From: allin
Overview of AI’s Impact
The recent rally in tech stocks has been significantly influenced by AI advancements [00:22:02]. While not the sole driver, a “radical realization” is underway regarding businesses whose services might be replaced by AI [00:22:57].
Market Trends and Valuation
The market saw a devastation in tech stocks in 2022, and the current rally is partly a “reversion to the mean” [00:14:10]. Tech stock multiples were 50-70% above their 10-year average in 2020-2021 and dropped to 30-40% below by the trough [00:14:22]. Currently, they are still trading below the 10-year average for internet and software companies [00:14:51].
Investor Sentiment and AI
The shift in investor focus this year is away from inflation and interest rates, and towards concerns about a potential hard economic landing [00:15:29]. The NASDAQ has moved 30% year-to-date [00:16:12]. However, some individual stocks, particularly AI-related ones like Nvidia and Microsoft, may have “gotten ahead of themselves” [00:17:45]. Nvidia, for example, saw its stock go from 400 after revising its Q2 guide from 11 billion [00:18:44].
There’s a psychological exhaustion from past losses and a desire to “will the market up” [00:31:31]. The belief that AI will lead to cuts in interest rates is challenged, especially with China’s likely economic stimulation [00:20:04].
AI’s Influence on Industries
The impact of AI on different industries is seen as a “distribution curve” [00:25:08].
- Immediate Impact: Demand for chips and infrastructure is the most obvious area of impact [00:23:32]. This leads to a “disproportionate bubbling” in the valuation and multiples of businesses like Nvidia [00:23:46].
- Mid-to-Long Term Impact: Services businesses like law firms and investment banks are expected to leverage AI in the future to improve margins and create new products, but this is a few years out [00:23:54]. Manufacturing industries might see changes in 10-15 years [00:24:32].
Redefining Web Architecture and Business Models
The “20 trillion dollar question” is how the open architecture of the web will be re-architected in the age of AI [00:34:48]. The “top of the funnel” in web search is now “up for grabs,” and those advertising dollars will be redistributed [00:35:27].
Google’s Position
Google is well-positioned to make a claim in this new AI landscape, but it will come at the expense of its traditional search business model [00:35:35]. The new user interface (UI) will not be about optimizing web pages or applications, but building a “conversational UI” – “the race to intimacy” [00:36:11].
Google’s Bard has shown rapid progress, now including images and menu items in search results, with the potential for paid links that could lead to higher CPMs (cost per mille) and CPCs (cost per click) [00:45:38]. Despite this, some believe Google’s competitive advantage in AI is less monopolistic than in search, as new competitors emerge [00:48:42].
Data Control and Monetization
A key question is whether users will have independent relationships with specialized AI services (e.g., travel agents, doctors, financial advisors) or a single aggregator [00:37:51]. The argument for fragmentation is user control and privacy, with users potentially becoming “servers” who can grant or rent access to their data to different service providers [00:39:06].
Case Study: Reddit’s API Dispute Reddit’s decision to charge for its API for AI model training led to a widespread mod strike, where 95% of subreddits went dark [01:04:22]. This situation echoes historical instances with Facebook and Twitter, where platform owners sought to control user experience and monetize data directly, often killing third-party apps [01:05:58].
The value of Reddit is deeply embedded in its community and the content creators, who have historically revolted to change platform rules [01:08:50]. Reddit’s management faced a “miscalculation” by not redefining revenue splits with mods before monetizing the API [01:13:22]. A potential solution could be allowing mods to monetize through subscriptions and splitting revenue with apps [01:13:50]. This trend highlights a shift where value “further erodes away from centralized apps and more towards the individual people or in this case these hubbed spokes these mods and not to the centralized organization” [01:14:44].
AI Investment Landscape
Recent AI startup funding rounds have been “insane,” with companies like Mistral AI raising 240 million valuation without having written any code [01:15:16]. These large rounds are often necessary to acquire expensive H100s and A100s for training AI models [01:16:00].
Critiques of Current AI Investment
Some argue that these large seed rounds are “stupid bets” and “financially illiterate” [01:17:03].
- Subsidizing CAPEX: Investing $100 million into a startup primarily to buy compute is seen as subsidizing capital expenditures (CAPEX), a role typically handled by banks with lower hurdle rates [01:25:12].
- Dilution: Giving up 40% of equity for CAPEX when equipment leases could be an alternative is questioned [01:38:33].
- Cost Curve Acceleration: The cost of training AI models is rapidly decreasing. For example, what cost OpenAI 5-10 million in 18 months [01:35:17]. This means early investments in model training may quickly lose value [01:35:41].
- Distraction for Founders: Large upfront funding can be a “huge distraction” for founders, potentially leading to inflated salaries and a lack of necessary “constraint” that fosters innovation [01:17:17].
- Regulatory Environment: The current regulatory climate in Washington D.C. makes it difficult for hyperscalers to acquire AI companies for over $1 billion, limiting exit opportunities for startups [01:32:55]. This creates a “two-sided problem”: high funding costs and undermined downside protection [01:34:31].
- Comparison to Past Bubbles: The current AI funding environment is compared to the internet search bubble of 1997-98 and the social networking boom, where many companies went to zero despite the overall market being right [01:19:24].
Opportunities in AI Investment
Despite the risks, AI is considered a “major platform disruption” [01:31:25]. Opportunities are seen in:
- Application and Tooling Layers: Businesses solving real problems in the application and tooling layers, rather than just large model development [01:31:54]. These can build sustainable advantages through customer base and features [01:37:33].
- First-Mover Advantage: While compute costs decline, securing a first-mover advantage in certain AI agent positions can be highly valuable [01:42:30].
Conclusion
The AI landscape presents both immense opportunities and significant financial risks. The ability to discern between genuine value creation and financially illiterate capital allocation is crucial for investors. The shift from traditional web models to knowledge extraction and intelligent agents is a “tectonic” shift [01:54:34], but the optimal investment strategies remain an evolving challenge.