Unit 8200 is the unit of the Israel Defense Forces (IDF) responsible for signals intelligence. With an estimated 5000 active duty soldiers, it is the largest military intelligence unit in the IDF. It handles open-source intelligence, threat assessment, cryptanalysis, and cyberwarfare. In terms of technical expertise, Unit 8200 has been described by experts as on par with the U.S. National Security Agency (NSA).1 Unit 8200 is believed to have played a key role in dramatic operations ranging from the Stuxnet virus that crippled Iran’s nuclear centrifuges in the 2000s to the September 2024 attacks on Hezbollah via compromised pagers and other communications devices rigged with explosives. Alumni of Unit 8200 have also played a key role in the Israeli software and startup sectors: they have founded hundreds of cybersecurity companies, including Check Point, Palo Alto Networks, CyberArk, Wiz, and the controversial NSO Group, as well as successful consumer tech startups like Viber and Waze. Since the renewed outbreak of war in the Middle East caused by the October 2023 attacks on Israel by Hamas, Unit 8200 has been a key organization applying new artificial intelligence technologies to problems in warfare.

Read more

Rare earth elements (REEs) are seventeen chemical elements with a wide range of critical uses across industry and electronics due to their unique magnetic and conductive properties. They are key to electronics when used for batteries, magnets, and luminescence, for example in the displays of smartphones and computers or in military sensors, as well as in the motors of electric vehicles. Rare earths are also commonly used as catalysts in automobiles and oil refining, and are used industrially to produce high-quality steel and glass. The term “rare earth element” is a misnomer: rare earths are generally more geologically abundant than many commonly used commodity metals like copper, tin, and lead. While rare earth deposits are mined much like any other mineral is mined, the bottleneck on rare earths is rather the concentration and purity of natural deposits and the need to refine mined minerals with energy-intensive processes. In 2025, China had an estimated 92.1% share of global rare earth refining capacity.1

Read more

The recent growth in optimism about artificial intelligence technology has led to a boom in demand for the computing power or “compute” that these new AI algorithms require for training i.e. development and inference i.e. usage. The high compute demands for AI are met through chips like Nvidia’s “graphics processing units” (GPUs), which are capable of processing vast amounts of computations in parallel. While GPUs and their equivalents from other companies like AMD or Google can be purchased and run on a company’s own premises in what is now commonly called a “data center” or “compute cluster,” the financial costs and organizational complexities of building such data centers, maintaining them, and inevitably upgrading them with new and more powerful chips are prohibitively high even apparently for well-funded frontier AI labs like OpenAI or Anthropic. As a result, many companies and individuals choose instead to rent compute from “cloud computing” providers like Amazon Web Services (AWS) or Microsoft Azure and, alongside these large and established companies, a number of smaller and newer “neocloud” companies have arisen to meet demand specifically for remotely renting out AI compute.

Read more

With annual profits of over $80 billion, Meta Platforms (formerly Facebook) is the seventh-most profitable company in the world and the eighth-most valuable with a market capitalization of over $1.5 trillion (META).1 Meta earns these profits from just under $200 billion in annual online advertising revenues from its globally dominant social media platforms: Facebook, Instagram, and WhatsApp. As of December 2025, Meta claimed 3.58 billion daily active users across its social media platforms.2 This is roughly 45% of the entire human population. Meta founder and CEO Mark Zuckerberg has made it a goal for the company to produce frontier artificial intelligence models at the cutting edge of capabilities, competitive with AI labs like OpenAI or Anthropic. With AI capabilities premised to a large extent on access to enough computing power or “compute” from advanced semiconductors from companies like Nvidia, Meta’s financial firepower makes it a key contender to examine in the AI race.

