Khan Capital | January 2023
Key Takeaways
- ChatGPT reached 100 million users in two months, the fastest consumer technology adoption in history, demonstrating AI capabilities that are immediately recognisable and applicable to knowledge work across industries.
- The AI investment opportunity spans three layers: infrastructure (Nvidia, TSMC, memory), platforms (Microsoft, Google, Amazon, Meta), and applications (vertical AI solutions), each with different risk-return profiles and competitive dynamics.
- Microsoft’s $10 billion OpenAI investment, combined with Azure and Office 365 distribution, creates the most compelling platform-layer position, while Nvidia’s 80%+ market share in AI accelerators makes it the highest-conviction infrastructure play.
- The timeline from demonstration to monetisation is measured in years: the sustainable business model for AI services has not been established, and each AI query costs approximately 10x more in compute than a traditional search.
- The parallel to the early internet (1994-1995) is instructive: the technology is genuinely transformative, the adoption curve will be steeper than consensus expects, and the hype cycle (enthusiasm, inflated expectations, correction, genuine buildout) is as inevitable as the eventual value creation.
On 30 November 2022, OpenAI released ChatGPT, a large language model chatbot, to the public. It reached 100 million monthly active users within two months, making it the fastest-growing consumer application in history. By January 2023, the financial markets have begun to process the implications: artificial intelligence, which has been a recurring technology theme for decades without producing a definitive market catalyst, appears to have found its “iPhone moment” – the consumer product that demonstrates a transformative technology’s capabilities to a mass audience and unlocks a wave of investment, adoption, and speculation.
The market response is still in its early stages, but the direction is unmistakable. Microsoft, which invested $10 billion in OpenAI and is integrating its technology into Azure, Office, and Bing, has rallied. Alphabet, whose dominance of internet search is suddenly threatened by conversational AI, initially declined before recovering as it rushed to showcase its own AI capabilities (Bard/Gemini). Nvidia, whose GPUs are the essential hardware for training and running large language models, is beginning a re-rating that investors do not yet fully appreciate. The AI trade has begun.
Why This Time Is Different (Perhaps)
AI has been a technology investment theme for years, featuring prominently in sellside reports, conference presentations, and venture capital pitches. Previous cycles of AI enthusiasm (IBM Watson, autonomous vehicles, early natural language processing) produced incremental advances but not the transformative, mass-market applications that would justify the scale of investment the technology required. Investors had grown sceptical of AI’s ability to deliver on its promise.
ChatGPT changed the calculus in three important ways. First, it demonstrated a capability that was immediately recognisable to non-technical users: the ability to generate coherent, contextually appropriate text that could draft emails, write code, summarise documents, answer questions, and engage in conversation. The product did not require explanation; it explained itself through use. Second, the speed of adoption (100 million users in two months, compared to nine months for TikTok and two years for Instagram) demonstrated genuine consumer demand rather than technology-push hype. Third, the business implications were immediately obvious: any knowledge-worker task that involves generating, analysing, or summarising text, code, or data is potentially automatable using this technology.
The corporate response has been equally rapid. Microsoft’s $10 billion investment in OpenAI positions it to integrate AI capabilities across its product suite. Google declared a “code red” and accelerated the development and deployment of its own large language models. Amazon, Meta, and every major technology company began racing to develop or acquire AI capabilities. The technology industry, which had been retrenching throughout 2022 (mass layoffs, hiring freezes, capex cuts), pivoted overnight to an AI-first strategy that would define its investment priorities for years.
The Investment Map: Where the Value Accrues
The AI investment opportunity can be decomposed into three layers, each with different risk-return profiles and competitive dynamics.
The infrastructure layer (“picks and shovels”). Training and running large language models requires enormous amounts of compute, primarily delivered by specialised GPUs. Nvidia is the dominant supplier, with its A100 and H100 chips commanding approximately 80%+ market share in data centre AI accelerators. TSMC manufactures the chips. ASML provides the lithography equipment. SK Hynix and Samsung supply the high-bandwidth memory. This layer offers the most immediate and direct exposure to AI spending, with revenue growth that is already accelerating.
The platform layer (“hyperscalers”). Microsoft, Google, Amazon, and Meta are the companies with the resources (capital, data, talent, distribution) to build and deploy frontier AI models at scale. They will compete to provide AI-as-a-service through their cloud platforms, embedding AI capabilities into enterprise software and consumer products. This layer offers large-scale exposure to AI’s commercialisation but with more diversified revenue bases that dilute the AI-specific exposure.
