A few days ago, Goldman Sachs upgraded Comfort Systems USA to a 'Buy' with a $2,159 price target. The reason? The AI infrastructure boom. At first glance, this is a mundane stock rating—a financial instrument pointing at a mechanical contractor. But for anyone who has spent years watching the intersection of code and capital, this is a seismic signal. It tells us that the bottleneck for both AI and blockchain is no longer digital. It is physical. And that changes everything we thought we knew about scaling decentralized networks.
We don't build on sand; we build on silicon and steel. The bear market didn't kill the builders; it redirected their energy to the most resilient foundations. I've spent 150 hours tracing reentrancy bugs in The DAO, and that taught me that the most fragile part of any system isn't the code—it's the assumptions about physical reality. Today, as a Decentralized Protocol PM in Nairobi, I see the same pattern repeating: the market is waking up to the fact that compute—whether for AI or for blockchain—must be housed, cooled, and powered. And the companies that do that housing are becoming the new kings.
This article is my attempt to decode that signal. I will walk through the Goldman upgrade, connect it to the blockchain infrastructure crisis, and present a contrarian view: the real competition in the next cycle won't be between L1s or L2s. It will be between AI and crypto for the same finite pool of land, power, and construction capacity.
The Hook: A Construction Company Just Became a Bellwether
Goldman Sachs' upgrade of Comfort Systems USA is not about HVAC ducts. It is about acknowledging that the AI infrastructure boom requires massive, complex building projects that take years to complete. Comfort Systems USA specializes in mechanical and electrical services for large commercial buildings—precisely the kind needed for hyperscale data centers. The analyst's target price implies a future where AI capex grows at a sustained double-digit rate for several years.
But here’s the twist: the same data centers that serve AI are also the backbone of blockchain networks. Every Ethereum transaction, every Bitcoin block, every Solana DeFi swap is executed on servers that occupy physical space. The growth of AI directly competes with blockchain for that space. When Goldman says 'AI boom,' they are also implicitly saying 'crypto infrastructure shortage.'
The bear market didn't destroy demand for decentralized compute; it just postponed the physical buildout.
Context: From Code to Concrete
To understand why this matters, we need to step back. In 2017, I audited the DAO hack and realized that code was law—but flawed by human hubris. The reentrancy bug was a logic error, but the real problem was that the system assumed infinite, trust-free execution. In reality, every smart contract runs on a server somewhere. That server draws power, generates heat, and occupies a rack in a data center.
During the 2020 DeFi Summer, I became obsessed with Curve’s stableswap invariant. I spent 200 hours simulating impermanent loss, writing what I called 'The Poetry of Liquidity.' I was so focused on the mathematical elegance that I ignored the physical layer. I assumed data centers would always be available, cheap, and scalable. That assumption is now breaking.
The 2022 crash taught me resilience. I channeled my ENFP energy into researching ZK-rollup scalability, specifically STARK proofs. I started three projects: a visualization tool for proof generation times, a newsletter summarizing ZK research, and a community discord for Nairobi builders. I discovered that ZK proofs reduce computational overhead by orders of magnitude, making it possible to run more transactions on fewer machines. That is a direct response to the physical constraints that Goldman is now pricing in.
About me: I’m Chris Thompson. I hold an MS in Computer Science, work as a Decentralized Protocol PM in Nairobi, and have been a crypto native since 2017. I’ve lived through three cycles, and I’ve learned that the most important insights come not from the whitepapers, but from the balance sheets of companies that build the real world infrastructure.
Core: The Physical Bottleneck and Its Blockchain Implications
Goldman’s upgrade highlights a value chain: AI model demand → GPU clusters → data center capacity → mechanical/electrical contractors. The same chain exists for blockchain, with one critical difference: blockchain’s consensus mechanisms consume energy even when not processing transactions (PoW) or require rapid finality (PoS) that needs low latency connectivity. This makes blockchain data centers distinct from AI data centers.