Read more

On February 28, 2026, the militaries of the United States and Israel jointly launched airstrikes on the Islamic Republic of Iran in an apparent surprise attack that resulted in the deaths of a number of top Iranian government and military leaders, including the defense minister, chief of the general staff of the armed forces, and commander of Iran’s Islamic Revolutionary Guard Corps (IRGC). Most notably, the strikes also killed the 86-year-old Ayatollah Ali Khamenei, the Shia Muslim cleric who served as Supreme Leader of Iran’s revolutionary and theocratic system of government from 1989 until his death. The U.S. and Israel have continued extensively bombing both military and civilian sites in Iran, with shifting public justifications ranging from regime change in Iran to more modest goals revolving around degrading Iran’s military capabilities or nuclear program.

Read more

Starlink is a satellite internet service provided by Elon Musk’s SpaceX. With just under 10,000 satellites in Earth orbit as of March 2026, Starlink is not just the largest and most advanced satellite internet constellation but, by volume, the single largest satellite operator on the planet: Starlink satellites account for roughly two-thirds of all active satellites in orbit of any kind.1 With the first Starlink launch occurring only in 2019, this feat has been achieved thanks to over 300 launches of SpaceX’s Falcon 9 reusable medium-lift rocket since then.2 SpaceX has greatly decreased the cost of launching mass to orbit, to the point of constituting nearly 90% of all mass launched to orbit worldwide in 2024, and has used this capability to rapidly build out Starlink.3 Starlink has, in turn, become SpaceX’s key source of revenue. In 2025, SpaceX reportedly generated $8 billion in profit on around $15 billion of revenue, with an estimated 69% of revenue coming from Starlink.4 Starlink is poised to be one of the key sources of cashflow to fund Musk’s lofty ambitions in space expansion and artificial intelligence.

Read more

This Bismarck Brief will contribute to our exclusive upcoming AI 2026 Bismarck Strategic Report, which will feature in-depth analysis of the state of the artificial intelligence sector and technology, together with a comprehensive profile of key players. Upgrade or subscribe now to receive full access and never miss a report.

Cohere is a Canadian artificial intelligence company founded in 2019 by Aidan Gomez, Nick Frosst, and Ivan Zhang. Gomez was one of eight co-authors of the seminal 2017 paper by Google researchers introducing the transformer architecture that underpins large language models (LLMs) today, and both Gomez and Frosst come from the tradition of knowledge in AI and computer science centered around Nobel Prize winner Geoffrey Hinton at the University of Toronto, which also produced OpenAI cofounder and former chief scientist Ilya Sutskever. Cohere has also tried to attract research talent by establishing its research arm, Cohere Labs, as an independent non-profit with the freedom to collaborate across academia and industry to publish its research. This move came as the major U.S. labs increasingly became closed off and stopped publishing any meaningful academic research.

Read more

This Bismarck Brief will contribute to our exclusive upcoming AI 2026 Bismarck Strategic Report, which will feature in-depth analysis of the state of the artificial intelligence sector and technology, together with a comprehensive profile of key players. Upgrade or subscribe now to receive full access and never miss a report.

Since its founding in early 2023, the U.S. artificial intelligence company xAI has rapidly become one of the world’s leading AI labs and one of the world’s most valuable privately-owned companies. After its most recent January 2026 fundraising round, xAI is believed to be valued at more than $230 billion, making it the third-most valuable private AI lab after OpenAI and Anthropic.1 Headquartered in Palo Alto, California, xAI’s core staff as of 2025 numbered roughly 300 people.2 Founded by Tesla, SpaceX, and OpenAI founder Elon Musk, who is currently 54 years old, xAI merged with X (formerly Twitter) in March 2025, resulting in a combined corporation that is approximately 59% owned by Musk personally.3 In Musk’s own words, the mission of xAI is to “understand the true nature of the universe,” and the company plays an increasingly prominent role in Musk’s constellation of companies.4

Read more

This Bismarck Brief will contribute to our exclusive upcoming AI 2026 Bismarck Strategic Report, which will feature in-depth analysis of the state of the artificial intelligence sector and technology, together with a comprehensive profile of key players. Upgrade or subscribe now to receive full access and never miss a report.

