Volkswagen overtakes Amazon as Rivian’s largest shareholder with 15.9% stake after $1B software milestone payment



TL;DR

Volkswagen has overtaken Amazon as Rivian’s largest shareholder after a one billion dollar share purchase triggered by a software joint venture milestone. VW now holds 15.9 per cent of Rivian, while Amazon’s stake has diluted from 20 per cent to 11.8 per cent without selling a share. The investment reflects VW’s failed internal software division and its dependence on Rivian’s zonal architecture for its next-generation vehicles.

When Rivian went public in November 2021, Amazon owned 20 per cent of the company. It had backed the electric vehicle startup with a 700 million dollar cheque in 2019, ordered 100,000 electric delivery vans, and watched its investment surge to more than 15 billion dollars on Rivian’s first day of trading. Four years later, Amazon has not sold a single share. Its stake has still fallen to 11.8 per cent, diluted by the successive capital raises that have kept Rivian alive while it burns through billions trying to become a real car company. On Monday, a new SEC filing confirmed what the dilution arithmetic had been pointing toward for months: Volkswagen has overtaken Amazon as Rivian’s largest shareholder, purchasing 62.9 million new shares on 30 April at 15.90 dollars apiece, a roughly one billion dollar investment that brought VW’s total holding to 209.8 million shares, or 15.9 per cent of the company. It is the first time since Rivian’s IPO that Amazon has not been its biggest backer, and it says more about what has gone wrong at Volkswagen than about what has gone right at Rivian.

The deal

Volkswagen’s latest billion-dollar tranche was triggered by the joint venture between the two companies hitting a specific testing milestone. The partnership, announced in 2024 with a total commitment of up to 5.8 billion dollars, centres on the development of a zonal electrical architecture for software-defined vehicles. Zonal architecture replaces the dozens of domain-specific electronic control units scattered throughout a conventional car with a handful of centralised computing zones, reducing wiring complexity, enabling over-the-air software updates, and providing the foundation for autonomous driving features. The milestone that unlocked the latest payment was the completion of winter testing of the production-intent architecture for VW’s first-generation software-defined vehicles. The joint venture now employs more than 1,500 people across international development centres, and reference vehicles from the Volkswagen, Audi, and Scout brands have entered the testing programme.

The architecture is the point. Volkswagen did not invest nearly six billion dollars in Rivian because it wanted to own shares in a company that delivered 10,365 vehicles last quarter and lost 3.6 billion dollars last year. It invested because its own software division, CARIAD, failed catastrophically. Launched in 2020 by then-CEO Herbert Diess with the ambition of making Volkswagen a software company, CARIAD consumed billions in development costs, pushed back the release of key Volkswagen, Audi, and Porsche models by nearly two years, and was eventually stripped of its lead development role. Volkswagen had already been cutting EV production as demand faltered across Europe, and the software delays compounded a strategic crisis that culminated in the company’s first German factory closure in 88 years and the planned elimination of 35,000 jobs. Rivian’s zonal architecture is the replacement for the platform CARIAD could not build. The equity stake is the price of admission.

The shift

Amazon’s displacement as Rivian’s top shareholder is symbolic but not accidental. Amazon’s investment was always strategic in a different sense: it bought a stake in its delivery van supplier, and the commercial relationship has proved durable. In Rivian’s first quarter of 2026, Amazon accounted for 468 million dollars of the company’s 908 million dollars in automotive revenue, more than half, through continued deployment of electric delivery vans across its logistics network. But Amazon has not increased its shareholding. It has watched passively as Volkswagen’s milestone-based investment programme steadily increased VW’s ownership while Amazon’s percentage shrank through dilution. The shift in the shareholder register reflects a shift in what Rivian is worth to its biggest investors: to Amazon, Rivian is a van supplier. To Volkswagen, Rivian is the software company that VW tried and failed to become.

