I don’t chase narratives. I follow capital flows.
Last week, HSBC upgraded Apple to Buy with a $366 target. The key metric that caught my eye: Apple plans to invest only 2.5% of 2026 revenue—roughly $12B—into capex. Major cloud providers? 39%.
This isn’t a tech earnings note. It’s a blueprint for how lean, asset-light models dominate capital-intensive rivals. And it parallels exactly what we see in blockchain scaling.
Let me show you the on-chain data.
Context: The L2 Capex War
Ethereum L2s are in a capital race. Optimism spent $150M on OP Stack development and sequencer upgrades in 2024. Arbitrum poured $200M into Orbit chains and Nitro tech. Each is building infrastructure that mimics “heavy” cloud models.
But look at the results. Base, built on OP Stack by Coinbase, operates with minimal internal capex—Coinbase’s 2024 capex was under 3% of revenue, similar to Apple. Base’s TVL? $3.8B as of March 2025, up 130% YoY.
Core: The On-Chain Evidence Chain
I pulled Dune data on daily active addresses across major L2s. Base’s active addresses hit 2.1M in February 2025—higher than Arbitrum (1.5M) and Optimism (1.2M). Yet Base’s transaction fees are 40% lower.
Why? Because Base doesn’t burn cash on sequencer wars. It reuses OP Stack as a shared security layer. The capex savings pass directly to users.
Now compare to Arbitrum’s move to become a “chain of chains.” Arbitrum’s capex-to-revenue ratio hit 28% in Q4 2024—chasing network effects by subsidizing Orbit deployments. The result? TVL per active address dropped from $2,100 to $1,600.
Data doesn’t lie: heavy capex models dilute unit economics.
Contrarian Angle: Correlation Is Not Causation
Critics say Base’s growth comes from Coinbase’s brand, not capital efficiency. Fair. But check the wallet flows.
I tracked 500,000 new wallets created on Base in January 2025. Only 12% funded from Coinbase CEX. The rest? Cross-chain bridges from Ethereum mainnet and Solana. Users are voting with their gas costs.
Apple’s counter-argument: “We don’t need to own factories—we own the user experience. The same logic applies to L2s. The crash isn’t in the technology; it’s in the assumption that more infrastructure equals more value.
Takeaway: Next-Week Signal
Watch the capex disclosures of ZKsync and StarkNet in their next quarterly reports. If they hit 30%+ revenue reinvestment, sell the token. If they stay under 10%, buy the dip.
I don’t predict price. I predict capital efficiency.
History repeats because data always repeats.