ByteDance, best known as the creator of the popular short-form video app TikTok, is also one of China’s largest cloud computing providers, hosts ByteDance Seed—one of the country’s largest frontier AI research labs—and owns the AI assistant app Doubao. The Doubao app is their main consumer product and had over 170 million monthly active users in China as of October 2025, meaning they have the most active users in the country, ranking above DeepSeek’s app’s 145 million monthly active users.1 ByteDance has the advantage of being able to train its multimodal models on user-submitted videos from TikTok and its Chinese equivalent, Douyin. Douyin, however, is not just a social media platform but also the third largest e-commerce platform in China.2 ByteDance’s stated aim is to integrate AI agents into every facet of how people interact with the internet and offer an immersive consumer entertainment, e-commerce, and social media experience within their own “super-app” ecosystem, something the company’s founder Zhang Yiming, who de facto still leads the company, believes will fundamentally transform user retention and satisfaction.

Read more

This Bismarck Brief will contribute to our exclusive upcoming AI 2026 Bismarck Strategic Report, which will feature in-depth analysis of the state of the artificial intelligence sector and technology, together with a comprehensive profile of key players. Upgrade or subscribe now to receive full access and never miss a report.

With a market capitalization of approximately $1.5 trillion (TSLA) as of January 2026, the U.S. electric vehicle manufacturer Tesla is by far the most valuable auto maker in the world as well as one of the most valuable companies overall.1 Elon Musk has served as CEO of Tesla since 2008, during which time the company has, more than any other, spearheaded the rise of electric vehicles as a viable commercial product and alternative to traditional internal combustion engine (ICE) vehicles. Tesla has also spearheaded another major shift in the design of automobiles: the liberal integration of software and electronics to improve the driving experience for Tesla drivers, which, for many years now, has been expected to culminate in fully autonomous self-driving vehicles with no need for human drivers or intervention. This outcome would be tantamount to an economic revolution in personal, public, and commercial transportation, by eliminating the need for costly and rote human labor in the form of driving.

Read more

This Bismarck Brief will contribute to our exclusive upcoming AI 2026 Bismarck Strategic Report, which will feature in-depth analysis of the state of the artificial intelligence sector and technology, together with a comprehensive profile of key players. Upgrade or subscribe now to receive full access and never miss a report.

Andrej Karpathy is a computer scientist, AI researcher, and expert in computer vision who was a founding member of OpenAI and then head of Tesla’s Autopilot driver assistance and self-driving car program from 2017 to 2022, where he was responsible for building the team and technology that today powers Tesla’s autonomous driving capabilities. Self-driving vehicles promise to be one of the most immediate and consequential areas where AI progress will obsolete human labor in the physical world. Whereas Google’s Waymo and other self-driving efforts have attempted to scale up safe autonomous taxis using multiple types of redundant sensors and extensive reference mapping of the environment, Elon Musk has instead pursued a “vision-only” approach focused on achieving self-driving AI software perceptive enough to maneuver vehicles using only cameras. Karpathy was handpicked by Musk to build this technology for Tesla and largely succeeded.

Read more

This Bismarck Brief will contribute to our exclusive upcoming AI 2026 Bismarck Strategic Report, which will feature in-depth analysis of the state of the artificial intelligence sector and technology, together with a comprehensive profile of key players. Upgrade or subscribe now to receive full access and never miss a report.

Corporate investments in artificial intelligence, especially in the construction of a new generation of data centers built to provide computational power or “compute” for generative AI, have driven over one-third of the United States’ GDP growth in the first nine months of 2025.1 These new data centers will also substantially increase both peak and total electricity demand, after two decades in which power demand was nearly flat, and grew at well under a 1% compound annual growth rate.2 Today, the entire U.S. data center sector draws a total of less than 15 gigawatts (GW) of power, but the pipeline of new data centers under construction will, if all planned facilities right now are completed, add 140 GW of new load to the current total U.S. peak demand of 760 GW, an increase of almost 20%.3

