Why Amazon bought Globalstar for $11.6 billion

The article highlights rapid AI progress, with industry producing over 90% of notable frontier models in 2025 and U.S.-China model quality described as roughly equal, though the U.S. has 10x more data centers. Key company-specific items include Amazon’s planned $11.6 billion acquisition of Globalstar, Meta and Broadcom expanding their custom chip partnership toward 1 gigawatt of deployment, and IDC reporting a 4.1% decline in Q1 global smartphone shipments, the first drop since 2023. The backdrop is mixed: strong AI and infrastructure investment, but weaker device demand and ongoing supply-chain pressure from memory shortages and trade/geopolitical तनाव.

Analysis

The clearest read-through is that AI capex is shifting from “model race” to “infrastructure moat.” That favors the picks-and-shovels layer with the most control over power, packaging, networking, and custom silicon—especially where buyers are trying to de-risk Nvidia exposure without slowing training/inference deployment. Broadcom looks better positioned than Meta on this tape because the former monetizes the capex race directly while the latter is still in the investment phase; the market is likely rewarding the supplier with visible revenue conversion and punishing the buyer for near-term margin dilution.

The satellite deal is more strategically important than the headline size suggests: it signals that low-Earth-orbit connectivity is becoming a distribution channel for device ecosystems, not just a niche telecom service. If Amazon can make “always connected” a premium feature for mobile devices, the real winner could be downstream e-commerce, ads, mapping, and logistics monetization rather than the satellite operator itself. The near-term constraint is launch cadence, meaning this is a years-long optionality story, but the option value is meaningful if terrestrial connectivity becomes a more salient coverage gap in remote and disaster-prone regions.

Memory scarcity is the underappreciated macro input here. Tight supply should pressure lower-end handset volumes first, but it also creates a second-order benefit for premium ecosystems: affluent consumers and enterprise buyers will trade up instead of out, which helps the most resilient device franchises. The flip side is that broad consumer hardware demand could stay weak into next year even if AI enthusiasm remains strong, so the market may be too complacent about margin support in the component chain.

Consensus is probably underestimating how much of the current AI spend is a land grab for control points, not just compute. That argues for owning the enablers and being selective on the platform owners that must fund the buildout. The weaker names are those with AI exposure but little pricing power or hardware leverage, where “AI narrative” can mask deteriorating unit economics.

Key Decisions for Investors

• Long AVGO vs. short META for 1-3 months: express the view that external AI infrastructure monetizes faster than internal buildouts; target is relative outperformance if capex guidance stays firm, with risk that Meta’s AI monetization narrative re-rates the stock.
• Buy AMZN calls or a call spread into the next 3-6 months: the satellite asset adds strategic optionality to AWS/mobile connectivity, but keep sizing modest because launch execution is a multi-quarter gating factor.
• Short SNAP on any strength over the next 2-4 weeks: the ad market remains concentrated in dominant platforms, and weak consumer/device sentiment plus AI budget pressure should keep smaller ad-supported platforms under earnings stress.
• Long AAPL vs. long hardware beta basket over 3-6 months: premium pricing and ecosystem lock-in should outperform in a memory-constrained environment; risk is that component inflation finally leaks into gross margin or demand.
• Avoid chasing UBER on AI productivity headlines; if anything, wait for a better entry after budget overrun fears compress the multiple, since the market may be extrapolating near-term savings too aggressively.