This is an edited excerpt of the book The Rulers - Corporate Power in the Age of AI and the Cloud, by Cecilia Rikap, to be released on 4 August 2026 by Verso Books.
In early 2024, the non-profit project ‘AI Impacts’ published the findings of its second survey, targeting almost 3,000 researchers who had published at top-tier AI conferences. Among the questions posed was one concerning the long-term value of AI progress and ‘between 38% and 51% of respondents gave at least a 10% chance to advanced AI leading to outcomes as bad as human extinction, . This concern, particularly focused on generative AI, became known as P(doom) – the probability that AI will lead to humanity’s downfall. The term P(doom) came up in dialogues between Google’s chief scientists and leading academics with whom the company frequently collaborates.
I will not engage here in the debate over whether AI’s misuses result in weapons of mass destruction, simply because AI is already being used as a weapon by corporations and states, and it is highly likely that the ecological breakdown – which AI is fueling – will lead faster to our doom anyway. P(doom) is interesting because it illustrates how companies shift the responsibility for a technology’s effects onto the technology itself.
By blaming the technology, discussions around AI-induced human extinction divert attention from scrutinizing the organizations deciding what AI is developed, how it is produced and for what purposes. The same applies to arguments claiming that AI is progressing too quickly to be halted. Certainly, AI is advancing rapidly, leaving many behind, but someone is stepping on the accelerator. Framing technology as the source of danger serves not only to mislead the public but also to fulfil an internal purpose.
When AI is portrayed as an autonomous, self-determining force, it allows scientists and engineers within Big Tech’s innovation systems to feel less accountable for their own decisions. If the technology itself is deemed responsible, these developers can continue modelling and coding without considering their ethical responsibilities, especially when AI is presented as the key to solving all the world’s problems.
But AI is not an independent entity. Like all technologies, AI is a product of human labour and decision-making. Technologies are not autonomous; in capitalism, they are shaped by labour, investments, alliances, corporate interests and the personal ambitions of CEOs, executives and chief scientists, among others.
By dominating the AI research field, Big Tech – especially Google and Microsoft given their centrality in the global knowledge network – exerts substantial influence over the AI currently being developed and, consequently, over how societies are affected by it. Although LLMs [large language models] possess a form of agency in the sense that developers cannot predict all the answers the model will generate, many human decisions are made throughout the process of creating them. The decision-making process begins with the choice to develop LLMs at all and continues with decisions about what data to use for training, what parameters to apply, what the model’s objectives should be, what type of training to conduct and many more.
The very act of calling ChatGPT and similar models ‘AI’ is itself a crafted narrative. ChatGPT is not intelligent; it is a proficient language predictor, provided the necessary information for prediction exists within its memory. LLMs – and AI models more broadly – are what linguistics professor Emily Bender and her colleagues have termed‘stochastic parrots’. The term highlights how LLMs predict responses stochastically based on text processing, without any understanding or assigning meaning to the word sequences they generate as the most likely answers.
Consider a robot footballer or a robot surgeon. Trained with every single relevant past data – all existing football plays or surgical procedures – they may become highly proficient, but they will work along a completely different basis than human knowledge. Because AI has no experiences, it lacks imagination and understanding. An experience is a situated compound of logical thinking, emotions and actions, a creative interplay where the subject and the object configure each other in history. An extemporaneous collection of data, by definition, is a frozen set of particles incapable of retrieving, even through multiple combinations, the dialectics of historical experience.
When Big Tech labels its prediction models as ‘AI’ and claims to be pursuing artificial general intelligence capable of surpassing human abilities across domains, it is implicitly endorsing a narrow definition of intelligence that is usually centered around solving complex coding and math problems. This definition fixates on individual optimization, leaving no room for collective intelligence or the possibility of choosing suboptimal individual outcomes for the sake of greater well-being and happiness for others or for causing less harm to the planet.