Read more

The depth and breadth of our long-form investigations at Bismarck Brief give rise to many new insights, observations, and models of economics and politics that do not fit neatly into any specific Brief, but are valuable and important to consider nonetheless. Every year, we have collected some of our favorite new insights from the year and published them for our readers—the full list of our 27 previous insights from 2021 to 2024 is available to read here. This year, the tradition will continue. On the cusp of the new year, here are our seven top insights from 2025:

#1 - The prime input for AI progress is elite math and computer talent. The so-called “scaling hypothesis,” that all that is needed to make concrete further progress in artificial intelligence technology is more data and more computing power, has become default consensus in Silicon Valley since the November 2022 release of ChatGPT. But although this theory of further advancement has spurred an unprecedented wave of investment and optimism about AI, it notably neglects the foundational input that necessarily must come before any other: individuals with elite talents in mathematics, computer science, and even software development, often drawn from the halls of academia.

Overlooking this basic fact is arguably what caused U.S. observers to drastically underestimate China’s potential for catch-up AI progress until the debut of DeepSeek in early 2025. Multiple Chinese AI labs drawing on China’s vast domestic reserve of mathematical and computer talent have since made global splashes of their own, such as Alibaba’s Qwen lab, Z.ai, or Moonshot AI. France’s Mistral has drawn on European talent in the same way, nor should it ever be forgotten that it was by leveraging academic computer science talent from North America, through top researchers like Ilya Sutskever, that companies like Google and OpenAI made their early great leaps in the first place.

#2 - Luxury goods are not intrinsically valuable, but instead social technologies designed by live players to generate money and prestige ex nihilo. It is deeply ingrained in our collective subconsciousness that items ranging from glittering gems to fine leather apparel are of a value so great they should not even be second-guessed, let alone haggled over. Yet whatever truly intrinsic value such goods may have, the bulk of profits made and status ascribed thanks to them never comes from this intrinsic value, but the intangible social and signalling value added on top. Such value appears to us also intrinsic, but is actually socially engineered by live players through elaborate marketing, diligently maintained rituals, and often counterintuitively exclusionary practices.

Whether one looks at the deliberately onerous shopping experience employed by Hermès against its own customers, the globe-spanning De Beers monopoly that until recently inflated the perceived value of diamonds, or the hallowed branding of the empire of Bernard Arnault’s LVMH, the notable fact is that such luxury goods are not fundamentally based in technical or logistical realities so much as the skillful manipulation of imaginations. It is this basic unreality of luxury that makes its producers catastrophically vulnerable to fickle changes in taste. But the takeaway for a live player is not that luxury goods are a capricious and risky arena to be avoided, rather that in this arena great wealth and influence can be generated almost purely through charisma rather than brutal political or economic competition.

#3 - A large enough organization is an empire that can contain hidden live players or multiple traditions of knowledge. We are perhaps accustomed to thinking of institutions as the secondary outputs of prolific captains of industry, or mere shells around a generative tradition of knowledge. But an organization with enough personnel, resources, and other owned power, that is sufficiently hard to steer or prune, can also function as an environment within which multiple key players and factions compete for power or quietly bend arms of the organization toward their own external goals.

It is a notable fact that, for example, the most influential person at Microsoft secretly remains Bill Gates, and the most influential person at Amazon remains Jeff Bezos, despite neither men holding majority voting power through shareholdings or the formal chief executiveship at their respective companies. Despite both being perceived as partially or fully retired, they are able to broker significant deals or effect pivotal changes in strategy behind the scenes that evidently no other person can, through mechanisms that are essentially personal and invisible. It was Gates who ultimately greenlit a fateful partnership with OpenAI and brought another live player in Sam Altman into the orbit of Microsoft. Bezos’ nominally independent space company Blue Origin builds rockets that will carry Amazon internet satellites—and perhaps AWS orbital data centers—into space.

