Amazon's Satellite and AI Ambitions Converge in a $200 Billion Infrastructure Push

Amazon eyes Globalstar acquisition to boost Project Kuiper, while a leaked OpenAI memo reveals overwhelming enterprise demand for its AWS AI partnership, validating Amazon's massive infrastructure bets.

Amazon's stock, having climbed roughly 35% from its April low, finds itself at the intersection of two major strategic moves this week. The e-commerce and cloud giant is reportedly in advanced talks to acquire satellite operator Globalstar for an estimated $9 to $10 billion. Simultaneously, a leaked internal memo from OpenAI reveals overwhelming enterprise demand for its new partnership with Amazon Web Services (AWS), validating the company's colossal infrastructure investments.

The potential acquisition of Globalstar addresses a pressing strategic need. Amazon's own satellite internet project, Project Kuiper, has only 241 production satellites in orbit, a stark contrast to the over 10,000 operated by SpaceX's Starlink. Regulatory pressure adds urgency: the FCC requires Amazon to have approximately 1,600 operational satellites by July 2026—a deadline the company views as unrealistic and is seeking to push to 2028. A Globalstar deal would instantly provide critical licensed spectrum and infrastructure that is difficult to build organically. One significant hurdle remains: Apple owns a 20% stake in Globalstar and has secured 85% of its network capacity for emergency features on the iPhone, meaning any transaction requires direct negotiations between Amazon and Apple.

On the cloud front, a newly surfaced internal memo from OpenAI provides a powerful endorsement of AWS's strategy. Denise Dresser, OpenAI's new head of sales, described incoming customer demand for the AWS partnership as "overwhelming," according to CNBC. The partnership, announced in late February, is architecturally distinct: while Microsoft's Azure remains the exclusive provider for stateless APIs, AWS secures the rights to sell stateful runtime environments where AI models retain memory and context across ongoing workflows.

This demand is translating into tangible financial metrics. In his latest shareholder letter, CEO Andy Jassy revealed that AWS's AI business is now running at an annualized revenue rate of over $15 billion, based on Q1 performance. This represents roughly 10% of AWS's total $142 billion in annual revenue. The company's custom silicon division is growing even faster. Combined revenue from Graviton, Trainium, and Inferentia chips has surpassed a $20 billion annualized run rate, doubling from the figure at the start of the year. Jassy hinted these chips may eventually be sold externally, positioning Amazon as a direct competitor to Nvidia and AMD.

Analysts suspect Amazon's in-house chips are already powering leading AI models, with Evercore analysts noting that Anthropic's recently unveiled "Mythos" model was likely trained on Amazon's Trainium chips. By funding and providing infrastructure for both leading independent AI labs—OpenAI and Anthropic—AWS establishes itself as a fundamental base layer, regardless of which model ultimately prevails commercially.

The financial scale backing this dual-pronged strategy is staggering. Amazon has outlined capital expenditures of $200 billion for 2026, primarily for AI infrastructure. A portion of this spend is already contractually secured; OpenAI alone has committed to over $100 billion as part of its partnership with AWS. This investment has impacted free cash flow, which declined significantly in 2025. The market has responded positively, with shares gaining nearly 11% in the past week to trade at 203.55 euros on Monday.

All eyes now turn to Amazon's first-quarter 2026 earnings report on April 29th. The company has guided for an operating income between $16.5 and $21.5 billion. AWS, which grew 20% in 2025 to reach $129 billion in annual revenue, faces heightened expectations. Investment bank Citi has already raised its estimates, forecasting AWS revenue growth of 28% year-over-year, driven directly by the ramping partnerships with Anthropic and OpenAI. The upcoming results will be a critical test of whether Amazon's massive bets in space and silicon can sustain their current momentum.