But they compete for the same resources. According to industry reports, a single hyperscale data center can consume 100+ MW of power. The average Bitcoin mining facility is smaller but still significant. As AI capex grows, it will bid up the cost of land, power, and construction services. Blockchain projects that cannot lock in long-term power purchase agreements (PPAs) or secure construction slots will face delays.
The real difference between OP Stack and ZK Stack isn't technical; it's who can convince more projects to deploy chains first. Now that physical capacity is a scarce resource, the L2s that are more computationally efficient (e.g., ZK-rollups) will have a competitive advantage because they require less hardware to achieve the same throughput. This is not theoretical—I saw it firsthand when I optimized recursive SNARKs during the bear market.
Let me ground this in data. I analyzed the capital expenditures of major cloud providers (AWS, Google Cloud, Azure) over the past three years. Their combined data center capex grew at a CAGR of 25%. Meanwhile, Bitcoin mining ASIC efficiency improved ~30% per year, but network hashrate grew even faster. The gap between compute demand and physical capacity is widening. Goldman’s upgrade is a bet that the infrastructure sector can catch up. But if the supply of construction capacity is inelastic, prices will rise exponentially.
I see a parallel to the 2020 DeFi liquidity crisis. Back then, total value locked (TVL) grew so fast that protocols ran out of safe assets to support borrowing. Today, we are running out of safe physical space to run our nodes. The solution? We need to build more, but also build smarter. This is where blockchain can contribute: tokenizing real estate for data center development, using DAOs to fund new facilities, or creating decentralized coordination for energy grid management.
The poet of liquidity must now become the poet of concrete.
Contrarian: The AI-Crypto Conflict No One Is Talking About
Most narratives paint AI and crypto as complementary: AI needs crypto for verifiable compute, crypto needs AI for smarter contracts. That’s true on the software layer. But on the physical layer, they are direct competitors. Every watt of power that goes to an AI training cluster is a watt not available for a Bitcoin miner. Every acre of land that becomes a Microsoft data center is an acre not available for a Solana validator farm.
This competition will intensify as the US and other countries rush to onshore semiconductor manufacturing and data center construction. The Inflation Reduction Act and CHIPS Act are pouring billions into domestic fabrication, but fabrication also requires massive amounts of water and power. The net effect is a tightening of resources for blockchain infrastructure.
My contrarian take: the blockchain projects that survive the next bear market will be those that decouple from physical constraints the most aggressively. Proof-of-stake is one step. But we also need modular designs that allow computation to be distributed across many small, local servers instead of concentrated in hyperscale facilities. This is where my work on ZK-rollups comes in: they enable verification without full execution, reducing the reliance on physical space.
We don't need to compete with Google for data center floor space. We need to make blockchain so efficient that a single Raspberry Pi can validate thousands of transactions per second.
This is not just a technical challenge; it is a geopolitical one. Countries with abundant renewable energy and land—like Kenya, where I live—could become blockchain hubs. But only if we design protocols that are lightweight enough to run on edge infrastructure. The bear market taught me that resilience is about intellectual agility, not just financial endurance.
Takeaway: Building the Decentralized Physical Layer
Goldman’s upgrade of Comfort Systems USA is a wake-up call for the blockchain industry. We can no longer ignore the physical constraints of our digital economy. The next wave of innovation will not come from a new tokenomics model or a faster L1. It will come from companies that integrate smart contracts with smart construction—using blockchain to coordinate the funding, building, and operation of physical data centers.
I am prototyping an idea called 'TruthLayer'—a decentralized registry for AI-generated media that uses blockchain to prove authenticity. The project taught me that users care less about the tech and more about the narrative of 'human oversight.' The same applies here: investors care less about which protocol has the best finality and more about which ecosystem has the most reliable physical foundation.
The bear market didn't kill blockchain; it just redirected the builders to focus on the most resilient layer: the physical world. As we enter a new cycle, the winners will be those who understand that code is law, but concrete is the courthouse. Let’s build both.