At Google, meanwhile, the resources of the world’s most profitable company have not been deployed on a narrow and focused mission but doled out to multiple independent technical traditions of knowledge such as Waymo, Google Brain, and Demis Hassabis’ DeepMind, under the often distant watch of both the company’s two idiosyncratic founders Sergey Brin and Larry Page, as well as the much closer watch of highly effective but very different lieutenants like Eric Schmidt and Sundar Pichai. An organization is sometimes a dead player overall even as its constituent parts may be piloted or merely mined for resources by live players; multiple such organizations now count among the world’s most profitable companies.

#4 - Developed democracies can and will rapidly economically decline if they remain dead players. The very term “developed” is misleading, since it implies that there is an end to economic, technological, and industrial development. Yet the finality of this term, among other similar narrative devices, has lulled us into a false sense of security with regards to the robustness of the wealth, prestige, and public order of the world’s industrialized liberal democracies. Rather than being a static and irreversible state achieved at the end of history, we have seen that developed democracies can take lasting economic damage even in peacetime.

A country like Greece stands as an extreme but early and telling example, having used every tactic including outright fraud to escalate social transfers way beyond fiscal sustainability until a so far permanent relative impoverishment became inevitable. Even much larger countries like Italy and Spain, despite far more industrial virtues and far higher baselines, are slowly yielding their productive sectors to foreign competition and domestic rent-seeking in the absence of decisive leadership. Japan, within living memory the world’s second-largest and most dynamic economy, succumbed to stagnation after succession failure in the commanding heights of its economy. These economies are in the same reference class as those of the United States and Western Europe. Live player leadership is not optional to avoid decline. Greece was perhaps a canary in the coal mine for the rest of us, not an exceptional case from the periphery.

#5 - The militarization of space is beginning and will reshape the global balance of power. After the initial spurt of progress in rocketry and satellite technology caused by the Cold War-era Space Race in the 1950s and 1960s, space technology became the dominion of sluggish defense and aerospace contractors like Lockheed Martin, Boeing, France’s ArianeSpace, and later the Russian successors of the Soviet “design bureaus.” But since the creation of reusable space rockets and the resulting revolutionary decreases in the cost of carrying mass to space pioneered by Elon Musk’s SpaceX, a whole new universe of possibilities for space capabilities has opened up—not just commercial, but military too.

China, foremost, cannot allow the U.S. to develop an uncontested advantage in military capabilities in space. This already is and will further drive Chinese ambitions to match and surpass SpaceX. But neither can lesser industrial and military powers including Russia, India, Japan, and others rest on their laurels. Whether a middle power is able to spin up a serviceable rocket and satellite program to match the state of the art may well become a key determiner of middle power status at all in the relatively near future, in the same way that today a key determiner of this status is whether a power can run an effective air force.

#6 - Progressive philanthropy and activism is increasingly shaped by none other than tech billionaires. Although prominent voices in software and venture capital have taken a political turn conciliatory to or even vocally supportive of President Donald Trump and the Republican Party since 2024, a perhaps equally prominent and influential contingent of key players in Silicon Valley have become, and remain, new nodes tying tech wealth to the Democratic Party and wider progressive activist movement.

Steve Jobs’ widow Laurene Powell Jobs is a major financial patron of progressive prestige media and a personal confidante of former Democratic presidential candidate Kamala Harris. When Elon Musk acquired Twitter in a bid to loosen censorship and speech restrictions, it was the fortune of eBay founder Pierre Omidyar that financed his most vociferous opponents, who ran a successful campaign to get advertisers to drop the company. Peter Thiel’s perhaps former close friend and LinkedIn cofounder Reid Hoffman has made it a personal mission to oppose Trump even as he plays a notable behind-the-scenes role in AI. Despite the outsized popularity and visibility of some stalwart contrarians, Silicon Valley’s politics were until recently monolithic, and the era of a truly bipartisan tech scene is still brand new and fragile. Nor is the U.S. likely, anymore, to see a political future free of the touch of tech wealth.

