The Intel Capex Playbook for Crypto: Why Ethereum L2s Are the Overlooked Infrastructure Bet

Goldman Sachs is bullish on Japanese semiconductor equipment stocks. They see Intel’s 30-billion-dollar capex increase as a catalyst for Lasertec, Tokyo Electron, and Disco. The logic is clean: higher spending on advanced nodes means more orders for the tools that make those nodes possible. But Goldman is missing the bigger story. The same capital-expenditure-driven thesis applies directly to Ethereum’s Layer 2 ecosystem—and most crypto investors are asleep at the wheel. The semiconductor playbook is simple. Identify the structural shift: Intel’s IDM 2.0 strategy, AI-driven demand, and regional onshoring. Then find the bottleneck suppliers—companies with monopoly or near-monopoly positions in critical process steps. For Intel, those are Lasertec (EUV mask inspection), Disco (precision dicing for chiplet packaging), and Tokyo Electron (etch/deposition for advanced nodes). The market rewards them with premium multiples because they capture a disproportionate share of the capex dollar. Now map that onto crypto. The structural shift is Ethereum’s rollup-centric roadmap, accelerated by the Dencun upgrade and EIP-4844. Billions of dollars in value are migrating from L1 to L2. The bottleneck suppliers are not the L2s themselves—they are the infrastructure providers that every rollup must use: data availability layers, sequencer networks, and proving systems. These are the Lasertecs and Discos of crypto. Their revenue scales with L2 activity, not with speculative token price. Let me walk through the seven-dimensional framework I used to dissect Goldman’s semiconductor thesis. Apply it to crypto, and the parallels are striking. Technology Architecture: Rollups vs. Monolithic Chains Goldman’s semiconductor analysis focuses on transistor architecture—GAA, FinFET, backside power delivery. In crypto, the equivalent is rollup design: optimistic vs. zk-rollups, EVM-equivalence vs. custom VMs. The winners are not necessarily the most innovative L2s but the ones that provide the best infrastructure for others to build on. Arbitrum’s Arbitrum One uses optimistic fraud proofs; zkSync employs zero-knowledge proofs. The critical technology bottleneck? Data availability. Without a robust DA layer, no rollup can scale securely. This is where the capex analogy sharpens. Intel’s 18A node requires High-NA EUV lithography from ASML and mask inspection from Lasertec. Ethereum’s blob space requires data availability sampling—a service provided by chains like Celestia and Avail, or by the Ethereum consensus layer itself. The DA layer is the High-NA EUV of crypto. It is the non-negotiable component for scaling. And just like Lasertec holds an 85% share in EUV mask inspection, Celestia holds a dominant mindshare in modular DA. Its token capture is directly tied to the total blob capacity consumed by L2s. Supply Chain Handcuffs: Sequencers and Provers In semiconductors, the supply chain is concentrated. Intel cannot easily swap out Tokyo Electron for Applied Materials because the process integration is years deep. Similarly, rollups cannot easily switch sequencers or provers without forking their entire state management. The sequencer is the bottleneck—it orders transactions and determines MEV distribution. Equity in sequencer infrastructure (like Espresso or Astria) is the equivalent of owning shares in Lasertec. The more L2s launch, the more demand for shared sequencers. Goldman’s analysis highlighted customer concentration as a risk for Japanese equipment makers. In crypto, the risk is even starker. L2s are heavily reliant on Ethereum as their settlement layer. If Ethereum’s blob space becomes congested, L2 fees spike. But that dependency also creates a moat. No L2 can fully escape Ethereum without losing security guarantees. The supply chain is locked. Capital Expenditure: The Blob Market vs. The Node Buildout Intel is spending 30 billion extra to build fabs. Ethereum does not have a central capex committee, but the network’s capital expenditure is happening in aggregate across the L2 ecosystem. Each rollup deploys its own set of validators, state machines, and bridge contracts. The total economic cost of L2 infrastructure—hardware, development, sequencer operation—is conservatively in the hundreds of millions per year and growing exponentially as blobs fill up. The key metric is blob utilization. In February 2024, before Dencun, blob space was zero. One year later, blobs regularly exceed 80% of capacity. That is the equivalent of Intel’s fab utilization rate. When fabs are full, equipment orders surge. When blobs are full, demand for DA and proving infrastructure surges. The play is to own the suppliers of that infrastructure before the capacity crunch hits. Market Demand: AI Inference on L2s Goldman’s bull case for Intel hinges on AI demand for HPC chips. The same trend is driving crypto demand. AI agents are already using L2s for micropayments and data attestation. A single AI model query on-chain can generate dozens of L2 transactions. This is the killer use case for block space. And just as AI pulled forward Intel’s capex cycle, it is pulling forward L2 capacity expansion. The infrastructure that scales with AI agent activity—not human retail trading—will be the long-term winner. Disco benefits from chiplet packaging for AI chips. The