a16z Former Partner's Major Tech Report: How AI is Consuming the World?

Written by: Bu Shuqing

Source: Wall Street Journal

“AI is devouring the world, and we haven't even seen what it looks like.”

In the latest report “AI eats the world,” renowned technology analyst and former a16z partner Benedict Evans provided a judgment that is enough to shake the entire tech world: generative artificial intelligence is triggering a major platform migration in the tech industry every ten to fifteen years, and we still do not know where it will ultimately lead.

Evans pointed out that the foundation of the technology industry has been completely rewritten every decade or so, from mainframes to PCs, from the internet to smartphones, and the emergence of ChatGPT in 2022 is likely the starting point of the next “fifteen-year change.”

Global tech giants are rushing into an unprecedented investment race. Capital expenditures for Microsoft, Amazon AWS, Google, and Meta are expected to reach $400 billion by 2025—this figure exceeds the annual investment scale of approximately $300 billion in the global telecommunications industry.

“The risk of underestimating AI is far greater than the risk of over-investing,” said Sundar Pichai, CEO of Microsoft, in a quote from the report, capturing the essence of industry anxiety.

The report also cites the 1956 U.S. Congressional Automation Report and the case of the disappearance of elevator operator jobs to remind us that when technology truly takes root, it quietly becomes infrastructure and is no longer referred to as “AI.”

The Change of Fifteen Years Again: Historical Laws of Platform Migration

Evans noted in the report that the technology industry experiences a platform shift approximately every ten to fifteen years, from mainframes to personal computers, from the World Wide Web to smartphones, with each shift reshaping the entire industry landscape. Microsoft's case illustrates the brutality of this shift: the company once held nearly 100% of the operating system market share during the personal computer era, but became almost irrelevant when the focus shifted to smartphones.

Data shows that the share of Microsoft's operating system in global computer sales has sharply declined from its peak around 2010, dropping to below 20% by 2025. Similarly, Apple, which dominated the personal computer market in its early days, was also marginalized by IBM-compatible machines. Evans emphasizes that early leaders often disappear, which seems to be an iron law of platform shifts.

But three years have passed, and we still know very little about the nature of this transition. Evans cited failed ideas from the early days of the internet and mobile internet, such as America Online (AOL), Yahoo portal, Flash plugins, etc. Now it is the turn of generative AI, and the possibilities are equally dazzling: browser forms, agent forms, voice interactions, or some entirely new user interface paradigms, and no one really knows the answer.

Unprecedented investment frenzy: a $400 billion gamble

Tech giants are investing in AI infrastructure on an unprecedented scale. By 2025, capital expenditures by Microsoft, AWS, Google, and Meta are expected to reach $400 billion, compared to the global telecommunications industry's annual investment of about $300 billion.

It is worth noting that this 2025 growth plan has nearly doubled within the year.

The construction of data centers in the United States is surpassing the scale of office building construction, becoming a new driver of the investment cycle. Nvidia is facing supply bottlenecks due to its inability to keep up with demand, and its quarterly revenue has exceeded Intel's accumulated revenue over many years. TSMC is also unable or unwilling to expand capacity quickly enough to meet Nvidia's order demands.

According to Schneider Electric's industry survey, the main limiting factor for data center construction in the United States is the supply of public electricity, followed by chip acquisition and fiber access. The electricity demand in the United States is growing at about 2%, while AI could add an additional 1% to this demand, which is not a problem in China but is difficult to build quickly in the United States.

Model Convergence: The Moat Disappears, AI May Be “Commercializing”

Despite the massive investment, the gap in benchmark tests among top large language models is narrowing to single-digit percentages. Evans warns that:

If the model performance is highly convergent, it means that large models may be turning into “commodities”, and value capture will be reshuffled.

In the most general benchmark tests, the gap between the leaders has become very close, and the model leadership is changing weekly. This indicates that the models may be becoming commodities, especially for general-purpose use.

Evans pointed out that after three years of development, there has been more progress in science and engineering, but there is still a lack of clear understanding in terms of market shape. Although models are still being improved, more models, participation from Chinese manufacturers, open-source projects, and new technology acronyms have emerged, the moat is not obvious.

In his view, AI companies must find new moats in terms of computing power scale, vertical data, product experience, or distribution channels.

User Engagement Dilemma: ChatGPT's 800 million weekly users cannot mask the real lack of stickiness.

Despite ChatGPT claiming to have 800 million weekly active users, user engagement data paints a different picture. Multiple surveys show that only about 10% of American users use AI chatbots daily, with most still in the occasional trial stage.

Deloitte's survey data shows that the number of people who occasionally use AI chatbots is much larger than those who use them daily.

Evans calls it a typical “participation illusion”: AI is penetrating at an astonishing speed, but it has not yet become a daily tool for everyone.

He analyzed the reasons for this engagement dilemma: how many use cases are obviously simple adaptations? Who has a flexible work environment and consciously seeks optimization methods? For others, is there a need to package AI into tools and products? This reflects a significant gap between technological capability and practical application.

Enterprise deployment is equally slow. Reports cite surveys from multiple consulting firms showing that despite the high enthusiasm for AI among enterprises, there are still few projects that have actually entered production environments.

Deployed: 25%

Plan to deploy in the second half of 2025: about 30%

Deployment at least in 2026: about 40%

Currently, successful cases are still concentrated in the “absorption phase” of programming assistance, marketing optimization, and customer support automation, and there is still a distance to true business reconstruction.

Advertising and recommendation systems are undergoing a disruptive rewrite.

Evans believes that the area where AI will undergo the most rapid changes is in advertising and recommendation systems.

Traditional recommendations rely on “relevance”, while AI has the ability to understand “user intent” itself. This means:

The underlying mechanisms of the trillion-dollar advertising market may be rewritten.

Google and Meta have disclosed early data: AI-driven ad placements can lead to a 3%-14% increase in conversion rates. The cost of ad creative production may also be further reshaped by automated generation technology from the annual $100 billion market.

Historical lessons: When automation succeeds, it is no longer referred to as “AI”.

Evans brings the perspective back to the 1956 U.S. Congress automation report, noting that each wave of automation triggers significant social discussion, but ultimately fades quietly into the infrastructure.

The disappearance of elevator operators, the inventory revolution brought by barcodes, the internet transforming from a 'new thing' to infrastructure… all prove that:

When technology is truly implemented and accessible, people no longer refer to it as “AI”.

Evans emphasized that the future of AI is both clear and vague: we know it will reshape industries, but we do not know what the final product will look like; we know it will be ubiquitous in enterprises, but we do not know who the dominant players in the value chain will be; we know it will require immense computing power, but we do not know where the growth will stop.

In other words, AI is becoming the protagonist of a new fifteen-year cycle, but the direction of the entire play has yet to be written.

We may be standing on the fault line of the next technological earthquake.

The Future of Value Capture: From Network Effects to Capital Competition

For research-intensive and capital-intensive commoditized products, value capture becomes a key issue. If a model becomes a commodity and lacks network effects, how will model labs compete?

Evans proposed three possible paths: expanding downstream to win by scale, expanding upstream to win through network effects and products, or seeking new dimensions of competition.

Microsoft's case demonstrates a shift from competition based on network effects to competition based on capital acquisition capability. The company's capital expenditure as a percentage of sales has significantly increased from a historical low, reflecting a fundamental change in the competitive landscape.

OpenAI has adopted a “yes to everything” strategy, including infrastructure deals with Oracle, Nvidia, Intel, Broadcom, and AMD, e-commerce integration, advertising, vertical data sets, as well as a diversified layout in application platforms, social videos, web browsers, and more.