Silicon Valley AI Inspection: Many Ships, Few Pirates

At the entrance of the Computer History Museum in Silicon Valley, there is a black flag on display. It bears a skull and crossbones with one of the eye sockets overlaid with Apple’s rainbow logo. The flag was originally flown in 1983 atop the Bandley 3 building, where Steve Jobs led the Macintosh team. At the time, Jobs had been ousted from the Lisa project and took over the relatively disregarded Macintosh team, which was physically separated from the main office. There, he reportedly proclaimed, “It’s better to be a pirate than to join the Navy.”

More than 40 years later, Apple has become one of the largest “navies” in the world, and Silicon Valley is riding the waves of a new tech revolution: artificial intelligence. From early May 2026, the author spent approximately two weeks in Silicon Valley, engaging in discussions with founders of AI startups, engineers from major tech companies, and investors. The findings paint a starkly different reality from the rebellious spirit symbolized by Jobs’ pirate flag.

The current AI industry in Silicon Valley resembles a sea crowded with massive corporate “ships,” leaving little room for startups that truly disrupt the status quo like pirates. There are plenty of crew members, but few pirates—this is the true picture of Silicon Valley’s AI landscape in the spring of 2026.

First Impressions of Silicon Valley AI

On the first day of arrival, the author attended a public event in San Francisco with around 100 attendees. Seated to the left was an engineer from ASML, while in front sat a graduate of Tsinghua University, now holding a Ph.D. and working at Roblox. Despite jet lag, the sheer density of talent in Silicon Valley was apparent.

During private sessions, the author gained deeper insights into the realities of the AI industry. The most striking revelation was the growth rate of Anthropic. The company’s Annual Recurring Revenue (ARR) skyrocketed from $9 billion at the end of 2025 to over $40 billion by May 2026. This growth was largely driven by their code-generation products, which have completely overturned the conventional wisdom that the coding market was capped at around $10 billion. Similar exponential growth has been observed with OpenAI’s Codex and Cursor, which SpaceX has acquired an option to purchase.

Unlike Chinese AI companies that are often evaluated based on narratives and future potential, Silicon Valley’s AI firms are assessed according to actual revenue. After completing its Series H funding round, Anthropic reached a valuation of $965 billion—approximately 25 times its ARR. The justification for this valuation is clear, with discussions between investors and founders focusing on practical implementation challenges.

How AI Agents Are Changing Business Models

AI startups in Silicon Valley take a different approach than Chinese companies that enhance mid-range models through frameworks in a “cultivation” methodology. Instead, they focus on leveraging top-tier models and implementing them as agents. Initially, they provided tools for vertical domain experts, but the focus has shifted towards business-to-business (B2B) customers who pay for tangible outcomes.

The transformation from “selling tools” to “using tools to directly perform tasks” has brought dramatic changes to AI startups. Some companies have reached a level where their technology can replace entire departments within client organizations. Corgi AI, which has found success in the insurance sector, serves as a prime example.

Two founders born in the 2000s dropped out of college, acquired a traditional insurance company, and dismissed all underwriting staff. They reconstructed the underwriting process entirely with AI, achieving unprecedented speed and pricing that traditional insurers couldn’t match. By early 2026, their ARR exceeded $40 million, and their Series B funding round valued the company at $1.3 billion. While it’s uncertain whether such radical restructuring will spread across industries, it stands as an extreme case of success.

Emergence of Headless Startups

Another intriguing trend is the rise of startups specializing in infrastructure for AI agents. These companies design products not for human users but as APIs that can be directly accessed by AI, adopting a “headless” model.

For instance, exa is developing a search engine exclusively for agents, serving over 5,000 companies including Cursor, Devin, and Alibaba. Unlike traditional search business models dependent on ad revenue, exa proposes a pay-per-use model similar to utilities like water and electricity, redefining the search market itself.

