The market is missing the signal. The headline is clear: by 2027, the combined AI capital expenditure of five US tech giants—Alphabet, Amazon, Meta, Microsoft, Oracle—will hit $1.1 trillion. That is 3.2% of US GDP, surpassing the entire defense budget for the first time. A stunning pace, as Kobeissi Letter calls it. But the crypto crowd reads this and shrugs.
They should be terrified.
Liquidity is a zero-sum game. Every dollar that flows into NVIDIA’s GPU clusters, into data center cooling towers, into 800G optical cables, is a dollar that could have flowed into Bitcoin, into ETH staking, into DeFi yields. When a single sector demands $1.1 trillion of capital in three years, it doesn’t conjure that money out of thin air. It pulls it from other asset classes. Crypto will feel the vacuum.
Context: The Five-Headed Hydra
The Kobeissi Letter report is not a prediction—it’s a budget. These five firms have committed to this trajectory in their earnings calls, their supply chain contracts, their multi-year lease agreements with power utilities. The numbers are staggering: 2025 spending at ~2.5% of GDP, 2026 crossing $800 billion, 2027 breaking $1 trillion. The primary driver is not some theoretical belief in AGI. It’s competitive FOMO. No CEO wants to be the one who under-invested in AI while rivals captured the next platform. So they spend.
For the crypto market, this is the macro event that matters more than any ETF flow or halving. Because institutional capital allocation is a portfolio problem. A pension fund or endowment has a fixed risk budget. If AI infrastructure equity becomes the hottest sector—with government backing, visible demand, and a narrative of national security—then crypto’s allocation gets squeezed. I saw this firsthand in 2024 when I helped a Brazilian pension fund structure their first crypto allocation. The conversation was always the same: “We have 5% for alternative assets. Should that go into AI venture capital or Bitcoin?” AI wins every time when the story is “surpassing defense spending.”
Core: The Crypto Capital Flow Mechanics
The $1.1 trillion is not just a headline number. It translates into real effects on crypto’s three primary capital sources: institutional investors, retail speculators, and energy markets.
Institutional Investors: The big allocators—sovereign wealth funds, endowments, insurance companies—are long-term, slow-moving entities. Their total addressable capital for “new” assets is finite. If they see AI infrastructure as a must-own theme with government tailwinds, they will rotate out of existing positions. Crypto, still viewed as a high-risk, unproven asset class, will be the first to be trimmed. I predict that the next 18 months will see net institutional outflows from crypto products (ETFs, funds) as managers rebalance toward AI-linked equities.
Retail Speculators: Retail attention is even more fickle. The current bear market in altcoins has already dampened speculative fervor. The AI narrative is stealing retail oxygen. Every new GPT model, every robotics demo, every data center ribbon-cutting reinforces the idea that “real” innovation is happening in AI, not crypto. Retail traders will chase momentum, and momentum is in AI stocks and semi-conductors. Crypto needs its own breakthrough narrative—perhaps a killer app for DePIN or decentralized AI—but that hasn’t materialized yet. Until it does, AI will cannibalize crypto’s speculative liquidity.
Energy Markets: This is the most overlooked channel. Crypto mining and AI data centers both crave cheap, abundant energy. But the scale of AI demand is orders of magnitude larger. A single training cluster for GPT-5 will consume as much power as a mid-sized city. To meet the 2027 AI capex projections, the US will need to build dozens of new gigawatt-scale renewable or nuclear facilities. This will drive up power prices and lengthen lead times for new connections. Crypto miners, who operate on thin margins, will be priced out of prime locations. The era of “clean” energy mining using stranded renewables is ending—because those renewables will be contracted to AI users at higher rates. Miners will be forced to migrate to less efficient, dirtier power sources, increasing regulatory scrutiny on the entire sector.
But the core insight goes beyond these mechanics. The $1.1 trillion AI capex is a structural decision to centralize intelligence. It builds massive silos of compute in a handful of corporate data centers. This is the opposite of crypto’s ethos of decentralization. And that contradiction is exactly where the opportunity—and the trap—lies.
Contrarian: The Decoupling Trap
The conventional bullish take is that AI and crypto are complements: AI needs decentralized compute, crypto provides it; AI creates demand for verifiable data, oracles provide it. This is “Utility is dead. Long live speculation.” But that narrative is a trap. It assumes that the centralized tech giants will willingly cede control to decentralized alternatives. They won’t. Their entire capex strategy is built on vertical integration—owning the GPUs, the software stack, the customer relationship. They will use crypto only when forced by regulation or when it provides a clear cost advantage. Right now, it doesn’t. Centralized cloud is cheaper, faster, and more reliable for AI workloads.
The real contrarian angle is that AI capex will eventually destroy itself.
History tells us that massive infrastructure investment booms always lead to overcapacity. The dot-com fiber bubble, the 2000s housing bubble, the 2010s shale oil boom—each ended with a sharp correction when supply overwhelmed demand. AI is no different. By 2028 or 2029, the big five will realize they have built far more compute than current applications can monetize. Returns on capital will collapse. Shareholders will revolt. CEO heads will roll.
And that is when crypto benefits. Because the same capital that fled to AI will flee back to alternative assets. Scarce, non-productive assets like Bitcoin—which require no ongoing investment to maintain their value—will look attractive. “Yields are taxes on risk you don,” and during an AI capex bust, the only safe bet will be assets that don’t need constant feeding.
I’ve seen this cycle before. In 2017, I wrote a report warning that 80% of ICO tokens would fail because their tokenomics were unsustainable. In 2020, I profited from DeFi arbitrage but recognized it was liquidity-driven, not utility-driven. In 2021, I shorted NFT ETFs I had analyzed as bubbles. Each time, the crowd chased the hot narrative; each time, the narrative collapsed into a liquidity vacuum. AI capex today is the same—just at 100x the scale. The difference this time is that crypto is more mature, with actual cash flows from DeFi and staking. But those cash flows are dwarfed by AI’s demands.
Takeaway: Positioning for the Cycle
The smart play is not to fight the AI liquidity drain. It is to acknowledge it and wait. Between now and 2027, crypto will be in a secular bear market relative to AI equities. Bitcoin will survive—it always does—but altcoins that rely on speculative trading volume will suffer. The protocols that bleed LPs will be the ones without real yield. The ones that survive will be those that offer genuine utility or are directly tied to AI infrastructure (e.g., decentralized compute networks like Akash or storage like Filecoin, but only if they secure major AI clients).
But the endgame is clear: AI capex is a forced march toward centralization, and centralization breeds fragility. When the correction comes—and it will—crypto will be the safe harbor. The cycle is already priced into the fear. The takeaway: survival matters more than gains. Manage your risk budget. Wait for the yield on AI risk to be taxed. Then deploy.