#7 - Judges and bureaucrats are resilient to populist groundswells. The tug-of-war for electoral and cultural primacy between credentialed elites and populist coalitions enabled by the internet has now been ongoing for a decade in North America and Europe. There have been victories by both sides, even lasting populist victories in countries such as Hungary. While this conflict remains live and unresolved, it is notable that it has clear parallels well outside of the economic and cultural environment of the developed world.

In India, for example, a winning populist middle-class coalition led by Narendra Modi and with the backing of the country’s wealthiest businessmen like Mukesh Ambani has so far failed to dislodge a sclerotic civil service or overpowered supreme court. This suggests that institutional factors are as crucial to consider as cultural, demographic, or economic ones in this ongoing struggle. One of the most dramatic governance turnarounds of the decade, that of President Nayib Bukele in El Salvador, began not as a populist revolt but an arm’s-length alliance with established but lethargic and morally discredited parties. What lessons do these examples and others hold for us?

If you haven’t already, you can read the full list of our 27 previous insights from 2021 to 2024 here.

In early 2026, Bismarck Analysis will be releasing our first-ever comprehensive annual report on a key industry, titled AI 2026: Bismarck Strategic Report. At an estimated final length of between 225 to 275 pages, this industry-grade document will feature all of the information and analysis a live player needs to navigate the ramifications of our society’s financial, political, and human-capital investment in AI, and what we can expect from this transformative technology. Read more here or, if you wish to receive the report, become a subscriber to Bismarck Brief at the AI 2026 tier here.

If you aren’t already a paid subscriber, I warmly invite you to become one and join us on this ongoing investigation into the global power landscape, with a new in-depth investigation in your inbox every Wednesday at 2pm GMT sharp:

Subscribe now

If you are already a paid subscriber, you can also give the gift of Bismarck Brief to a friend, family member, or colleague this holiday season.

Happy New Year!

Samo Burja

Founder and President, Bismarck Analysis

In November 2025, Blue Origin became the second organization in the world to demonstrate reusability of an orbital space rocket, with the vertical landing and recovery of the first stage or “booster” of its New Glenn heavy-lift rocket.1 The first organization to accomplish this feat was Elon Musk’s SpaceX, which demonstrated reusability of its Falcon 9 medium-lift rocket in December 2015 and, as a result, has greatly reduced the cost of launching mass to orbit, accounting in 2024 for roughly 90% of all mass launched to orbit worldwide.2 With New Glenn, Blue Origin has become the closest contender in the international race to match or surpass SpaceX. Blue Origin was founded and remains generously funded by the personal fortune of Amazon founder and executive chairman Jeff Bezos, who is the de facto leader of Blue Origin although he does not hold the title of CEO.

Read more

This Bismarck Brief will contribute to our exclusive upcoming AI 2026 Bismarck Strategic Report, which will feature in-depth analysis of the state of the artificial intelligence sector and technology, together with a comprehensive profile of key players. Upgrade or subscribe now to receive full access and never miss a report.

Ilya Sutskever is one of the world’s leading AI scientists. As of December 2025, he is the CEO of Safe Superintelligence (SSI), a startup he co-founded upon his departure from OpenAI in May 2024, where he was also a co-founder and worked for over eight years as Chief Scientist. Under his technical leadership, OpenAI, which began as a small non-profit, became the world’s leading generative AI company with a sizable lead over its competitors in both technology and paying user base, a lead it still retains today.

Read more

This Bismarck Brief will contribute to our exclusive upcoming AI 2026 Bismarck Strategic Report, which will feature in-depth analysis of the state of the artificial intelligence sector and technology, together with a comprehensive profile of key players. Upgrade or subscribe now to receive full access and never miss a report.

