Rocket Lab Targets SpaceX's Starlink Dominance in New Deal | Bloomberg Tech 6/29/2026

Rocket Lab is pursuing an $8 billion acquisition of Iridium to strengthen its position in the orbital economy and compete more directly with SpaceX. Separately, South Korean firms including Samsung and SK Hynix plan to spend at least $880 billion on chips and data centers to preserve their AI leadership, while Anthropic regained U.S. approval to restore access to part of its Mythos 5 model after security concerns were addressed.

Analysis

The most interesting read-through is not “more AI capex,” but a potential reordering of where bargaining power sits in the stack. If Korean memory and compute ecosystems are locking in a multi-year build cycle, the beneficiaries are not just the obvious chip names but also power, cooling, networking, and construction bottlenecks that convert capex into revenue with less pricing cyclicality. The scale matters because once a sovereign/industrial base commits to this level of spend, it tends to pull forward adjacent infrastructure orders for years, not quarters.

For semis, the second-order effect is that this reinforces a floor under advanced memory utilization and nodes, but it also raises the odds of a supply response that eventually caps margin expansion. The market is likely underestimating how much of this spend becomes “good” for vendors with tight execution and how much leaks to overcapacity in data-center real estate and lower-quality equipment names later in the cycle. In other words, the initial trade is quality-leader multiple support; the later trade is discrimination within the capex chain.

On the M&A/regulatory side, the Rocket Lab–Iridium angle is more important as a signaling event than as a single deal. It suggests the orbitally linked defense/comms complex is moving toward consolidation to secure recurring revenue and control launch-plus-connectivity economics; that is favorable for integrated platforms and unfavorable for pure-play niche providers that lack scale or government relationships. The approval for Anthropic’s model access is a reminder that AI commercialization remains gated by national-security politics, which can create asymmetric upside for firms that can navigate compliance faster than competitors, while creating headline risk for frontier-model peers.

Contrarian view: the consensus will likely overpay for “AI exposure” and underpay for regulated distribution moats. If the capital intensity of AI keeps rising, the best risk-adjusted winners may be infrastructure enablers and incumbents with balance-sheet capacity, not the most visible model names. The main risk is that any export controls, antitrust scrutiny, or permitting delays can turn a 6-12 month buildout story into a 12-24 month digestion phase.

Key Decisions for Investors

• Long SMH / short a basket of lower-quality semiconductor equipment and memory cyclical laggards for 3-6 months: expression is that the capex wave rewards scale and execution first, while second-tier names face margin compression if order timing slips.
• Initiate a pair trade long NBIS or EQIX-style digital infrastructure exposure vs. short a high-multiple, unprofitable AI software basket over 6-9 months: the market is likely overpricing model-layer growth versus picks-and-shovels cash conversion.
• Buy call spreads on RKLB into 6-12 months, funded with upside sales above the announcement zone: the catalyst is consolidation optionality and defense/comms revenue quality, but position sizing should reflect deal-execution risk and financing uncertainty.
• Long defense/comms infrastructure beneficiaries with recurring government exposure; avoid chasing pure launch optimism. Use 3-6 month horizon and prefer names where regulatory approval or classified demand creates switching costs.