equivalent in crypto is the proving market for zk-rollups. Every zero-knowledge proof requires computational resources. As proof generation gets cheaper (via zkASICs and FPGA accelerators), the volume of proofs will explode. Companies like =nil; Foundation and Hype are building zk-proving infrastructure. They are the Disco of crypto: precision tools for a critical step that every rollup needs, but no rollup builds in-house. Geopolitics: Regulation as the Onshoring Driver Goldman’s semiconductor analysis correctly flagged that regional onshoring (CHIPS Act) pumps orders for Japanese equipment suppliers. In crypto, regulation is the onshoring equivalent. The MiCA framework in Europe and spot ETF approvals in the US are driving institutional capital into compliant infrastructure. Regulated staking providers (like Coinbase Custody) and compliant L2s (like those with KYC bridges) will capture the bulk of this inflow. But here is the contrarian twist. The most regulation-friendly L2 is also the most infrastructure-dependent: Base. It uses OP Stack, Celestia for DA (via Alt-DA), and Coinbase for sequencer. Base’s success is not purely its own; it is a reflection of the modular stack it rents. That makes Base’s growth a passive tailwind for the underlying suppliers. The same way Intel’s fab expansion benefits Lasertec regardless of whether Intel wins the CPU war. Competitive Landscape: The Tokyo Electron of Crypto Tokyo Electron holds strong positions in etch and deposition but faces fierce competition from Applied Materials and Lam Research. In crypto, the equivalent is the general-purpose L1—Solana, Sui, Aptos. They compete head-to-head for developer mindshare and liquidity. But the infrastructure suppliers—DA layers, shared sequencers, proof markets—have much wider moats. Celestia faces competition from EigenDA and Avail, but the modular narrative is still Celestia’s to lose. Its market cap is a fraction of what a monopoly equipment supplier would trade for in traditional markets. Goldman’s analysis pointed out that Lasertec and Disco have stronger investment theses than Tokyo Electron because their monopolies are more defensible. The same is true in crypto. L2s are disposable; the infrastructure they depend on is sticky. The best risk-adjusted play is not on any single L2 token. It is on the infrastructure providers that capture value from aggregate L2 growth. Financials and Valuation: The Premium Multiples Lasertec trades at 45–50x earnings. Market participants accept that premium because they price in years of monopoly-driven growth. Comparable crypto infrastructure tokens—TIA, EIGEN, and others—trade at much lower earnings multiples (if they have earnings at all). The disconnect exists because institutional capital has not yet rotated into this thesis. But the macro winds are shifting. As L2 revenue grows, these tokens will start to show real cash flows. When they do, the multiple expansion will be violent. Goldman’s recommendation on Lasertec at 70,000 yen was a call on multiple expansion justified by capex visibility. The same logic applies to TIA at current prices. The risk is that Intel’s capex plan fails. The risk in crypto is that L2 usage sputters or that Solana captures the entire rollup mindshare. Both are real. But the infrastructure suppliers have a hedge: even if Ethereum loses market share, other chains will still need DA and proving. Disco does not care if Intel wins or loses; it only cares that Intel spends. Similarly, Celestia does not care if Arbitrum beats Base; it only cares that both use blobs. Contrarian Angle: The Decoupling Thesis The market believes that crypto infrastructure tokens are correlated with ETH. That is true in the short term. In the structural cycle, they decouple. When the bull market matures, investors rotate from beta (ETH) into alpha (specific infrastructure plays). This is the same decoupling that happened in semiconductors between Intel (the end-product) and Japan’s equipment makers (the picks-and-shovels). Intel stock underperformed Lasertec by 3x over the past five years. The same dynamic will play out in crypto. Betting on the L2s directly is betting on individual execution. Betting on the infrastructure is betting on the entire sector’s growth. That is the contrarian edge. Most liquidity chases L1 tokens. The real alpha is in modular infrastructure that works across all rollups. The market has not yet priced in the scale of capex that L2s will deploy over the next 18 months. When blob utilization hits 95%, the bidding war for sequencer capacity will make today’s fees look quaint. Takeaway: Cycle Positioning Intel’s capex surge is a signal, not a trade. The signal is that the bottleneck suppliers in any technological scaling story capture disproportionate value. In crypto, the bottleneck is shifting from L1 throughput to L2 composability and data availability. The infrastructure layer is the new Japanese equipment sector—and it is still early. Position for the blob wars. Accumulate exposure to modular DA and proving infrastructure. Monitor blob utilization and L2 revenue as leading indicators. When institutions start writing “Lasertec of crypto” narratives, the multiples will already be repriced. The ledger does not sleep, but the analyst must. Yield is a lie; liquidity is the truth. And right now, liquidity is flowing into the picks and shovels of the rollup era.