John Carmack is an American software engineer and video game developer who cofounded the game studio id Software, whose game series including Doom, Quake, and Wolfenstein have been widely credited with not just popularizing but laying the technical groundwork for fast-paced video games utilizing three-dimensional graphics in the 1990s, in particular the “first-person shooter” (FPS) genre. Carmack was the lead programmer on these series and has been credited with developing multiple techniques that made it possible to smoothly render computationally-intensive graphics. Carmack’s early software work thus helped to drive adoption of specialized, high-performing computer chips called graphics processing units (GPUs), designed by Nvidia. Decades later, Nvidia’s powerful GPUs have become the standard for developing and using computationally-intensive artificial intelligence software, and, since 2022, Carmack has in turn devoted his time to running his own AI research lab aiming to build artificial general intelligence (AGI).

Read more

Just over four years ago, we issued the first Bismarck Brief. Since then, our talented analysts have written well over 1 million words on the technological, geopolitical, and institutional state of t…

Read more

This Bismarck Brief will contribute to our exclusive upcoming AI 2026 Bismarck Strategic Report, which will feature in-depth analysis of the state of the artificial intelligence sector and technology, together with a comprehensive profile of key players. Upgrade or subscribe now to receive full access and never miss a report.

It is rare for someone to be a competent researcher as well as able to navigate sometimes cutthroat corporate politics. Because he can do both, currently 49-year-old Demis Hassabis is one of the most influential technologists working on artificial intelligence today. Hassabis is currently CEO of Google DeepMind, an organization with notable research capabilities demonstrated by their many technical firsts spanning over a decade of AI research. Such research capacity is surprisingly rare in the crowded field of giant companies eagerly contending to lead the AI industry, perhaps matched by only a few other exceptions such as OpenAI and DeepSeek. For example, under Hassabis, DeepMind built the program AlphaFold to predict protein structures, for which Hassabis was awarded the 2024 Nobel Prize in Chemistry and was knighted in the United Kingdom.

Read more

This Bismarck Brief will contribute to our exclusive upcoming AI 2026 Bismarck Strategic Report, which will feature in-depth analysis of the state of the artificial intelligence sector and technology, together with a comprehensive profile of key players. Upgrade or subscribe now to receive full access and never miss a report.

Z.ai, known as Zhipu AI until 2025, is a Chinese artificial intelligence startup founded in 2019 by Tang Jie and Li Juanzi, both professors at Tsinghua University, Chinese paramount leader Xi Jinping’s alma mater and one of China’s most prestigious universities. Alongside companies like DeepSeek and Moonshot AI, it is one of the leading AI startups in China, known primarily for the “GLM” series of open-source large language models (LLMs) with low costs of inference i.e. usage and high performance, especially on coding and “agent” tasks. The company is headquartered in Beijing and reportedly employs around 500 people, with a core research team of between 100 to 200 staff members.1 It is led by CEO Zhang Peng, an AI researcher and Tsinghua alumnus, and was one of the first companies in China to build a large language model before the viral release of OpenAI’s ChatGPT in November 2022.

Read more

This Bismarck Brief will contribute to our exclusive upcoming AI 2026 Bismarck Strategic Report, which will feature in-depth analysis of the state of the artificial intelligence sector and technology, together with a comprehensive profile of key players. Upgrade or subscribe now to receive full access and never miss a report.

Alibaba is the second-most valuable technology company in China, with a market capitalization of nearly $400 billion (BABA) as of November 2025 and approximately $130 billion in revenues in 2024.1 Although it is primarily an e-commerce company that has been dubbed “China’s Amazon,” it today also hosts an elite artificial intelligence research lab and is one of China’s major providers of cloud computing services i.e. remotely renting out access to servers for computing and data storage, and building the large data centers to do this. Tongyi Lab oversees the development of the Qwen series of large language models (LLMs), which have become some of the most popular models in China for businesses, academics, and consumers. Almost all the Qwen models are released on highly permissive open-source licenses, allowing individuals to download them and external providers to offer them to customers without paying Alibaba. The models are also winning customers outside of China for their very low costs of inference i.e. usage and high performance: for example, Airbnb CEO Brian Chesky recently revealed that Airbnb’s customer service chatbot is largely powered by Qwen models.2

Read more