The broader EV market in the United States has turned hostile, with at least a dozen electric vehicle models discontinued, paused, or cancelled in 2026 as 25 per cent import tariffs, the expiration of the federal tax credit, and rising import costs have made selling electric cars in America increasingly uneconomic. Honda wrote off 15.7 billion dollars after cancelling its entire 0 Series. Tesla discontinued the Model S and Model X. Volkswagen’s own US ambitions have been battered: its Scout brand, a revived American nameplate intended to compete with Rivian’s trucks, has been delayed to mid-2028, in part because Rivian’s software was designed exclusively for battery-electric vehicles and VW’s software division CARIAD must now integrate combustion engine controls for a range-extended version. The tariff environment and the Scout delay make the Rivian investment simultaneously more important to VW’s long-term platform strategy and more precarious as a near-term financial bet.

The money

Rivian’s financial position is the controlled demolition that EV startups call a path to profitability. Revenue in the first quarter was approximately 1.4 billion dollars, up 11 per cent year on year. The company achieved its first full year of positive gross profit in 2025, at 144 million dollars, though the automotive segment swung back to a 62 million dollar gross loss in Q1. Full-year guidance calls for deliveries of 62,000 to 67,000 vehicles, an adjusted EBITDA loss of 1.8 to 2.1 billion dollars, and capital expenditures of roughly two billion dollars. The stock trades at approximately 15.43 dollars, down more than 80 per cent from its peak. The company ended the quarter with 4.83 billion dollars in cash, bolstered by the latest billion-dollar Volkswagen payment.

The R2, Rivian’s cheaper SUV intended to reach a broader market, began customer production at the Normal, Illinois, factory on 22 April, after an EF-1 tornado struck the facility without, according to CEO RJ Scaringe, delaying the launch schedule. Initial pricing came in at 57,990 dollars, with the targeted 45,000 dollar entry-level version delayed to 2027. A new factory in Georgia, now designed for 300,000 vehicles per year after a 50 per cent capacity increase, is under construction with production targeted for 2028. The Georgia plant will also build the smaller R3 and up to 50,000 robotaxis for Uber, which struck a 1.25 billion dollar robotaxi deal with Rivian targeting a fleet of up to 50,000 autonomous R2 vehicles across 25 cities by 2031. The DOE loan for the Georgia plant was renegotiated down from 6.57 billion dollars to 4.5 billion dollars after the Trump administration’s review of Biden-era EV commitments, but survived intact.

The architecture

Volkswagen’s restructured software strategy now runs on three parallel architectures: a Global Architecture developed by CARIAD for legacy models, an SDV East architecture from a partnership with Chinese automaker Xpeng for the Asian market, and an SDV West architecture from the Rivian joint venture for Western markets. CARIAD, once positioned as the central software organisation for the entire group, has been reassigned as a coordinator rather than a developer. The admission is remarkable for a company that employs more than 680,000 people and sold 9 million vehicles last year: Volkswagen cannot build the software its future cars need.

Chinese EV manufacturers have built the kind of vertically integrated, software-first platforms that Volkswagen has spent six billion dollars trying to acquire. BYD sold 2.26 million battery-electric vehicles in 2025, overtaking Tesla as the world’s top EV seller. Xiaomi delivered more than 410,000 cars in its first full year of production. Both companies control their own software stacks. Kia, part of the Hyundai Motor Group that also owns Boston Dynamics, has responded by cutting its EV sales target, expanding into hybrids, and planning to deploy Atlas humanoid robots in its Georgia factories from 2028. The legacy automakers are all converging on the same conclusion: the vehicle is becoming a software platform, and the companies that cannot build the software are paying the companies that can.

The bet

Volkswagen’s 15.9 per cent stake in Rivian is not an investment thesis about electric trucks. It is a confession. The largest automaker in Europe, with 88 years of manufacturing history and a software division that absorbed billions before being demoted to a coordination role, has determined that a startup in Normal, Illinois, which has never posted an annual profit and whose stock has lost 80 per cent of its value, has built something VW cannot replicate internally. The zonal architecture that the joint venture is developing will underpin vehicles across the Volkswagen, Audi, Porsche, and Scout brands for the rest of the decade. If it works, VW will have bought the most important component of its future vehicles for less than the cost of a mid-sized acquisition. If it does not, VW will have spent 5.8 billion dollars on a software partnership with a company that is still trying to prove it can sell cars.