It is not me, it’s you
Big Tech companies have also significantly shaped other knowledge fields beyond those directly involved in developing AI models. This influence is particularly evident in AI ethics, which has been moulded since its early days through Big Tech funding almost every prominent scholar working on AI ethics at leading universities. As a result, this field – and, even more so, the narrower ‘AI safety’ sub-field – is predominantly concerned with problems that overlook corporate power and Big Tech’s responsibility.
A typical example is the study of biases. When models are found to be biased, the blame is placed on biased datasets. Since datasets are socially constructed, they are said to reflect societal biases; therefore, if societies discriminate, then models will inevitably do the same. Big Tech companies conveniently position themselves as the first responders to combat these biases, thus framing themselves as problem-solvers rather than taking responsibility for the bias.
A similar pattern of co-optation is occurring within the social sciences and humanities more broadly, through initiatives such as Microsoft’s Social Media Collective, where many prominent media studies and social sciences scholars are affiliated. Even the author of a widely cited article explaining how the term ‘platform’ originated from the industry’s cultural vocabulary to convey a sense of horizontal, democratic structure – while obscuring their underlying power – later served as a senior principal researcher at Microsoft.
An economic piece co-authored by another Microsoft researcher and university scholars is titled ‘Digital Addiction’. The article argues that individuals are to blame for their addiction to smartphones and social media, as if there were no mechanisms employed by platforms to nudge, persuade and even shape online behaviours. It was published in the American Economic Review, one of the top five economics journals, thereby clearly influencing the entire discipline while hiding that Big Tech explicitly designs algorithms to keep us glued to the screen.
Whistleblower Frances Haugen provided evidence that Meta was aware of the harms caused by its addictive algorithms, and it later emerged that Google and Meta had made a deal to enhance Instagram’s appeal to under-18s through a YouTube campaign, circumventing Google’s own rules on advertising to minors.
Another scientific publication, this time by Google’s chief economist Hal Varian, concludes that tech is ‘a dynamic, competitive industry with high levels of investment in R&D and capital equipment, high wages for workers, and rapid growth’. The article gives the impression that this is the healthiest industry of all time. What Varian does not mention is that he considers an industry competitive only by looking at prices and consumers. He neglects the power imbalance between Big Tech and the myriad of start-ups and other organizations that operate within their spheres of control.
Published in reliable scientific outlets, these articles illustrate that Big Tech’s aim to dominate discourses goes beyond computer science. They also seek to influence the research agendas of fields that study the social, ethical and economic impacts of their controlled technologies, partly by shifting the conversation away from holding them accountable for harmful consequences.
The same strategy is now being employed to divert attention from AI’s high consumption of resources – particularly energy and water. Amazon, Microsoft and Google emphasize that their public clouds are more efficient than alternative solutions for training and running AI models, such as private clouds or on-premises datacentres. But even when the overall equation points to a more efficient solution, their push leads to expanded use of their clouds – further boosting their profits – resulting in higher resource consumption.
Microsoft’s CEO Satya Nadella stated it himself on his LinkedIn account in early 2025: ‘Jevons paradox strikes again! As AI gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can’t get enough of.’ This skyrocketing use translates into skyrocketing consumption of energy and water. The way in which Amazon, Microsoft and Google present their clouds as the absolute best is an example of a broader practice among intellectual monopolies. To showcase themselves as the most innovative and cutting-edge companies, they even design frontier technology solutions without intending to directly profit from them.
Bullshit innovation
Intellectual monopolies exploit their extraordinary financial leeway to produce what can be described as ‘bullshit innovation’: products whose ultimate purpose is not to be sold or widely adopted by the public, nor to generate direct revenue, but rather to position the corporation as a driving force of frontier technological change. (In my definition of bullshit innovation I draw on David Graeber’s idea of bullshit jobs). Such products reinforce their identity as innovative companies, likely enhancing the perception of other products that could achieve mass adoption and, consequently, contribute to corporate accumulation.
Elon Musk excels at creating bullshit innovation or, more accurately, constantly promising bullshit innovation. Since it is impossible to keep track of developments across every corner of the world, some might believe there are cities where Tesla self-driving cars roam around as commonly as bicycles. Whether this is true or false is irrelevant; it effectively establishes a narrative of Tesla as a cutting-edge innovator.
