Goldman Sachs Partner: 11 Stages of a Tech Bubble — and the Market Is in One
N.R. Finch
Goldman Sachs macro trader Bobby Molavi maps the current market onto the 11-stage arc of a tech bubble — semiconductors now account for ~19% of S&P 500 market cap, more than double the dot-com peak, while AI capex is reshaping the market's pricing anchor on a scale last seen during the railroad era.
Semis at 19% of the S&P 500 — how extreme is that?
Semiconductor stocks now represent roughly 19% of total S&P 500 market cap — a record, and more than double the concentration at the peak of the 2000 dot-com bubble.
Goldman's momentum pair strategy — a trade that bets "winners keep winning, losers keep losing" — closed Friday at an all-time high, helping the S&P 500 post gains in 11 of the past 12 weeks.
This means → the market isn't just rising; it's rising in an extremely concentrated, self-reinforcing loop — capital is piling into the same direction.
How much money is AI capex actually going to burn?
The market expects hyperscale cloud providers (Microsoft, Amazon, Google and peers) to spend between $2 trillion and $3 trillion on AI capex over the next three years.
In plain terms = before the AI era, these companies' entire cumulative spend was $1.5 trillion — they're set to outspend their own history in just three years.
Consensus forecasts put 2026 spending at $757 billion and 2027 at $920 billion; if investment scales to 2–3% of GDP — comparable to railroad- and automobile-era infrastructure buildouts — 2027 capex approaches $1.1 trillion, up roughly 45% year-on-year.
Everyone is making the same bet — where is the risk?
Molavi describes a dangerous "paradigm": active and passive strategies, fundamental hedge funds and systematic quant funds, institutions and retail — all adding AI exposure simultaneously.
He cites one investor: "Micron was a $100 billion company last year; now it's $1.1 trillion — if it resets, it won't land comfortably at $700 or $800 billion."
This means → when everyone crowds onto the same boat, the boat doesn't tip gently — it either keeps rising or it capsizes. There is no graceful middle.
The VIX is only 17 — so why is volatility actually elevated?
Implied volatility on core holdings — SK Hynix, SpaceX, Marvell, Micron, Kioxia — has exceeded 100, while the VIX (a gauge of broad-market fear) sits at just 17.28.
In plain terms = the headline index looks calm, but the handful of stocks actually driving returns are churning violently beneath the surface.
This reflects a structural mismatch: the market's apparent "calm" masks extreme swings in the assets that matter most — investors may be severely underpricing actual risk.
What did the final stage of past bubbles look like?
Molavi draws a parallel to the late dot-com era: in the 6–12 months before the final crash, the market absorbed multiple 5% drawdowns, each time snapping back via mean reversion.
Until one 5% dip failed to bounce — and cascaded into a genuine systemic collapse.
His blunt assessment: "If a 10% decline is breached, whether there's any floor beneath it — nobody has an answer right now."
Can this trade achieve a soft landing?
No one wants to exit early — because leaving is a guaranteed miss on returns, while staying is only a "maybe" risk.
But the market leans ever more heavily on one theme (AI) and one factor (momentum) — and that concentration is the core variable that determines whether a soft landing is even possible.
This means → the danger of a bubble isn't that it must burst; it's that when everyone bets "this time is different," the exit door narrows to almost nothing.
Content is for reference only, not financial advice.