Ben Thompson’s 2025 Stratechery year in review marks a personal and professional shift, as he notes that after starting the publication in 2013 in the United States and spending most of the intervening years in Taiwan, he has now moved back, making Stratechery once again a U.S.-based publication in its 13th year. He frames 2025 as a year dominated by artificial intelligence and by growing anxiety over the United States’ manufacturing competitiveness, asking whether artificial intelligence can both “save America” and be matched by the country’s capacity to build the required physical infrastructure. Over the year, Stratechery published 26 free Articles, 109 subscriber Updates, and 39 Interviews, with the review offering a curated guide to the most-read pieces and key themes.
The five most-viewed articles highlight the breadth of Thompson’s focus within artificial intelligence and geopolitics. “DeepSeek FAQ” examines how DeepSeek has reset expectations for artificial intelligence and competition with China, while “Google, Nvidia, and OpenAI” argues that OpenAI and Nvidia are both under threat from Google and that OpenAI must develop an advertising model to beat Google as an Aggregator. “The Agentic Web and Original Sin” explores Microsoft’s proposals for an Open Agentic Web and stresses that digital payments will be essential for a new content marketplace for artificial intelligence. “U.S. Intel” critiques the idea of the U.S. government taking an equity stake in Intel as a terrible but possibly least-bad path to making Intel Foundry viable. “The Benefits of Bubbles” contends that an artificial intelligence bubble may be justified if it results in enduring physical infrastructure and coordinated innovation.
Beyond individual companies, Thompson groups many articles under broader themes. In “Analyzing artificial intelligence and its impact on society,” he highlights pieces like “AI’s Uneven Arrival,” which uses o1/o3 to argue that progress toward artificial general intelligence that can complete tasks may outpace enterprise adoption, as suggested by the history of digital advertising, and “Deep Research and Knowledge Value,” which portrays Deep Research as an artificial general intelligence product that makes public information easy to find and therefore increases the value of secret knowledge. Other entries in this section include a survey of the Big Five tech companies’ artificial intelligence positions, a framework for assessing winners and losers by tech philosophy and business potential, and an argument that content is becoming the organizing principle for virtual communities rather than geographic ones.
Thompson then turns to “Big Tech,” emphasizing how the largest technology companies are investing heavily in artificial intelligence and face the greatest risk if they misstep. He writes about Meta’s “blowout” earnings coinciding with Mark Zuckerberg’s artificial intelligence pivot and revisits Meta in a separate update on Sora, “AI Bicycles,” and the threat of creativity-unlocking tools to Meta’s business. He underscores YouTube’s centrality to Google’s artificial intelligence prospects, profiles OpenAI’s strategy to become the “Windows of artificial intelligence” by controlling both hardware suppliers and software developers, and explores how robotaxis could shift the balance between suburbs and cities in ways that might hurt urban centers and Uber. He also addresses Netflix’s role in Hollywood’s endgame, suggesting the company is betting it can enhance the value of intellectual property while keeping YouTube at bay.
Apple receives its own section as the big tech company struggling most conspicuously with artificial intelligence. Thompson describes Apple’s artificial intelligence effort as delayed and possibly overextended, arguing in “Apple AI’s Platform Pivot Potential” that Apple should focus on enabling developers to build artificial intelligence applications rather than trying to do everything itself. In “Apple and the Ghosts of Companies Past,” he warns that Apple’s long-term outlook is unusually uncertain and that the company needs to change while it still can. “Apple Retreats” interprets Apple’s WWDC as a retreat and potential strategic reset that nonetheless made for a strong presentation, while “Paradigm Shifts and the Winner’s Curse” and “iPhones 17 and the Sugar Water Trap” express concern that Apple, like Amazon, may struggle with paradigm shifts and that its flagship iPhone launches risk feeling increasingly peripheral to what is transforming the wider world.
The “American Challenges” section ties together artificial intelligence leadership and industrial vulnerability. Thompson argues that America is at the forefront of artificial intelligence but lags in manufacturing, forcing policymakers to weigh artificial intelligence’s promise against structural weaknesses. In “AI Promise and Chip Precariousness,” he writes that the artificial intelligence industry is more exciting than ever, but the chip situation is very precarious and requires drastic action. “American Disruption” reframes Donald Trump’s tariffs through a disruption lens to explain U.S. manufacturing problems and proposes a better strategy that uses demand instead of suppressing it. “Resiliency and Scale” examines how lower transportation and communications costs increase resiliency in theory but hurt it in practice, asserting that true resiliency requires accepting less efficiency.
Thompson also recaps a dense year of interviews and updates that expand on these themes. He lists conversations with public company leaders such as the CEOs of ServiceNow, Uber, Snowflake, Google Cloud, Meta, SAP, Nvidia, Cloudflare, YouTube, Booking, Asana, Unity, Atlassian, and Rivian, along with startup founders including OpenAI’s Sam Altman, Anduril’s Brian Schimpf, and several others building products in areas from security to developer tools. He notes analyst interviews on subjects like YouTube, semiconductors, artificial intelligence unknowns, streaming, CHIPS policy, Apple in China, artificial intelligence infrastructure, rare earths, and e-commerce.
The year’s favorite Stratechery Updates, as highlighted by Thompson, range across content moderation, encryption, Netflix earnings, Alexa’s history, Nvidia GTC and ASICs, Google’s product strategy, talent wars, artificial intelligence money, TSMC earnings, Figma’s artificial intelligence potential, KPop-influenced content, SpaceX’s terrestrial spectrum ambitions, speech norms on YouTube, and the costs of Meta’s Reality Labs. He closes by thanking subscribers for making his work possible, extends holiday greetings, and expresses optimism about heading into 2026 after a year defined by artificial intelligence, strategic shifts in big tech, and pressing questions about American industrial strength.
