The New ‘Fall of Constantinople’: How AI Scaling Could Reshape Global Power by 2030
Generative AI capabilities have been increasing fast. For some bystanders, this has been just another example of a tech hype and bubble that will soon burst. Some others frame the development as a significant turning point. A similar division of worldviews is visible within the people who work in the industry. A vast majority remains “realists”, who predict and even hope for a new AI winter or at least an AI fall, claiming that the development cannot continue in such a fast pace. A small and vocal minority, partly from the less convincing tech evangelist side, are accelerationists. They say that the development will continue, and that it should continue.
In this ongoing debate, I was enjoying the ride but firmly on the realist side. The consensus by most industry people and in foresight discussions with experts has long been that AI’s explosive development — currently growing at 4x annually — would eventually slow down after we’d hit physical and technological limits.
However, recent research has shed clarity to my realist perspective. Even if current research on AI development bottlenecks indicates that AI development will continue, my claim is that AI scaling cannot proceed as unchecked, because unlimited AI development will trigger geopolitical consequences that slow down its development.
Changing Winds
To comprehend the magnitude of current geopolitical shifts, we must revisit a pivotal moment in history: the fall of Constantinople in 1453. This event was not merely the capture of a city; it represented a seismic shift in global power dynamics, altering trade routes, sparking new eras of exploration, and redefining sovereignty in the known world.
For over a millennium, Constantinople stood as the crossroads between Europe and Asia, a vibrant hub where East met West. Its strategic location on the Bosporus Strait made it the linchpin of the Silk Road and other critical trade networks. The city’s wealth and influence were unmatched, serving as a cultural and economic bridge that facilitated the exchange of goods, ideas, and technologies.
By the mid-15th century, however, the Byzantine Empire had dwindled to a fraction of its former glory. The once-mighty empire had been eroded by relentless conflicts with neighbouring Balkan states and internal strife exacerbated by religious schisms, notably the Great Schism of 1054, which divided Eastern Orthodox and Roman Catholic Christianity. The Fourth Crusade’s sacking of Constantinople in 1204 had further strained relations with Western Europe and inflicted lasting damage on the empire’s infrastructure and morale.
Despite these challenges, Constantinople remained a beacon of hope for European powers wary of the rising Ottoman Empire. The city’s formidable walls and strategic position made it a crucial bulwark against Ottoman expansion into Europe. Yet, in 1453, Sultan Mehmed II launched a well-orchestrated siege, utilizing advanced military technology of the time, including massive cannons capable of breaching the city’s ancient defences. After 53 days, Constantinople fell, marking the end of the Byzantine Empire and a significant shift in the balance of power.
The fall sent shockwaves throughout Europe. Not only did it symbolie the rise of the Ottoman Empire as a dominant regional power, but it also effectively severed the overland trade routes that European merchants had relied upon for centuries. The Ottomans now controlled the passage between the Black Sea and the Mediterranean, imposing hefty taxes and restrictions on trade. European nations, particularly those along the Atlantic coast like Portugal and Spain, were compelled to seek alternative routes to access the wealth of Asia.
This necessity became the mother of invention, igniting the Age of Exploration. Portuguese explorers, under the patronage of Prince Henry the Navigator, began charting the West African coast, seeking a sea route to India. In 1492, Christopher Columbus, funded by Spain, embarked on his voyage westward, inadvertently discovering the Americas. These explorations not only opened new trade routes but also led to the colonization of vast territories, the exchange of goods and cultures, and, unfortunately, the exploitation and displacement of indigenous populations.
The consequences of this expansion were profound and far-reaching. European colonisation triggered a seismic shift in the demographic, environmental, and cultural landscapes of the Americas. The arrival of Europeans introduced diseases such as smallpox, measles, and influenza to which Native American populations had no immunity. Historians estimate that these epidemics, coupled with warfare and forced labor, led to the deaths of an estimated 56 million indigenous people — a catastrophic decline that represents one of the largest demographic collapses in human history.
This massive depopulation had unintended environmental effects. Vast tracts of previously cultivated land were abandoned, leading to widespread reforestation. Some scientists suggest that this regrowth of forests across the Americas sequestered significant amounts of carbon dioxide from the atmosphere. The reduction in atmospheric CO₂ may have contributed to the global cooling observed during the “Little Ice Age” in the 16th and 17th centuries. This period was marked by colder winters, shorter growing seasons, and frequent crop failures in Europe, exacerbating economic hardships and social unrest across the continent.
Yet, amidst these tragedies, the Age of Exploration also facilitated a profound exchange of ideas and philosophies. European explorers, missionaries, and settlers encountered complex societies with rich cultural traditions, sophisticated governance structures, and deep philosophical insights. Indigenous leaders like Kandiaronk, a Huron-Wendat statesman and diplomat, engaged in extensive dialogues with European intellectuals and missionaries. Kandiaronk’s sharp critiques of European societal norms — particularly the entrenched hierarchies, materialism, and lack of communal solidarity — challenged the Europeans to reflect on their own social and political systems.