A similar scenario unfolds with SpaceX plans to land on the Moon, despite Tesla and SpaceX being two separate companies. Musk founded them both, so if SpaceX can land on the Moon, why not simply accept Tesla as the best mobility option on Earth? Although Musk’s histrionic personality may suggest that such promises are exclusive to the same allegedly unique man, this strategy is far from unique to Musk and his companies. Intellectual monopolies frequently combine bullshit innovation that directly loses money with systematic innovations co-produced with others, which enable intellectual monopolies to capture intellectual rents.
DeepMind’s AI model for predicting protein structures, AlphaFold, can be considered bullshit innovation. It is a frontier scientific achievement that even won a Nobel Prize, but it is not intended to directly become the basis of a Google product. AlphaFold is free to use, and Google has no intention of creating an ecosystem around it as a means of generating profit, unlike in the case of Android. Instead, it serves to position DeepMind, and by extension Google, as a frontier innovator making a substantial contribution to healthcare research.
Digital technologies, partly because laypeople often do not understand how they work, are a perfect ally for bullshit innovation. SAP’s [the main provider of software to manage operations like finance, HR, and supply chains] lab in the US produces AI and robotics demos and one-offs to showcase the power of the technology underpinning its new Enterprise Resource Planning software, HANA (High-performance ANalytic Appliance). The company does not expect clients to ever use those demos. As a SAP AI engineer explained, the purpose was merely to demonstrate that SAP HANA was a frontier and powerful tool.
Technologies can also start as bullshit innovations and have the potential to make a future business. Generative AI could be considered such an innovation that might eventually prove itself otherwise. So far, while OpenAI continues to lose millions annually, generative AI has served as a magnet to attract more customers to Amazon, Microsoft and Google clouds. This includes those developing solutions on top of existing models that all run on the cloud, as well as those migrating to the cloud drawn by generative AI but who ultimately use other computing services, even different AI models.
Speaking of the use of AI as bullshit innovation, in 2020 L’Oréal released Perso, a sort of Thermo Mixer for beauty that uses AI to deliver personalized, on-the-spot skincare and cosmetic formulas. One might wonder how many people would even consider buying such a device if they could afford it, but Perso sales are not L’Oréal’s primary motivation.
The aim is to present L’Oréal as an innovator attuned to the digital age, with Perso serving as a piece of evidence.
Likewise, Amazon invests in home robots, although, so far, their most advanced versions are test products. The goal is twofold, with both motivations complementing each other and reinforcing Amazon’s intellectual monopoly. On the one hand, it feeds the narrative that Amazon is an innovator developing cutting-edge technology. On the other hand, Amazon’s focus on robots encourages others to explore this area further – whether by creating start-ups aiming to be acquired by or sell technology to Amazon, or by inspiring curious scholars eager to develop technology relevant to those at the frontier. In other words, Amazon motivates others to research and develop home robots. These motivations could be coupled with various others in the future, ranging from feminist advocacy for socializing domestic duties to military interest in robots for purposes that could later be easily adapted for home use. All these motivations, together with technological advancements, could eventually transform home robots into a widely consumed product. Amazon will be prepared to capture that demand. In the meantime, its home robots remain a form of bullshit innovation.
[...]
As I hope becomes evident after reading The Rulers, we live in a world where the production of knowledge – the fundamental resource for organizing individual and collective life – is overwhelmingly controlled by Amazon, Microsoft and Google. Their dominance has reached a point where even other powerful leaders have become dependent on their monopolized technologies. For this end, Big Tech companies systematically transform knowledge and information produced by each and every individual and organization into their assets. They are, in essence, the world’s largest knowledge predators.
Cecilia Rikap is associate professor in economics and head of research at the University College London’s Institute for Innovation and Public Purpose. She is also an associate researcher at COSTECH lab, Université de Technologie de Compiègne, France. She is the author of Teoría de la Dependencia Digital and of Capitalism, Power and Innovation: Intellectual Monopoly Capitalism Uncovered, which won the EAEPE Joan Robinson Prize Competition in 2023.
