Is Alphabet Losing the AI Race?

For years, Google has been perceived as the frontrunner in AI technology. However, this perception began to shift in late 2022 when OpenAI unveiled ChatGPT, sparking debates about Google's position in the AI landscape. A prevailing concern is whether Google's foundational technology in AI has been overshadowed by the emergence of GPT models. This begs several critical questions:

1. How significant is the gap between OpenAI's and Google's foundation AI model technologies?

2. What challenges might Google face in closing this gap?

3. What does the future market landscape for foundation AI models look like?

In this article, I will provide my current views on these questions, discuss the pick-and-shelf strategy in the AI industry, and discuss some key signals to watch to adjust my understanding of Alphabet’s status in the future.

Google's Dual Role in AI

Google plays a dual role in the AI ecosystem: it is both a provider of foundation models and a developer of AI applications, such as its AI-powered search engine. The foundation model is a focus of the AI race among big tech companies. Drawing parallels with the cloud computing industry, where a few key players dominate due to the high costs of infrastructure and the scaling effect, it's plausible that only a handful of foundational model providers will eventually lead the market. Among today's contenders, OpenAI (GPT), Google (Gemini), Anthropic (Claude), and Meta (LLAMA) are promising candidates, given their resources, talent, and technological advancements.

Assessing the Technology Gap

OpenAI is undeniably at the forefront of foundational model innovation. Yet, when comparing their most advanced models—GPT-4 and Gemini 1.5—the superiority of GPT is not clear-cut.

Factors such as reasoning capability, multimodality, and cost must be considered. The industry's secrets, like model architecture and training methods, have become more standardized, with GPT-4's success attributed to its Mixture of Experts (MoE) architecture and Reinforcement

Learning from Human Feedback (RLHF) training method. Unless OpenAI introduces a radically new model architecture, Google's chances of catching up are plausible. I have seen conflicting reviews on OpenAI’s latest GPT-4o model, so will need to give it some more time for a thorough evaluation.

OpenAI's significant advantage lies in its first-mover status and the associated switching costs. The convenience of sticking with a familiar, robustly tested model like GPT-4 through its API service is considerable. However, for enterprises that have invested heavily in customizing their applications around a specific model, the incentive to switch diminishes unless a new model offers substantial improvements or unique features.

The Challenge of Catching Up

If OpenAI fails to rapidly evolve its models, other tech giants with the necessary talent, technology, and capital could close the gap. In a market with 3-5 foundational model providers, user loyalty will hinge on cost and performance rather than brand. Interestingly, as models become more sophisticated, the cost of switching may decrease, further intensifying competition.

Future Market Landscape of Foundation AI Models

The foundational model market is likely to consolidate around 3-5 key providers, mirroring the cloud computing industry. Decisions will be based on specific features, pricing, and partnerships rather than fundamental technological differences. Google, with its vast resources and existing cloud infrastructure, is well-positioned to be a leading player in this space.

The Pick-and-Shovel Strategy in AI

If determining the outright winner of the AI race proves challenging, an alternative investment strategy could be focusing on the "pick-and-shovel" companies, like Nvidia, that supply essential technology to the entire industry. While investing in Nvidia seemed like a great strategy a year or two ago, the widespread bullish sentiment around the company now might indicate an overheated stock. This reminds me of Peter Lynch’s cocktail party story, where even the dentist in a party recommends a stock to Lynch when a stock skyrockets.

Moreover, as the demand for GPUs and AI chips has become apparent, a surge of competitors has entered the market, threatening Nvidia's margins and market share. Tech giants are acutely aware that relying solely on Nvidia's dominance is not sustainable. Companies such as Amazon, Google, Meta, and Microsoft have all embarked on developing their in-house AI chips.

Sam Altman has also hinted at OpenAI's ambitions in the AI chip space. While the impact of these endeavors on the market remains to be seen, they signify a clear shift in the industry's dynamics.

Alphabet's Advantages and Challenges

Advantages:

● Ample capital

● A deep pool of AI talent

● An established cloud infrastructure business

● Diverse AI use cases across its products, including Search, Google Workspace, YouTube, and potentially Waymo

Challenges:

● General inefficiency within Google

● Lack of first-mover advantage, making catch-up efforts more challenging

Key Signals to Watch

● The release of Google's foundational models that match or surpass OpenAI's offerings would be a positive indicator. However, if Google struggles to release models that are comparable to OpenAI or other providers’ models, that will be a warning sign.

● High profile partnership could signal market confidence in Google's AI capabilities. Recent reports indicate that Apple is considering Google and OpenAI's AI technologies for its next iPhone, with a decision expected to be released at June's WWDC. Should Apple opt for OpenAI's technology, it could be seen as a cautionary signal for Google. Conversely, if Apple selects Google's AI solutions, or a combination of Google and OpenAI, it would serve as a strong endorsement of Google's AI capabilities.