The application layer (“software and services”). The companies that build vertical AI applications, using the foundation models as a platform to deliver specific business solutions (legal, healthcare, finance, education, customer service), represent the longest-duration but highest-uncertainty opportunity. This layer is where the eventual value creation from AI will be greatest, but the competitive landscape is nascent, and identifying the winners at this stage is highly speculative.
What the Market Is Misunderstanding
The timeline from demonstration to monetisation is measured in years, not months. ChatGPT is a proof of concept, not a business model. The path from a consumer product that generates excitement to an enterprise technology that generates revenue at scale requires product development, sales cycles, integration work, and customer adoption that takes years. The market’s tendency to front-load multi-year value creation into near-term price action creates the risk of a “hype-to-reality” correction before the genuine earnings impact materialises.
The cost structure is prohibitive at current scale. Running large language models is extraordinarily expensive. Each ChatGPT query costs an estimated 10x more than a traditional Google search in compute resources. At current cost levels, the economics of AI-powered services depend on either dramatic efficiency improvements (which are being actively researched) or a willingness by providers to subsidise AI services as a customer acquisition strategy (which Microsoft and Google are currently doing). The sustainable business model for AI services has not yet been established.
Nvidia is the most obvious beneficiary and the most mispriced. The market is beginning to appreciate Nvidia’s position as the essential infrastructure provider for AI, but the magnitude of the demand acceleration has not yet been fully priced. Data centre GPU procurement is in the early innings of a multi-year buildout that will make Nvidia’s revenue trajectory unlike anything seen in semiconductor history. The stock trades at what appears to be an elevated multiple on current earnings, but the multiple on forward earnings, once the AI revenue ramp is fully appreciated, may prove surprisingly reasonable.
The disruption risk to incumbents is real. Google’s search monopoly, which has generated over $100 billion in annual advertising revenue, is potentially threatened by conversational AI interfaces that answer questions directly rather than presenting a list of links. While Google has formidable AI capabilities of its own, the defensive position of responding to a competitive threat is fundamentally different from the offensive position of disrupting an existing market. The competitive dynamics of the AI era may produce winners and losers among the current mega-caps, not just add value to all of them simultaneously.
Implications for Investors
AI is a multi-year investment theme that deserves strategic, not tactical, allocation. The comparison to the early internet (1994-1995) is apt: the technology is genuinely transformative, the adoption curve will be steeper than consensus expects, and the investment opportunity extends across the entire technology stack. Investors who missed the internet’s early years because the stocks seemed “expensive” on trailing metrics paid a far higher price for their caution.
The infrastructure layer is the highest-conviction opportunity in the near term. Nvidia, TSMC, and the semiconductor supply chain benefit regardless of which platform or application ultimately wins the AI race. The picks-and-shovels approach, while not without risk, reduces the uncertainty associated with predicting application-layer winners at this early stage.
Microsoft has the most compelling platform-layer position. The OpenAI partnership, Azure’s enterprise cloud position, and the Office 365 distribution channel create a unique combination of AI capability and commercial distribution that no other company can match. The stock’s re-rating is in its early stages.
Expect a hype cycle. The pattern from previous technology revolutions suggests an initial surge of enthusiasm, a period of inflated expectations, a correction when monetisation timelines disappoint, and then a sustained buildout as the genuine applications emerge. Investors should own the theme with awareness that the correction phase is as inevitable as the eventual value creation.
Conclusion
The release of ChatGPT marks the beginning of the most significant technology investment theme since the advent of cloud computing, and possibly since the commercial internet itself. The speed of adoption, the breadth of potential applications, and the scale of corporate investment being redirected toward AI capabilities suggest a multi-year cycle that will reshape the technology sector and, eventually, every sector of the economy. The AI trade is no longer a speculative bet on a distant future; it is a present reality that is already being priced into the companies best positioned to build, deploy, and profit from it. The trade has begun. It is still early.
Related Reading
ChatGPT sparked a new investment theme after the brutal selloff covered in Growth Stock Carnage: The 2022 Tech Wreck. The AI trade intensified through Nvidia’s AI Supercycle and the Magnificent Seven, before facing scrutiny in The AI Bubble Debate.
Leave a Reply