Amazon’s quiet dilution from 20 per cent to 11.8 per cent tells the other half of the story. The company that helped create Rivian has decided that the delivery van relationship is sufficient and that increasing its ownership of an unprofitable automaker is not worth the capital. Volkswagen has decided the opposite: that Rivian’s software is worth more than Rivian’s cars, and that the price of building the wrong platform internally is higher than the price of buying the right one from someone else. The shareholder register now reflects that judgment. For the first time since its IPO, Rivian’s most important investor is not the company that buys its vehicles. It is the company that needs its code.



Source link

Leave a Reply

Subscribe to Our Newsletter

Get our latest articles delivered straight to your inbox. No spam, we promise.

Recent Reviews


An analysis of the praxis of human intelligence versus artificial intelligence, exploring embodiment, intentionality, ethics, and meaning in the age of AI.

Conceptual illustration contrasting human intelligence and artificial intelligence, showing a human brain and a robotic AI head representing the praxis of human cognition versus machine computation

Human Intelligence vs. Artificial Intelligence

The rapid evolution of artificial intelligence (AI) has intensified one of the central philosophical and technological questions of the twenty-first century: what distinguishes human intelligence from artificial intelligence in practice? While AI systems demonstrate remarkable capabilities in pattern recognition, optimization, and prediction, their operation differs fundamentally from the embodied, experiential, and purposive nature of human cognition.

The concept of praxis provides a useful framework for exploring this distinction. Originating in Aristotelian philosophy and later developed by thinkers such as Karl Marx and Paulo Freire, praxis refers to the integration of theory and action through reflective practice (Freire, 1970). Human intelligence operates not merely as abstract reasoning but as a dynamic process of perception, judgment, intention, and lived action within the world.

Artificial intelligence, by contrast, functions through computational processes grounded in statistical inference, algorithmic architecture, and large-scale data training. Even the most advanced machine learning systems remain fundamentally different from human cognition because they lack subjective experience, embodied awareness, and existential intentionality.

This essay examines the praxis of human intelligence in contrast with artificial intelligence, focusing on five dimensions: embodiment, intentionality, experiential learning, ethical judgment, and meaning-making. Through this analysis, it becomes clear that while AI can replicate certain cognitive functions, it does not participate in the same praxis-driven structure of intelligence that characterizes human beings.

Understanding Praxis: Action Informed by Conscious Reflection

The term praxis originates from Aristotle’s distinction between theoria (contemplation), poiesis (production), and praxis (action informed by moral and practical reasoning) (Aristotle, trans. 2009). Praxis describes a form of activity in which knowledge is enacted through deliberate engagement with the world.

In contemporary philosophy and social theory, praxis refers to the cyclical process of reflection, action, and transformation. Freire (1970) described praxis as “reflection and action upon the world in order to transform it” (p. 51). Human intelligence unfolds through such iterative engagement with reality.

Human cognition therefore operates within a feedback loop:

  1. Perception of the environment
  2. Interpretation and meaning-making
  3. Intentional action
  4. Reflection on outcomes
  5. Adaptation and learning

This cycle is not merely computational but phenomenological, grounded in subjective experience. Humans perceive the world through senses, emotions, cultural frameworks, and personal histories. These factors shape how knowledge becomes action.

Artificial intelligence, however, operates differently. AI systems do not experience the world; they process representations of it. Their learning occurs through optimization algorithms adjusting statistical weights within models trained on datasets. While this process can mimic aspects of learning, it lacks the reflective and experiential dimensions central to praxis.

Embodiment: Intelligence in the Living Body

Human intelligence is fundamentally embodied. Theories of embodied cognition emphasize that cognition arises from the interaction between brain, body, and environment (Varela, Thompson, & Rosch, 1991). Perception, movement, and sensory feedback form the basis of human understanding.

For example, a photographer tracking a bird in flight relies on a complex integration of sensory perception, motor coordination, anticipatory judgment, and situational awareness. The act is not simply analytical; it is a form of embodied praxis.

The photographer reads the wind, anticipates motion, adjusts posture, and responds dynamically to environmental cues. Experience accumulated over years shapes intuitive responses. Such intelligence emerges through physical engagement with reality.

