© Tapati Rinchumrus / Shutterstock.com
The Invesco QQQ Trust (NASDAQ:QQQ | QQQ Price Prediction) has ripped higher in the spring rally, climbing 10% in the past month and 19% year to date to trade near $727. That rebound from the early-April volatility spike, when the VIX briefly punched above 25, has been powered almost entirely by the same handful of AI infrastructure names that dominate QQQ’s index weight. Investors holding QQQ today own a fund where the top five positions account for roughly 37% of net assets, and where one company alone, NVIDIA (NASDAQ:NVDA), carries a roughly 10% weight.
That concentration is the entire story for the next 12 months. The fund’s 0.20% expense ratio and Nasdaq-100 wrapper are almost beside the point when six holdings drive the outcome.
The macro factor that matters: AI capex follow-through
Forget the Fed for a moment. The federal funds upper bound sits at 3.75%, and the 10-year Treasury yield at 4.56% is roughly where it has lived all year. The swing factor is whether hyperscaler capital expenditure budgets keep climbing through 2026.
The numbers explain why. Microsoft just spent $30.88 billion on capex in a single quarter, up 84% year over year. Alphabet ran capex up 107%, Amazon hit $44.2 billion, and Meta Platforms raised its 2026 guide to $125 to $145 billion. That spending is the revenue line for NVIDIA, whose Data Center segment grew 92% last quarter. Combined, these five names sit at over 29% of QQQ.
What to watch concretely: each quarter’s hyperscaler earnings call, specifically the capex guide and the Microsoft commercial RPO line (currently $627 billion, up 99%) and Google Cloud backlog ($460 billion). If two of the four hyperscalers trim 2026 capex guidance on the same reporting cycle, the AI trade unwinds, and QQQ would absorb most of that move. A useful early tell is GPU rental pricing, the sort of data point Reddit’s stockmarket community has already started flagging, with one widely-read post this week titled “GPU Rental Prices Notably Decline through 2nd half of May; H200 -38%”. Falling rental prices would point to supply finally catching demand, which historically precedes capex digestion. Check those guides quarterly, with the next round in late July.
The fund-specific factor: circular AI exposure
QQQ’s bigger structural risk is that its top holdings are not independent bets. NVIDIA sells chips to Microsoft, Amazon, Alphabet, and Meta. Those four buy the chips with cash generated partly by AI services that depend on those same chips. QQQ owns all of them. A capex deceleration at the hyperscalers also hits NVDA’s growth rate, and a slip in NVDA’s 75% gross margin would compress the single largest holding by index weight.
What to monitor: the quarterly Nasdaq-100 rebalance and the NVDA weight specifically. NVDA at roughly 10% is approaching levels that historically prompted the index’s special rebalance mechanism, which capped a single stock’s growth contribution back in 2023. If the methodology forces a trim, QQQ buyers get less NVDA exposure mechanically, regardless of how the stock performs. Invesco publishes holdings daily on its fund page.
Investors who want the same AI thesis without the concentration can look at equal-weight Nasdaq products, which dilute the top names to roughly 1% positions each. Those vehicles underperform sharply when NVDA leads, and outperform when leadership broadens.
The bottom line
Watch the July hyperscaler earnings round for any cut to 2026 capex guidance, and watch the next Nasdaq-100 special rebalance for any forced trim of NVDA’s weight. Those two events, more than the Fed, will determine whether QQQ’s 43% one-year run extends or stalls.
