AI Model Commoditization: The Next Platform Shift or a New Cycle?

In August 2011, Marc Andreessen penned the influential essay "Why Software Is Eating the World," which detailed how software was transforming industries and reshaping the global economy. More recently, Benedict Evans, a former partner at a16z, presented on the generative AI platform shift three years after ChatGPT’s launch. His central argument is that while we know this matters, the specifics remain unclear. In this article, I explore why the evidence points more toward commoditization of AI models rather than durable competitive advantages at the model layer.

Why it Matters

The technology industry has historically reorganized around new platforms every 10-15 years: mainframes in the 1960s-70s, PCs in the 1980s, the web in the 1990s, and smartphones in the 2000s-2010s. Generative AI appears to be the next major platform shift, but it could also break this cycle entirely. The potential outcomes range from "just more software" to a single unified intelligence that handles everything. Understanding where we are on this spectrum is crucial for investors and businesses.

Key Risks

The hyperscalers-Microsoft, Google, Amazon, and Meta-are pouring unprecedented amounts of capital into AI infrastructure. In 2025 alone, these companies will invest approximately $400 billion in AI capex, surpassing global telecommunications capex. Microsoft is now spending over 30% of its revenue on capex, double what Verizon spends. Despite this massive investment, the results have produced models that are more capable but less defensible.

When ChatGPT launched in November 2022, OpenAI had a significant quality advantage. However, today, dozens of models cluster around similar performance levels. DeepSeek demonstrated that with $500 million, any entity can build a frontier AI model. The pricing for large language models (LLMs) has also collapsed. Since the launch of GPT-3, OpenAI’s API pricing has dropped by 97%, and each year brings an order of magnitude decline in inference costs.

The Opportunity

While $500 million is still a substantial barrier, it is one that only a few dozen entities globally can meet with acceptable risk. However, the rapid commoditization of AI models suggests that value will flow up the AI value chain to applications rather than being captured by model providers. This trend aligns with historical patterns where new platforms initially create high barriers but eventually lead to widespread adoption and innovation.

For instance, GPT-4’s performance on complex reasoning tasks, Claude’s extended context windows of up to 200,000 tokens, and Gemini’s multimodal capabilities highlight the rapid advancements in AI. These developments are driving a shift towards more specialized and application-focused use cases, rather than creating monopolistic control over the model layer.

Conclusion

The evidence suggests that while generative AI represents a significant platform shift, it is likely to follow a path of commoditization similar to previous technology cycles. This scenario implies that the real value will be in building innovative applications on top of these increasingly accessible models. For investors and businesses, this means focusing on how to leverage AI to solve specific problems and create unique value propositions rather than trying to compete at the model layer.