The numbers arriving in my inbox this week carried a familiar weight. A 44-year-old who has watched three market cycles in crypto โ who helped design governance frameworks for DAOs managing hundreds of millions, who negotiated clauses ensuring institutional capital funded open-source infrastructure โ recognizes certain patterns regardless of the sector. When I read that Cohere was in advanced talks to raise up to $3 billion, I didn't see an AI story. I saw a replay of every bull market narrative I have witnessed since 2017: capital flooding into a transformative technology, valuations untethered from fundamentals, and the quiet voices of dissent getting drowned in the euphoria.
This is not an article about Cohere specifically. I have not audited their smart contracts โ because they likely have none. This is about what the AI funding surge reveals about the investment landscape that blockchain builders now operate within, and the lessons that remain stubbornly relevant despite the sector switch.
The Capital Has Moved, But the Architecture Remains
Let me be precise about what we know. Cohere, the enterprise AI company founded by Transformer paper co-author Aidan Gomez, is reportedly negotiating a funding round that could reach $3 billion. To put this in context: the entire blockchain industry raised approximately $8 billion in 2023 across all categories โ DeFi, infrastructure, gaming, and everything else. A single AI company is reportedly seeking nearly 40% of that sum in one round.
The comparison is not meant to diminish AI's potential. It is meant to calibrate our understanding of where institutional capital is flowing and why. In my experience advising an Australian pension fund on crypto allocation, I learned that large institutions move in herds, and that herd behavior creates both opportunity and hazard for those operating at the frontier. The $3 billion figure tells us something specific: the institutional FOMO that drove Bitcoin ETF approvals has metastasized into something broader. Capital allocators are no longer asking whether AI will transform enterprise productivity. They are asking which horse to back.
This matters for blockchain because the narratives are not mutually exclusive. When I designed quadratic voting systems for Community DAO in 2020, we were wrestling with questions that enterprise AI companies are now confronting at scale: How do you prevent whale dominance? How do you align incentives across diverse stakeholders? How do you ensure that the people who contribute value capture value? These are governance questions, and they are domain-agnostic.
What the DeFi Reckoning Taught Me About Valuation Gravity
I spent three months in withdrawal after the Community DAO treasury drain in 2020. Fifty thousand dollars lost to a signature replay attack โ not because the code was technically flawed in a novel way, but because we had designed governance structures that assumed good faith actors would not exploit edge cases. The trauma recalibrated my relationship with narrative. I began to see how stories โ about token valuations, about network effects, about the inevitability of adoption โ could detach from underlying economic reality and sustain themselves through pure momentum.
The AI funding landscape is exhibiting similar symptoms. A $3 billion raise for Cohere would imply, if we apply typical venture dilution of 10-20%, a post-money valuation somewhere between $15 billion and $30 billion. For context, Cohere's valuation in mid-2024 was approximately $5.5 billion. That is a 3x to 5x jump in valuation based on a single funding event, with no disclosed revenue multiples, no confirmed burn rate, and no clarity on whether the $3 billion represents equity, convertible notes, or strategic commitments in kind.
The silence around terms is itself informative. When I refused to sign off on EtherTrust's smart contract in 2017 because of reentrancy vulnerabilities, the founders called me a blocker. But the code did not lie. The patterns in the code โ the assumptions baked into the logic โ revealed exactly what the founders would not admit: that they were building for exit, not for permanence. I suspect the same pattern recognition applies to funding rounds that arrive with maximum headline value and minimum disclosure. The opacity is not accidental.
The Institutional Mirror and the Blindness It Reveals
When I negotiated the clause ensuring 5% of the pension fund's crypto allocation would flow to open-source infrastructure, traditionalists thought I had lost my bearings. In their mental model, allocation meant yield. Yield meant return. Return meant the familiar instruments they had priced for decades. The notion that capital should serve infrastructure โ that the plumbing matters as much as the product โ seemed unorthodox.
I see the same blind spot crystallizing in AI investment. The $3 billion flowing to Cohere is not flowing to open-weight model research, to interpretability tooling, or to the governance frameworks that would make AI systems accountable to the people they affect. It is flowing to enterprise sales teams, to cloud compute commitments, to the apparatus of commercial deployment. This is not inherently wrong โ someone has to build the bridges between laboratory capability and business workflow. But when 90% of the capital is chasing the same deployment layer, the infrastructure layer atrophies.