Organizational Restructuring: A Work in Progress

Many companies are currently in the phase of merely integrating AI into existing workflows. This process is akin to “cramming an electric motor into a steam engine.” While development efficiency has greatly improved, previously insignificant stages like requirement design and testing have now become bottlenecks, requiring adjustments across the board.

At Silicon Valley’s SaaS AI Summit, the fragmentation of the industry was starkly visible. Bold declarations like “SaaS is dead” filled the venue, creating an atmosphere reminiscent of simultaneous funerals and weddings. While some mourned the end of an era, others sought to prove they were a new breed of AI-native businesses.

Anthropic, with 3,000 employees, achieved over $40 billion in ARR, while Salesforce required a workforce of 80,000—including 30,000 sales staff—to generate $41.5 billion in revenue for fiscal 2026. Anthropic’s Claude integrates various SaaS tools to automate processes like information retrieval, report generation, and proposal drafting without demanding companies to completely restructure their existing workflows.

Conversely, a new startup called Monaco aims to replace fragmented tools with an AI-native platform that unifies the entire customer journey from acquisition to deal closure. Having raised $50 million in a Series B funding round, Monaco’s beta testing has already yielded an ARR exceeding $1 million. Both approaches are currently driving real revenue growth, suggesting they may coexist in the long term.

Differences Between U.S. and China’s B2B Markets

Silicon Valley’s AI startups primarily focus on B2B, with few large-scale business-to-consumer (B2C) companies emerging since 2015. This contrasts starkly with China’s situation, where B2B startups often face growth ceilings.

This divergence stems from differences in business structures between the two countries. High labor costs in the U.S. push companies to purchase external tools, creating highly fragmented demand and standardized modules after decades of development. U.S. firms are more willing to test new vendors, enabling startups to grow rapidly by targeting niche needs and scaling quickly.

In China, comparatively lower labor costs lead large companies to prefer building in-house teams. As a result, demand is highly customized, making it difficult to establish industry standards. Startups often end up handling custom projects without generating profits through product replication, thus limiting growth potential.

These ecosystem differences significantly impact the success patterns of AI startups. Silicon Valley startups thrive on standardized B2B demand, while Chinese startups struggle to apply the same model effectively.

The Shadow of the AI Boom

The rapid changes in the AI sector mean that products not up to par are quickly eliminated from the market. Founders must constantly adapt to the shifting tides of the industry, and faster monetization often equates to greater security. However, this rush to monetize has created certain distortions.

Many AI companies are pursuing ARR growth at all costs. Some engage in practices where multiple firms purchase services from each other to inflate their ARR, creating figures that look impressive but lack genuine business value. Signs of such a bubble are increasingly recognized by investors.

AI is already replacing human labor in standardized, repetitive roles. After AI raises individual productivity by three to five times, companies often find they only need one-third to one-fifth of their original workforce, forcing displaced workers to pivot their careers.

While the software sector faces intense competition and constant risks of disruption, the hardware sector benefits from the AI infrastructure boom. Hardware demand is surging, and entry barriers in core areas—built over decades—enable related companies to earn profits far exceeding industry averages.

True Pirates Still Dormant

Currently, annual investments in AI infrastructure amount to hundreds of billions of dollars. Yet, the total revenue from application-side AI remains under $100 billion, mostly limited to B2B cost-cutting and efficiency improvements. Transformative B2C products that significantly impact users’ lives have yet to emerge.

In 1983, Steve Jobs led a small team of “pirates” to create the Macintosh, revolutionizing the personal computing world. However, this success was predicated on the maturity of semiconductor and software infrastructure. The current AI revolution may similarly require further infrastructure development before producing game-changing products.

The pirate flag at the Silicon Valley Computer History Museum continues to quietly remind visitors of this legacy. For now, large corporate ships dominate the waves, leaving little room for pirates. However, in unassuming offices scattered across the valley, the next generation of pirates may be quietly preparing to change the world. The direction of the AI wave remains a topic that warrants close attention.