Berkshire Hathaway’s 83% increase in its Alphabet stake, now valued at $38 billion, is the kind of headline that traditionally signals a paradigm shift. For a firm that spent decades avoiding tech like a plague, this move suggests Warren Buffett’s lieutenants have finally capitulated to the AI narrative. But as someone who spent six months auditing the Tezos mainnet launch in 2017—identifying 14 critical vulnerabilities in its consensus mechanism—I’ve learned that the market’s most celebrated moves often mask deeper structural flaws. Truth is immutable, unlike the price action.
I wrote that line in a whitepaper titled “Code is Law, But Only If It Compiles.” It was a response to the ICO euphoria, where promising projects raised millions without a single line of auditable code. Today, I see a similar pattern in the AI frenzy: centralized power wrapped in the language of innovation. Berkshire’s bet on Alphabet is not a bet on technology; it’s a bet on centralization. And for those of us who have spent years building decentralized alternatives, this is both a warning and an opportunity.
Let me step back and provide context. Alphabet owns Google, which dominates AI research, cloud computing, and data aggregation. Its AI models—from Gemini to DeepMind—are trained on a corpus of user data that is effectively a monopoly. The company’s control over the AI stack is staggering: it owns the compute, the data, the distribution, and the talent. Berkshire’s increased stake is a vote of confidence in this vertical integration. But the question we must ask is not whether Alphabet will profit from AI, but whether the structure of AI itself is becoming a threat to the very principles of decentralization that blockchain was built to protect.
I recall the 2020 DeFi Summer vividly. I founded OpenLedger Lab, a non-profit that mentored 50 junior developers from underrepresented backgrounds. We deployed their first ERC-20 tokens, and I wrote a guide on “Democratic Governance in DAOs” that was downloaded 15,000 times. The guiding philosophy was that financial sovereignty is a human right. That same philosophy applies to AI. If AI is controlled by a handful of corporations, then sovereignty itself is at risk. Truth is immutable, unlike the price action.
Now, let’s dive into the core technical analysis. The AI industry today relies on a few centralized providers for compute, storage, and model training. Amazon Web Services (AWS) and Google Cloud together control over 60% of the cloud infrastructure, while Nvidia dominates GPU supply. This concentration mirrors the oracle problem in DeFi, where a single point of failure can cascade into systemic collapse. I’ve seen this firsthand: in 2020, I audited a DeFi protocol that relied on a single Chainlink node for price feeds. The node went down for 12 minutes, and the protocol lost $4 million in liquidations. The same logic applies to AI. If Alphabet’s servers fail, or if its model alignment is compromised, the consequences are global.
But the problem is deeper than hardware. It’s about data sovereignty. Alphabet’s AI models are trained on user data without explicit consent, often through opaque terms of service. This is not a bug; it’s a feature of centralized architecture. In contrast, blockchain-based AI projects like Bittensor and Render are building decentralized networks where compute and data are distributed, and users retain control. Bittensor, for instance, uses a subnet structure that allows anyone to contribute computational power and earn rewards. Render tokenizes GPU cycles, enabling a peer-to-peer rendering network. These are not just technical alternatives; they are ethical imperatives.
I experienced this tension during the 2022 bear market, when the Terra-Luna collapse shattered my idealization of algorithmic stability. I retreated to a cabin in rural Virginia for six weeks, disconnected from all devices. In that solitude, I drafted the manuscript for “The Soul of Sovereignty,” a book arguing that blockchain must serve human dignity, not just capital efficiency. That experience taught me that the most dangerous illusions are the ones we embrace collectively. Berkshire’s bet on Alphabet is one such illusion: it assumes that centralized AI can be benevolent, that the invisible hand of the market will align incentives. But we know from blockchain history that code is law, and law without enforcement is just a suggestion.
Let me offer a contrarian angle. Many in the crypto space will see Berkshire’s move as a validation of AI’s potential, and they will rush to invest in anything AI-related. But I see a different story. Berkshire’s increased stake is a sign of institutional capture. The same forces that co-opted the 2024 Bitcoin ETF—turning a decentralized asset into a Wall Street product—are now co-opting AI. The ETF approval, which I criticized in an op-ed titled “Institutionalization vs. Ideology,” showed that 95% of custody providers were centralized third parties. The same pattern is emerging in AI: venture capital is pouring into centralized models, while decentralized alternatives struggle for visibility.