This is the lesson from my work with indigenous Australian artists on their NFT project. We raised $150,000, and I faced intense pressure to flip the assets for quick profit. I resisted because I understood something that the speculators could not see: the cultural integrity was the product. When you strip the story to isolate the asset, you destroy the thing that made it valuable. In AI investment, I see the same dynamic in reverse. Capital is so focused on the deployment layer โ the enterprise contracts, the API call volumes, the ARR multiples โ that it cannot see the infrastructure dependencies underneath. The models are trained on data scraped without consent. The infrastructure runs on GPUs controlled by a single company. The governance is whatever the founding team decides, accountable to no one except their investors.
The Contrarian Position That Isn't Really Contrarian
Here is the uncomfortable truth that I have learned to speak plainly: the blockchain industry, for all its failures and frauds and spectacular implosions, built something that AI has not yet replicated. The idea that infrastructure should be open, that protocols should be auditable, that governance should be contestable โ these are not utopian fantasies. They are the lessons of repeated failure in systems that concentrated control in the name of efficiency.
When rollups post blob data to Ethereum, when DeFi protocols lock value in smart contracts, when DAOs attempt to coordinate without hierarchical authority โ these are experiments in a particular theory of organizational design. The theory holds that transparency reduces agency costs, that open-source development produces better security outcomes, and that distributed decision-making is more resilient than centralized command. These are empirical claims, and the evidence is mixed. But they are serious claims, subjected to real-world stress tests in which billions of dollars were at stake.
The AI funding surge is happening almost entirely outside this framework. The models are proprietary. The training data is opaque. The governance is traditional venture-backed hierarchy. The infrastructure dependencies are hidden behind cloud service agreements. And yet the same investors who would demand smart contract audits and governance token distributions in a blockchain deal are writing $3 billion checks to AI companies without equivalent scrutiny.
This is not a critique of AI as a technology. It is an observation about the differential standards applied across sectors, and the consequences those standards produce. When I write about DeFi interest rate models being arbitrary, I am not suggesting that traditional finance does better. I am suggesting that the scrutiny applied to blockchain โ however inconsistent and however often captured by hype โ at least creates the conditions for accountability. You can read the code. You can trace the transactions. You can identify the concentrated wallets.
Can you do the same for an AI model making decisions about credit, employment, or medical treatment? The answer today is no. And the $3 billion flowing into Cohere is not flowing toward changing that reality.
The Blob Data Saturation Problem and Its AI Parallel
I have written previously about my expectation that post-Dancun blob data will saturate within two years, forcing rollup gas fees to increase again. The analogy to AI infrastructure capacity is direct. When compute demand outpaces supply โ when the number of enterprises seeking to deploy AI models exceeds the available GPU capacity and inference optimization โ the price of deployment rises until either supply expands (requiring massive capital expenditure) or demand contracts (requiring price sensitivity that enterprise buyers have not yet developed).
A $3 billion funding round for a single AI company is, among other things, a bet that this dynamic will resolve favorably. The bettor is assuming that compute supply will expand fast enough, that inference costs will fall fast enough, and that enterprise adoption will generate enough revenue to justify the valuation premium before the next model cycle renders the current capability obsolete.
This is not an irrational bet. It might be correct. But it is a bet on execution in an environment where the fundamental inputs โ compute availability, model capability, enterprise budget allocation โ are all volatile. And it is being made with institutional capital that has, historically, been remarkably bad at pricing technology transition risk.
Forward
The question I hold โ and the question I would put to any builder navigating this moment โ is not whether AI or blockchain will win. The question is whether the infrastructure being built will be accountable to the people it affects. The $3 billion flowing to Cohere is capital that could, in principle, fund open-weight research, interpretability tooling, or decentralized inference markets. It will instead fund enterprise sales teams and cloud compute commitments, producing returns for investors while leaving the governance questions unaddressed.
I have spent twenty-eight years watching technology cycles. I have watched capital flow into ICOs, into DeFi summer, into NFTs, and now into AI. The specific technologies change. The human patterns of narrative, momentum, and blind spot remain constant. The builders who survive are not the ones who predict the cycle correctly. They are the ones who build infrastructure that outlasts the cycle โ who make choices in the bull market that do not require amnesia to defend in the bear.
For blockchain builders watching the AI funding surge: this is not a distraction. It is a mirror. The capital is moving somewhere. The question is whether the infrastructure you are building will be there when it arrives.
The answer depends, as it always has, on what you chose to build when no one was watching closely enough to make you explain yourself.