Software vs Hardware Tech Performance Divergence: Why Microsoft and Meta Are Breaking

The software vs hardware tech performance divergence starts with a split. Microsoft (MSFT) closed at $373.02 on June 30, 2026, 22.5% below its 2026 high and 16.2% below its 200-day moving average. Meta (META) closed at $563.29, 23.6% below its 2026 high and 12.9% below its 200-day moving average, based on Yahoo Finance chart data and Meta chart data checked June 30.

Now look at iShares Semiconductor ETF (SOXX). It closed at $640.76 on June 30, up 104.5% from its first 2026 close and only 2.2% below its 2026 high, according to SOXX chart data. Todd Sohn made the same point on The Real Eisman Playbook: Microsoft looked close to retesting spring lows while the chip trade looked far stronger.

That is the disputed claim. AI is still the story. But the market is separating AI buyers from AI sellers.

Key facts at a glance

• MSFT was 16.2% below its 200-day average on June 30, based on Microsoft chart data.
• META was 12.9% below its 200-day average on June 30, based on Meta chart data.
• SOXX was 71.9% above its 200-day average, based on SOXX chart data.
• Microsoft spent $30.9 billion on property and equipment in the March quarter, while Meta spent $19.0 billion, according to SEC company facts.

The bull case

The bull case is serious. Microsoft, Meta, Alphabet (GOOG) and Oracle (ORCL) are spending because AI infrastructure is now table stakes.

Microsoft has enterprise software, Azure, security, GitHub and AI distribution. Meta has the ad engine and user scale. Alphabet has search, YouTube, cloud and TPU silicon. Oracle has become a cloud infrastructure name, not just a database vendor.

The spending is ugly, but these are not weak businesses. Microsoft reported $82.9 billion of revenue, $30.9 billion of property and equipment purchases, and $46.7 billion of operating cash flow for the March 2026 quarter in Microsoft SEC company facts. Meta reported $56.3 billion of revenue, $19.0 billion of property and equipment purchases, and $32.2 billion of operating cash flow in Meta SEC company facts.

The bulls say the spending becomes a moat. Fine. Then the next test is simple: show margin on the other side.

The bear case

The bear case is about who gets paid first. Nvidia and the chip complex sell picks and shovels. Microsoft and Meta buy them, install them, power them and then have to prove the new capacity creates high-return revenue.

That is a different business problem.

Alphabet is the cleaner counterexample. GOOG closed at $353.33 on June 30, up 12.2% from its first 2026 close and 12.5% above its 200-day moving average, according to Alphabet chart data. Alphabet also spent heavily: $109.9 billion of revenue, $35.7 billion of property and equipment purchases, and $45.8 billion of operating cash flow for Q1 2026, according to Alphabet SEC company facts. But Alphabet has the TPU argument.

Oracle is the knife-twist. ORCL closed at $146.62 on June 30, down 24.6% from its first 2026 close and 40.9% below its 2026 high, according to Oracle chart data. Its FY2026 filing showed $67.4 billion of revenue, $55.7 billion of property and equipment purchases, and $32.0 billion of operating cash flow in Oracle SEC company facts. They spent 83% of revenue on property and equipment purchases. That is not a clean software multiple story. That is a capital cycle story.

What the data shows

The table uses Microsoft chart data, the same price-history method for each ticker, and SEC company facts filed between April 29 and June 22, 2026.

Name	Price signal	Spend signal	What it implies
MSFT	Down 20.8% from first 2026 close, 16.2% below 200-day average	March-quarter equipment purchases were 37% of revenue	Investors want proof that Azure and AI software earn the spend back
META	Down 13.2% from first 2026 close, 12.9% below 200-day average	March-quarter equipment purchases were 34% of revenue	The ad engine is strong, but the capital bill is being marked down
GOOG	Up 12.2% from first 2026 close, 12.5% above 200-day average	March-quarter equipment purchases were 32% of revenue	TPU silicon gives the market a cleaner explanation
ORCL	Down 24.6% from first 2026 close, 26.9% below 200-day average	FY2026 equipment purchases were 83% of revenue	Cloud demand is real, but the stock looks like a financing story
SOXX	Up 104.5% from first 2026 close, 71.9% above 200-day average	ETF proxy for chip suppliers	Suppliers are getting paid before software buyers prove returns

Sohn also said semiconductors had gone from roughly 2% to 18% to 19% of the S&P 500 over ten years in the episode’s SOXX segment. Treat that as his market framing, not audited index math. The audited point is enough: the chip proxy is acting better than the software buyers.

What would change our mind

First, margin proof. If Microsoft or Meta show that AI infrastructure spend is turning into visible operating margin expansion, the bear case weakens fast. Revenue growth alone is not enough.

Second, chart repair. Microsoft back above its 200-day average, Meta reclaiming its January high, and Oracle holding a real base would say the market has stopped punishing software capex.

Third, a semiconductor break. If SOXX loses its trend while Microsoft and Meta stabilize, this stops being a clean buyer-versus-supplier split. Then it becomes a broader AI digestion problem.

What this does not tell you

This does not tell you to buy chips or sell software. It does not tell you that Microsoft or Meta are broken companies. It does not tell you that Alphabet is safe because its chart is better.

It tells you that the market is drawing a line. On one side are companies collecting AI infrastructure revenue. On the other side are companies funding AI infrastructure and asking investors to wait.

The fact is simple: revenue growth is no longer the question. Returns on the spend are.

FAQ

What is the software vs hardware tech performance divergence?

It is the split between AI infrastructure buyers and suppliers. Microsoft, Meta, Alphabet and Oracle pay for data centers. Semiconductor suppliers are closer to the revenue side.

Does this mean Microsoft and Meta are bad businesses?

No. It means the market is questioning the price it should pay while AI spending rises. A strong business can still have a weak chart.

Why does SOXX matter here?

SOXX is a public proxy for the chip side of the AI cycle. Its June 30 trend shows how differently the market is pricing hardware exposure.

What would prove the bears wrong fastest?

Operating margin. If the software names show that AI infrastructure spend is turning into higher-margin revenue, not just bigger sales, the warning loses force.