The Great AI Rotation: Why Investors Are Abandoning Big Tech for the Picks and Shovels

For two years, the AI trade was simple: buy the biggest tech companies and watch them soar. Microsoft, Meta, Nvidia, Amazon, Alphabet — the so-called Magnificent 7 were the undisputed winners of the artificial intelligence revolution.

That trade is breaking down.

Microsoft has fallen 33% from its recent highs. Meta is down 28%. Tesla, Amazon, Nvidia, and Alphabet are all trading more than 10% below their peaks. Only Apple has held relatively steady, down about 7%. Bitcoin, often a proxy for speculative appetite, has plunged roughly 50% from its October all-time high.

The money hasn't left the market. It has just moved — from the companies spending billions on AI to the companies supplying the physical infrastructure that makes AI possible. It's a rotation as old as investing: when the gold rush stalls, buy the pickaxe makers.

## $725 Billion and Counting

The scale of AI spending has reached a point that would have seemed absurd even a year ago. Alphabet, Amazon, Microsoft, and Meta are expected to spend a combined $725 billion on capital expenditures this year — a 77% increase over last year's record pace, according to recent estimates.

That spending is no longer self-funding. Free cash flow at the big tech companies has been swallowed by data center construction, GPU procurement, and the massive energy costs of running AI infrastructure. To bridge the gap, Alphabet, Amazon, and Meta collectively borrowed $93 billion in the corporate bond market last year — roughly 6% of total issuance.

Meanwhile, share repurchases, a key source of demand for these stocks, have fallen 33% to $132 billion. Companies that once bought back their own shares aggressively are now redirecting that capital into AI infrastructure. For investors, that's a double hit: lower buyback demand and higher spending with uncertain returns.

## Where the Money Is Going

The beneficiaries of this rotation are not household names — and that's the point.

Sandisk (SNDK), a memory-chip maker that most consumers associate with SD cards, has surged roughly 800% this year. The Global X Artificial Intelligence & Technology ETF, which focuses on memory companies essential to AI processing, is up about 140%. Micron Technology (MU), which makes the DRAM chips that AI servers depend on, has gained approximately 230%. The VanEck Semiconductor ETF (SMH) is up 67%.

The pattern is clear: investors are betting not on which AI application will win, but on the infrastructure that all of them require. Memory chips, semiconductor manufacturing, data center real estate, and power generation — these are the picks and shovels of the AI gold rush.

Then there's SpaceX. Elon Musk's space company raised $75 billion in the largest IPO in history last week, and a significant part of the enthusiasm was driven by SpaceX's growing AI capabilities — from satellite networks to autonomous spacecraft to Starlink's potential as an AI infrastructure backbone.

## Why Now?

Several forces are converging to drive the rotation:

**Returns are getting harder to prove.** The big tech companies have spent hundreds of billions on AI, but the revenue payoff remains uncertain. Copilot subscriptions haven't moved the needle at Microsoft. Meta's AI features are impressive but haven't generated clear incremental revenue. Alphabet's search business faces existential questions from AI-powered alternatives.

**Performance is plateauing.** Each new generation of large language models costs more to train but delivers diminishing improvements. The jump from GPT-3 to GPT-4 was transformative; the jump from GPT-4 to the current frontier models feels incremental. Investors are asking whether the next $100 billion in spending will produce proportional returns.

**The infrastructure is the bottleneck.** AI compute demand is growing faster than the supply of chips, power, and data center space. Companies that solve those bottlenecks — memory producers, semiconductor equipment makers, energy providers — have clearer revenue visibility than the companies spending furiously to compete.

**Interest rates aren't coming down.** With inflation proving sticky (see: the Fed and Bank of England both holding steady this week), the cost of capital remains high. High-growth, profitless AI application companies are getting punished while infrastructure companies with real revenue and tangible demand are being rewarded.

## The Risk in the Rotation

Before you reallocate your portfolio, some caveats are in order:

**Infrastructure stocks are getting expensive.** An 800% gain in Sandisk means you're not buying a hidden gem — you're buying a stock that's already priced for perfection. The same applies to many semiconductor names.

**The Magnificent 7 aren't going anywhere.** These companies generate over $1 trillion in annual revenue. A 30% decline from the highs is painful, but it doesn't erase a decade of dominance. At some price, they become value plays again.

**AI spending could accelerate further.** If the next generation of models delivers a breakthrough — true reasoning, reliable autonomous agents, or a killer consumer application — the current spending levels could look cheap in hindsight. The infrastructure beneficiaries would still win, but the big tech companies would ride the wave too.

**Concentration cuts both ways.** The AI infrastructure market is dominated by a handful of companies (Nvidia, TSMC, ASML, Samsung, Micron). A supply chain disruption, regulatory intervention, or competitive shift could upend the thesis quickly.

## What This Means For You

- **Don't chase the rotation blindly.** The move from big tech to infrastructure is real, but many of the winners have already run. Buying Sandisk at 800% gains is not the same as buying it at the bottom. - **Consider the infrastructure ETFs.** If you want exposure to this trend, broad semiconductor and AI infrastructure ETFs (SMH, BOTZ, AIQ) offer diversified access without single-stock risk. - **The Magnificent 7 are on sale — maybe.** A 30%+ decline in Microsoft or Meta looks like a buying opportunity if you believe AI spending will eventually translate into revenue. But the gap between spending and earnings could widen further before it closes. - **Watch the capex numbers.** The next earnings season will be telling. If big tech companies signal they're moderating AI spending, the infrastructure trade could reverse quickly. If they accelerate, both sides of the rotation win. - **Bitcoin's decline is a signal, not an accident.** The 50% drop from highs reflects the same dynamic: capital is rotating out of speculative assets and into tangible infrastructure. Crypto isn't dead, but it's not where the smart money is going right now. - **This is a multi-year structural shift, not a trade.** The AI buildout will take a decade. The question isn't whether infrastructure companies will benefit — it's whether current prices already reflect that benefit. Buy the trend, not the hype.