Peter Schiff, the gold bug and Euro Pacific Capital chief who has spent two decades predicting the next dollar crisis, used his podcast The Peter Schiff Show episode “The Debt, the AI Bubble, and Strategy’s Liquidity Crisis… It’s All Connected” to frame the AI buildout as capital misallocation. His number is blunt. Companies are spending roughly $1 trillion annually on data centers, GPUs, and computing infrastructure, and the gear they are buying, in his words, “is going to be obsolete 5 or 6 years, maybe, maybe quicker.” Highways last fifty years. H100s do not.
The arithmetic of the hyperscalers makes Schiff’s framing concrete. Microsoft (NASDAQ:MSFT | MSFT Price Prediction) reported $30.88 billion in capex for its March quarter, up 84.39% year over year. Alphabet (NASDAQ:GOOGL) spent $35.67 billion, more than double the prior year. Amazon (NASDAQ:AMZN) put up $44.20 billion in a single quarter, an annualized pace near $175 billion, with trailing free cash flow collapsing to $1.2 billion. Meta (NASDAQ:META) raised its 2026 capex guide to $125 to $145 billion, an upward revision mid-year. Stack those four and you are tracking toward $500 billion. Add Oracle, CoreWeave, and the second tier, and Schiff’s trillion-dollar figure stops sounding rhetorical.
The chip vendors capturing the spend
NVIDIA (NASDAQ:NVDA) is the obvious cash register. Q1 FY27 revenue hit $81.62 billion, up 85.23%, with data center networking alone growing 199%. Jensen Huang calls it “the largest infrastructure expansion in human history.” The market prices NVIDIA at a $5.16 trillion market cap. Micron just crossed $1 trillion after UBS lifted its price target from $535 to $1,625. Micron stock is up 225.44% year-to-date and 865.62% over the past year. That move came from HBM bolted to GPUs rather than from a memory cycle.
Schiff’s actual question
Schiff’s argument goes past the bubble label. Plenty of strategists agree valuations are stretched. Vanguard’s 2026 outlook calls AI exuberance an “Economic upside, stock market downside” setup, and Goldman frames the economy as masking weakness through “long-term transformative investments.” Schiff goes deeper. “Where’s this trillion dollars coming from? What would all these companies have done with that trillion dollars if they weren’t using it to buy computer equipment?” His answer points to mass layoffs and foregone investment in other areas, with consumer confidence at an all-time record low running alongside Wall Street celebration.
On-the-ground evidence supports the misallocation thesis. Microsoft reportedly cut Claude Code access in favor of its own GitHub Copilot CLI, with The Verge noting Claude Code had “undermined Microsoft’s new GitHub Copilot CLI coding tool.”
Uber, per The Information, burned its entire 2026 AI budget by April after Claude Code spread across roughly 5,000 engineers faster than finance modeled. CTO Praveen Neppalli Naga said the company was “back to the drawing board on its assumptions.” AI coding tools in real enterprise use are turning out to be expensive in ways original capex models did not contemplate.
What price action is already saying
Microsoft is down 10% year-to-date. Meta is down 2.3%. The two heaviest capex spenders relative to expectations are also the two laggards. Google, where capex is matched by $460 billion in Cloud backlog, is up 22%. Amazon, where AWS grew 28%, is up 21%. The market rewards capex that converts to bookings and punishes capex that does not yet.
The bull rebuttal is direct. Satya Nadella says Microsoft’s AI run rate hit “an annual revenue run rate of $37 billion, up 123% year-over-year.” Polymarket assigns an 80% probability that Microsoft alone is worth more than OpenAI and Anthropic combined at year-end 2026. Reality is somewhere in between. Schiff is one voice, famously skeptical, often early. If AI gear depreciates on his timeline and revenue scales on Nadella and Huang’s timeline, both sides can be partially right. The spread between them is where this cycle will be decided.
Top Gaining Stocks
Top Losing Stocks