But here’s the truth: the decentralized AI ecosystem is more resilient than it appears. In 2025, I launched a “Human-Centric AI” initiative, collaborating with three key ethicists to draft the “Decentralized Trust Protocol.” This protocol uses zero-knowledge proofs to verify AI decisions without exposing sensitive data. It was cited by two EU regulatory bodies. The core insight is that zero-knowledge proofs can be the bridge between AI’s need for data and blockchain’s demand for privacy. This is not theoretical; it’s being implemented today. Projects like zkSync and Polygon are exploring zk-proofs for AI inference, allowing users to verify that a model’s output is correct without revealing the underlying data.
I’ve been told that this is too complex for the average reader. But I disagree. The average user understands that a centralized AI can be manipulated, censored, or hijacked. They understand that if Alphabet controls the oracle, they control the truth. The blockchain community has spent years building systems that are trustless, transparent, and immutable. The same principles apply to AI. Truth is immutable, unlike the price action.
Now, let’s address the market context. We are in a bear market, and survival matters more than gains. Over the past 7 days, several AI-focused protocols have lost 30-40% of their liquidity as capital flees to safe havens. But this is precisely the moment to build. The 2017 ICO crash taught me that the projects that survive are the ones with real utility, not hype. The same will hold for AI. Protocols that are bleeding dry are the ones that lack a clear value proposition. Those that integrate zero-knowledge proofs, decentralized compute, and user governance will emerge stronger.
I’ve embedded my experiences throughout this article because they are the foundation of my analysis. The 2024 ETF fallout taught me that institutional adoption is a double-edged sword. The 2022 bear market taught me that solitude breeds clarity. The 2020 DeFi summer taught me that community is the ultimate validator. And the 2017 ICO craze taught me that code is law, but only if it compiles. These are not just lessons; they are the lens through which I view Berkshire’s move.
Let me be clear: I am not arguing that Berkshire is wrong to invest in Alphabet. I am arguing that the narrative surrounding this investment is dangerously narrow. It ignores the systemic risks of centralized AI, the ethical imperative of data sovereignty, and the potential of decentralized alternatives. The contrarian view is not that AI is overhyped; it’s that the current structure of AI is unsustainable. The next bubble will not be a DeFi protocol or a meme coin; it will be a centralized AI company that fails to align with human values.
We are already seeing the cracks. Alphabet’s AI models have been criticized for bias, hallucination, and lack of transparency. The company’s handling of user data has led to multiple lawsuits. Meanwhile, decentralized AI projects are gaining traction. Bittensor’s market cap, though volatile, has grown 300% in the last year. Render has expanded beyond GPU rendering to AI compute. And a new generation of protocols, like Gensyn and Together, are building decentralized training networks.
The key metric to watch is not price, but distribution of compute. Centralized AI relies on a few data centers. Decentralized AI distributes compute across thousands of nodes. The latter is more resilient, more democratic, and more aligned with the original vision of the internet. I’ve seen this model work in DeFi, where Uniswap’s decentralized exchange handles billions in volume without a central operator. The same can happen for AI.
I’ll end with a forward-looking thought. The path to true AI sovereignty is not through Berkshire’s portfolio, but through the on-chain infrastructure we are building today. The decentralized trust protocol I helped draft is just one example. There are hundreds of teams working on zk-proofs, federated learning, and tokenized compute. The question is not whether they will succeed, but whether we will let them. The institutions are betting on centralized control. We are betting on distributed empowerment. Truth is immutable, unlike the price action.
So, as you read this, consider what Berkshire’s $38 billion bet really means. It means that the old guard is waking up to AI. But it also means that they are doubling down on the same centralized model that gave us the 2008 financial crisis, the 2022 Terra collapse, and the 2024 ETF custody mess. The blockchain community has a choice: embrace this institutional validation and risk co-option, or build the alternatives that make centralization obsolete. I know which path I’ve chosen. I hope you do too.