These interactions had a significant impact on Enlightenment thinkers back in Europe. Philosophers such as Jean-Jacques Rousseau and Denis Diderot were influenced by accounts of indigenous societies that emphasised equality, shared resources, and consensus-based governance. Kandiaronk’s perspectives, conveyed through works like the “Dialogues of a Huron” by Baron de Lahontan, questioned the legitimacy of absolute monarchy and the stark inequalities prevalent in European societies.
This infusion of new ideas contributed to the intellectual climate that eventually led to the French Revolution in 1789. The revolution was a watershed moment that dismantled feudal structures, challenged the divine right of kings, and promoted the principles of liberty, equality, and fraternity. It set the stage for the development of modern political systems in the Global North, emphasising individual rights, secular governance, and representative institutions.
The ripple effects of the French Revolution extended beyond Europe, influencing democratic movements worldwide and reshaping international relations. The post-World War II era saw the establishment of a world order based on these principles, with institutions like the United Nations promoting peace, human rights, and cooperation. This framework has been instrumental in shaping global politics, economics, and cultural exchanges for decades.
The fall of Constantinople reshaped global power structures, compelling nations to adapt rapidly to a new world order. Today, we stand at a similar crossroads with the advent of artificial intelligence. Just as control over trade routes once dictated a nation’s influence, dominance in AI technology now holds the key to future geopolitical power. The United States, leveraging its robust technological infrastructure, is poised to lead this new era — unless significant disruptions alter the course.
So the key question regarding world’s geopolitical order is: will AI scaling continue until 2030?
AI Scaling Will Continue Until 2030
The trajectory of artificial intelligence development suggests that we are on the cusp of another monumental shift — one that could redefine global power structures by 2030. Contrary to concerns that AI scaling might plateau due to physical or economic limitations, recent research indicates that AI scaling is not only continuing but accelerating at an unprecedented rate. If external events do not intervene, we could witness AI training runs performing up to 2×10^29 floating-point operations (FLOPs) by the end of this decade — orders of magnitude greater than today’s capabilities. This escalation is propelled by advancements in three critical areas: power availability, chip manufacturing capacity, and data growth.
The backbone of AI scaling is the energy that powers data centres worldwide. Projections show a significant ramp-up in global data center energy capacities by 2030. In the United States alone, data centres are expected to support between 2 to 45 gigawatts (GW) of power, according to recent estimates from industry analyses. This vast increase in energy availability is pivotal for enabling large-scale AI training runs.
This expansion is driven by substantial investments in renewable energy sources and more efficient power management systems. Tech giants and data center operators are increasingly committing to carbon-neutral or even carbon-negative operations, leveraging wind, solar, and hydroelectric power. Not only does this address sustainability concerns, but it also ensures a stable and scalable energy supply for the burgeoning demands of AI computations. Innovations in cooling technologies, such as liquid immersion cooling, further enhance energy efficiency by reducing the power required to maintain optimal operating temperatures for servers.
Another cornerstone of AI scaling is the availability of advanced semiconductors. Despite geopolitical tensions and sanctions that have limited certain countries’ access to cutting-edge chip technology, the global production capacity is expanding robustly. Companies like NVIDIA, along with semiconductor manufacturers like TSMC (Taiwan Semiconductor Manufacturing Company), are significantly increasing their output of high-performance graphics processing units (GPUs) and specialised AI accelerators essential for AI computations.
By 2030, the market could see the availability of over 100 million H100-equivalent chips — the latest in NVIDIA’s line of AI-optimized processors. This surge in hardware availability makes the development and deployment of larger, more complex AI models feasible. It lowers the barriers to entry for organisations aiming to leverage AI at scale, democratising access to advanced computational resources. Additionally, advancements in chip architecture, such as the development of application-specific integrated circuits (ASICs) and neural processing units (NPUs), are expected to enhance computational efficiency and performance, further fuelling AI scaling.
The fuel for AI models is data, and the volume of accessible data is expanding exponentially. While text data has been the primary driver of AI training thus far, the inclusion of multimodal data — encompassing images, video, audio, and sensor readings — is set to revolutionise the field. The proliferation of internet-connected devices, the expansion of the Internet of Things (IoT), and the increasing digitisation of various industries contribute to an ever-growing repository of data.
By 2030, this expansion could enable AI training runs requiring between 6×10^28 to 2×10^32 FLOPs. The richness and diversity of this data will allow AI systems to understand and interact with the world in more nuanced and sophisticated ways, pushing the boundaries of machine learning capabilities. For instance, advances in natural language understanding, computer vision, and sensory data processing will enable AI to perform complex tasks such as real-time language translation, autonomous navigation, and predictive analytics with unprecedented accuracy.
