is that 20% is an appropriate level, down from 40% operating margins," she said. She added, "As of yesterday, Salesforce was valued at 10x EV/FCF (about 14.5x including SBC). The market is already distinguishing between 'irreplaceable' assets, like the niche industry-specific software suites held by Constellation Software, and 'replaceable' ones, like Atlassian. I think the market is reading this correctly."

4-2. Platform-Type Companies as Exceptions

It is not that all SaaS is pessimistic. "Security companies like the platform-based CrowdStrike are more likely to be immune. Palantir is functioning as a guide for 'crossing the chasm.' AppLovin is falling as a software stock, but this is because it is in the IGV (software ETF); essentially, it is a mobile game advertising platform and is benefiting from AI utilization," she noted.

5. Actual Investments—Anthropic, Nebius, QXO, AppLovin

5-1. Private Investment in Anthropic

"We participated in the $380 billion valuation round. Under the 40 Act (public mutual fund) framework, we can invest 15% of fund assets in private companies. In the past, we have made early investments in Palantir and Chime," Crawford said.

"It is difficult for growth investors to participate only in public markets for companies that reach a $500 billion valuation before an IPO. Private investment is a differentiator for active management," a perspective that is a practical response to the current market structure where AI unicorns remain private for long periods while maintaining high valuations.

5-2. Nebius (NBIS)—"The Next AI-Native Hyperscaler"

"Nebius was originally Yandex (the Russian Google). After the invasion of Ukraine, the CEO and 1,500 engineers left Russia, moved to the Netherlands with $2 billion, and restarted as an AI data center company. We got a call from the NYSE out of the blue saying, 'We are listing on Monday,' and we spent from Thursday through the weekend preparing," she said. She also shared an anecdote: "I chased them down in Davos and finally got a meeting." Her assessment that "I believe this can become the next AI-native hyperscaler" is noteworthy.

5-3. QXO—"Growth Opportunity via Brad Jacobs-Style Industrial Consolidation"

"QXO is a company where Brad Jacobs (founder of URI and XPO) is trying to consolidate the building materials industry. He is acquiring Beacon Roofing and aiming to consolidate and technologize the entire industry. He has set a goal to increase EBITDA from the current approximately $1 billion to $5 billion by 2029-2030. He has done this many times before," she assessed.

6. Apple, Amazon, and the Outlook for the Screenless Era

"Among the Mag 7, Apple has a Halo—but it will lose it if it doesn't realize an agentic Siri immediately. I don't understand why Siri cannot book a flight," Crawford said, expressing both expectations and concerns for Apple.

"Amazon might be buyable right now at its lowest valuation ever, with a GAAP P/E in the low teens. It is 10 points lower than Costco and Walmart. Is this a reasonable valuation for an excellent grocery store?" This question highlights her bullish view on Amazon, which has an AI growth story through its AWS and Anthropic partnership.

Recommended Articles

Next Big Wave—Growth Stocks, Seeds of Ideas, and Deep Dives into Trends