AI systems, in contrast, are typically disembodied computational entities. Even robotic systems equipped with sensors operate through programmed control architectures and machine learning models rather than lived sensory experience. Their perception is mediated by sensors and interpreted through algorithms rather than consciousness.

Research in robotics and embodied AI attempts to bridge this gap by integrating perception and action systems. However, even advanced robotic agents lack the biological, phenomenological, and experiential dimensions of human embodiment (Clark, 1997).

Thus, while machines can simulate perception-action loops, they do not participate in the same embodied praxis that defines human intelligence.

Intentionality: The Directedness of Human Thought

Another defining characteristic of human intelligence is intentionality, the philosophical concept describing the mind’s capacity to be directed toward objects, goals, or meanings (Brentano, 1874/1995).

Humans act with purpose and intention. Decisions are guided by desires, beliefs, goals, and values. Intentionality shapes how humans interpret information and engage with the world.

Consider the difference between a human writer and a language model. A writer composes text with communicative intention—perhaps to persuade, inform, inspire, or critique. The act of writing is embedded in social, cultural, and personal contexts.

AI language models, by contrast, generate text by predicting probable word sequences based on training data. Their outputs may appear purposeful, yet the system itself possesses no intrinsic intentions or goals. It does not “want” to communicate; it calculates statistical likelihoods.

Philosopher John Searle (1980) famously illustrated this distinction through the Chinese Room argument, suggesting that computational systems may manipulate symbols without understanding their meaning.

Thus, AI can simulate intentional behavior but lacks genuine intentionality. Human intelligence, grounded in subjective consciousness, directs cognition toward meaningful goals and actions.

Experiential Learning and Tacit Knowledge

Human intelligence also develops through experiential learning, a process in which individuals acquire knowledge through direct experience and reflection (Kolb, 1984).

This type of learning often produces tacit knowledge—skills and understandings that are difficult to formalize or encode. For example:

  • A musician sensing subtle timing variations in performance
  • A surgeon adjusting technique during a complex operation
  • A wildlife photographer predicting bird flight patterns

Such expertise develops through repeated interaction with real-world situations. Over time, individuals internalize patterns and responses that operate below the level of conscious analysis.

AI systems learn through data-driven training processes. Machine learning models extract patterns from large datasets by adjusting parameters within mathematical architectures. While this can produce impressive predictive performance, it differs fundamentally from experiential learning.

AI does not possess personal experience, nor does it engage in reflective learning. Its knowledge is derived from statistical correlations within data rather than lived encounters with the world.

Furthermore, AI models often struggle when confronted with novel situations outside their training distribution. Humans, by contrast, can adapt creatively to new contexts because their intelligence is grounded in flexible experiential frameworks.

Ethical Judgment and Moral Agency

Human praxis also includes ethical reflection. Individuals evaluate actions in terms of moral principles, social norms, and personal responsibility.

Ethical judgment involves deliberation about right and wrong, fairness, and the consequences of decisions. Philosophers from Aristotle to Kant have emphasized that moral reasoning is a central component of human rationality (Kant, 1785/1993).

Artificial intelligence systems lack moral agency. They cannot experience responsibility, empathy, or moral concern. Instead, AI operates according to programmed objectives or optimization criteria defined by human designers.

For example, an AI algorithm used in hiring may optimize candidate selection based on patterns in historical data. However, if the data reflects social biases, the algorithm may perpetuate discriminatory outcomes.

Addressing such issues requires human ethical oversight, highlighting the limits of AI in moral decision-making. Machines can assist in analyzing ethical dilemmas, but they cannot independently determine moral principles.

Thus, the praxis of human intelligence includes not only action and reflection but also ethical accountability, a dimension absent from artificial systems.

Meaning-Making and the Human Search for Significance

Perhaps the most profound difference between human intelligence and artificial intelligence lies in the capacity for meaning-making.

Humans interpret experiences within frameworks of culture, identity, and existential reflection. Activities such as art, religion, philosophy, and storytelling arise from the human drive to understand the significance of existence.

Meaning-making involves questions such as:

  • Why does this matter?
  • What does this experience signify?
  • How should I live?

Artificial intelligence does not engage in such inquiries. It processes information but does not seek meaning or purpose.