Furthermore, the advent of synthetic data generation techniques can augment training datasets, addressing limitations related to data scarcity and privacy concerns. This approach not only amplifies the volume of training data but also enhances its diversity and representativeness, leading to more robust and generalisable AI models.
10 000 better generative AI
The contemporary global landscape is witnessing significant challenges to this established order. Emerging powers such as China are asserting their influence and proposing alternative models of governance and development. China’s rapid economic growth, technological advancements, and strategic initiatives like the Belt and Road Initiative have positioned it as a formidable contender on the global stage. This shift raises questions about the future of international norms, power structures, and the balance of influence among nations.
With 4x/year accelerating capability increases in generative AI, by 2030, generative AI by US-based companies would use 10 000 larger foundation models. This progress would likely bolster U.S. economic and military power, as AI advances are critical not only for automation and productivity but also for defence and cybersecurity. Such dominance would further marginalise China’s ability to compete in the global AI arms race, threatening its strategic goals and eroding its sovereignty in the technological arena.
A concrete example of how AI military capabilities could shape this competition can be seen in the deployment of AI-powered drone swarms versus AI-guided anti-air systems. Imagine a scenario where the U.S. deploys a swarm of autonomous drones programmed to overwhelm and disable an adversary’s AI-controlled anti-air defences. These drones, capable of real-time communication and coordination, could attack from multiple directions, adapting their strategies based on the response of the anti-air system. In turn, the anti-air system — armed with AI algorithms that predict and counter aerial attacks — would attempt to track and destroy the drones before they can penetrate critical defence zones.
In such engagements, the side with the more advanced AI will have the upper hand. If the drone swarm AI can outmaneuver the anti-air system, the defending country could be left vulnerable to further strikes, demonstrating the strategic importance of AI superiority in modern warfare. Conversely, an AI-driven anti-air system capable of predicting the behaviour of swarm attacks and efficiently neutralising them could hold the line against an otherwise overwhelming force.
This hypothetical battle highlights the critical role AI will play in military strategy by 2030. The ability to develop more sophisticated algorithms, faster decision-making systems, and superior computing power could be the difference between victory and defeat on the battlefield. In this context, if the U.S. continues its current rate of AI scaling, China’s military capabilities could become increasingly obsolete, pushing the nation to explore more drastic actions, such as targeting Taiwan’s semiconductor industry, to prevent falling too far behind in the AI arms race.
The Modern “Fall of Constantinople”
While the technical capabilities for AI scaling exist, the continuation of this trajectory is not guaranteed. Geopolitical factors could significantly alter the landscape. As AI becomes increasingly central to national security and economic prosperity, nations may take drastic measures to secure their positions.
China, recognising the strategic disadvantage posed by unchecked AI scaling in the United States, may be unwilling to allow this trend to continue unchallenged. Facing restrictions on access to advanced semiconductors and lagging behind in certain AI capabilities, China might consider taking drastic actions to disrupt the current trajectory of AI development.
One potential strategy could be a decisive move against Taiwan, the world’s leading producer of advanced semiconductors. Taiwan Semiconductor Manufacturing Company (TSMC) is critical to the global supply of cutting-edge chips essential for AI training and deployment. By asserting control over Taiwan — analogous to “conquering Constantinople” — China could secure access to these vital resources while simultaneously disrupting the semiconductor supply to the United States and its allies.
This scenario draws parallels to historical events where control over critical chokepoints reshaped global power dynamics. Just as the fall of Constantinople forced European powers to seek new trade routes, a disruption in the semiconductor supply chain could compel nations to reassess their strategies in AI development.
For China, initiating such a drastic action would not be taken lightly, given the significant economic and political risks involved. However, the potential benefits could be deemed worth the cost. By causing an external disruption to the system, China could delay the United States’ AI scaling for up to a decade, buying valuable time to enhance its own AI capabilities.
During this period, China could leverage its vast and diverse datasets — spanning over a billion citizens and extensive state-collected information — to develop AI systems that could outcompete those of the U.S. and Europe. China’s centralised data policies and less stringent privacy regulations provide it with an abundance of training data, potentially giving it an edge in creating more sophisticated and adaptable AI models.
Will China let AI scaling continue uninterrupted until 2030? While the technical and economic conditions appear favourable for continued exponential growth in AI capabilities, geopolitical realities introduce significant uncertainty. Nations like China may not passively allow competitors to achieve unassailable technological dominance. The potential for drastic actions, such as disrupting global semiconductor supply chains, cannot be discounted.
The future of AI scaling is thus entwined with global politics. The decisions made by nations in the coming years will shape not only the trajectory of AI development but also the balance of power in the decades to come. As history has shown, technological advancements often provoke strategic responses that redefine world orders.