Microsoft vs. Meta: Why AI Spending Splits Wall Street

Wall Street delivered a split verdict on Big Tech earnings this week, cheering one AI spending story and penalizing another. The divergence is worth unpacking for anyone tracking the AI investment cycle.

Key points

• Microsoft’s AI capital expenditure was rewarded by markets; Meta’s was punished, despite both companies reporting strong revenue growth.
• Samsung posted a record Q2 profit driven by AI server memory demand, with chip profit reportedly soaring more than 250-fold on AI memory shortages.
• Samsung’s smartphone division swung to its first-ever loss, with surging mobile memory costs (dubbed “RAMageddon”) crushing margins.
• ASML’s EUV lithography machines, priced around $400 million each, remain the critical chokepoint in the AI chip supply chain.
• India’s semiconductor production share is forecast to rise from 0% in 2025 to 13% by 2030, per Forrester.

Why did markets treat Microsoft and Meta so differently?

Both companies are spending heavily on AI infrastructure and both reported strong top-line growth. Yet markets applauded one and punished the other. The divergence comes down to perceived return visibility. Microsoft has been able to point to Azure AI revenue growth as direct evidence that its capital outlays are converting into billable cloud consumption. Investors can trace the spending to a product that customers are paying for today.

Meta’s situation is more complex. Its AI investments are largely directed at infrastructure and model development that supports its core advertising business indirectly, or at longer-horizon bets like augmented reality. When spending scales faster than the market can see a near-term revenue line tied to it, investors get nervous. The lesson for AI stock watchers: the market is no longer giving blanket credit for AI ambition. It wants to see the usage.

Samsung’s record quarter has a sharp asterisk

Samsung’s semiconductor division delivered a blockbuster Q2, with operating profit on the chip side reportedly up more than 250-fold year over year, per Bloomberg figures cited in aggregated coverage. AI server memory, particularly high-bandwidth memory, is the engine. SK Hynix and Micron have been the dominant HBM suppliers, but Samsung’s AI memory business is clearly gaining momentum.

The problem is that memory prices feeding AI demand are simultaneously strangling Samsung’s own mobile unit. Surging DRAM and mobile memory costs have hit the handset division so hard that it posted what appears to be its first-ever quarterly operating loss. Samsung also continues to trail TSMC in its foundry business, though management has pointed to rising demand for advanced process nodes as a sign of gradual improvement. Investors are also watching Chinese competition closely as a longer-run risk to memory pricing.

The internal contradiction, a chip business thriving while the device business that consumes chips is losing money, is an unusual dynamic. It suggests the AI-driven memory shortage is real and durable enough to reprice the whole supply chain, including for Samsung’s own downstream operations.

ASML and the $400 million bottleneck

One structural story running beneath all of this is the role of ASML, whose extreme ultraviolet lithography machines are essential to producing the advanced chips at the heart of the AI buildout. Each machine costs roughly $400 million, and ASML is the sole supplier of EUV tools. TSMC uses them to manufacture chips for Nvidia. Memory makers including SK Hynix, Samsung, and Micron depend on them as well.

That single-supplier position makes ASML one of the more defensible businesses in the AI hardware stack. Any sustained growth in data center construction requires more chip output, which requires more ASML machines. Lead times and export controls remain constraints, which is part of why India’s push toward domestic semiconductor production, projected by Forrester to reach 13% of chip output by 2030, faces a long runway even with strong policy tailwinds.

What to watch

The Microsoft-versus-Meta reaction is a useful template for the rest of earnings season. Markets are increasingly asking whether AI capex is tied to measurable revenue now, or whether it is a bet on future products. Samsung’s results show the memory cycle is running hot, which is good for memory-focused names but is creating cost pain elsewhere in the electronics supply chain. Both dynamics are worth tracking as more cloud and semiconductor companies report in the coming weeks.

This article is for informational purposes only and does not constitute investment advice.