Existential philosophers such as Jean-Paul Sartre and Martin Heidegger argued that human existence is defined by the capacity to reflect upon one’s being and to shape one’s life through choices (Heidegger, 1927/2010; Sartre, 1943/2007).

This existential dimension forms the deepest layer of human praxis. Intelligence becomes not merely a problem-solving tool but a means of navigating the human condition.

AI systems, lacking consciousness and existential awareness, remain fundamentally outside this domain.

Collaboration Rather Than Replacement

Recognizing these distinctions does not diminish the extraordinary capabilities of artificial intelligence. Instead, it clarifies the complementary roles of human and machine intelligence.

AI excels in areas such as:

  • Large-scale data analysis
  • Pattern recognition
  • Optimization and prediction
  • Automation of repetitive tasks

Human intelligence remains superior in domains involving:

  • Creativity and originality
  • Ethical judgment
  • Contextual interpretation
  • Embodied expertise
  • Meaning-making

The most productive future may therefore lie in human–AI collaboration, where computational systems augment human praxis rather than replace it.

For example, in medicine AI can assist doctors by identifying patterns in medical images or patient data. However, diagnosis and treatment decisions ultimately rely on human judgment informed by empathy, ethical reasoning, and experiential knowledge.

Similarly, in fields such as photography, journalism, and art, AI tools can assist with technical processes, but the creative vision and interpretive meaning remain human contributions.

The Limits of Artificial General Intelligence

Debates about artificial general intelligence (AGI) often assume that sufficiently advanced machines could replicate human intelligence entirely. However, the praxis perspective suggests important limitations to this assumption.

Even if AI systems achieve human-level performance across many cognitive tasks, they may still lack the phenomenological and existential dimensions of intelligence.

Without consciousness, subjective experience, and embodied engagement with the world, artificial systems remain fundamentally different from human agents.

Some researchers propose that consciousness could emerge from sufficiently complex computational systems. Yet this remains a speculative hypothesis with no empirical confirmation.

For now, the evidence suggests that AI represents a powerful form of computational intelligence, not a replacement for the full spectrum of human cognitive praxis.

Conclusion

The comparison between human intelligence and artificial intelligence often focuses on performance metrics: speed, accuracy, or problem-solving ability. However, examining intelligence through the lens of praxis reveals deeper distinctions.

Human intelligence operates as an embodied, intentional, experiential, ethical, and meaning-oriented process. It unfolds through continuous interaction with the world, guided by reflection and shaped by lived experience.

Artificial intelligence, by contrast, functions as a computational system optimized for pattern recognition and prediction. While it can simulate certain aspects of cognition, it lacks the subjective awareness and existential orientation that define human praxis.

The future relationship between humans and AI will likely depend on recognizing these differences. Rather than viewing AI as a replacement for human intelligence, it may be more accurate to understand it as a powerful technological extension of human capabilities.

Ultimately, the praxis of human intelligence remains rooted in consciousness, experience, and meaning—qualities that machines, at least for now, do not possess.

References

Aristotle. (2009). The Nicomachean ethics (W. D. Ross, Trans.). Oxford University Press. (Original work published ca. 350 BCE)

Brentano, F. (1995). Psychology from an empirical standpoint (A. C. Rancurello, D. B. Terrell, & L. L. McAlister, Trans.). Routledge. (Original work published 1874)

Clark, A. (1997). Being there: Putting brain, body, and world together again. MIT Press.

Freire, P. (1970). Pedagogy of the oppressed. Continuum.

Heidegger, M. (2010). Being and time (J. Stambaugh, Trans.). SUNY Press. (Original work published 1927)

Kant, I. (1993). Grounding for the metaphysics of morals (J. W. Ellington, Trans.). Hackett. (Original work published 1785)

Kolb, D. A. (1984). Experiential learning: Experience as the source of learning and development. Prentice-Hall.

Sartre, J.-P. (2007). Being and nothingness. Routledge. (Original work published 1943)

Searle, J. R. (1980). Minds, brains, and programs. Behavioral and Brain Sciences, 3(3), 417–457.

Varela, F. J., Thompson, E., & Rosch, E. (1991). The embodied mind: Cognitive science and human experience. MIT Press.



Source link