1. AI worries hit chipmakers.

Chip stocks in Asia and the United States dropped after memory chip giant Samsung Electronics' blowout earnings report failed to satisfy investors with sky-high expectations.
Samsung Electronics tanked 10% as investors looked beyond the South Korean electronics giant's forecast of a 19-fold jump in second-quarter earnings, with results failing to alleviate concerns about the sustainability of the AI-driven chip boom.
"The story of today is the story of the last few weeks, and that's rotation after the blistering run in the AI buildout, semis and memory. Expectations have gotten to be almost impossible to beat for these companies."
AI demand is still booming, but the market is starting to ask whether memory pricing, hyperscaler capex, and hardware margins can keep compounding from here. Morgan Stanley warned Monday that semi weakness could continue as investors brace for more capex discipline from the cloud giants.
Adding to worries about high-flying chipmakers, Reuters reported that Chinese startup DeepSeek is developing its own AI chip, a push that could reduce its dependence on Nvidia.

2. Barclays upgrades Legrand to “Buy” on data centre demand.

Barclays says investors are overlooking Legrand's shift toward higher-growth products and are too bearish on the impact of 800V DC architecture changes.
Price target goes to €185.
Barclays also stays "overweight" peer Schneider Electric on strong end-market demand.
Of the 21 analysts covering Legrand, 11 rate it "strong buy" or "buy," eight "hold," and two "sell" — LSEG data.

3. What is orbital compute?

These are compute servers in space with solar arrays kept in a sun synchronous orbit and a cooling radiator behind it constantly in the shade. It is a virtual data centre where racks are linked using lasers travelling through vacuum, which are technologies widely used in modern satellites.
Why put data centers in space? Nearly half of today’s cost of building a 1GW AI data center is spent on infrastructure, rather than compute. As reusable launch continues to drive costs lower, orbital platforms could replace a significant portion of this terrestrial infrastructure with solar-powered systems in space, potentially bringing orbital compute economics closer to terrestrial AI infrastructure over time.

Below: Orbital compute roadmap

4. 20 banks participated in the SpaceX IPO and their average target is 47% above yesterday’s close.

“SpaceX combines near-monopoly launch economics, the world’s largest LEO satellite network, and a fast-scaling AI infrastructure business. The company is one of the few platforms that can link real estate in orbit, global connectivity, and compute capacity into one infrastructure stack.”
Broker models see revenue rising from $45bn in 2026 to $319bn in 2030 and $3.3tn in 2040, with the largest upside tied to Starship, Starlink capacity, terrestrial compute, and orbital compute.
Four KPIs are key to the stock: revenue per watt, cost per watt, cost per kg to orbit, and Starlink subscribers/connected "nodes".
These map to four debates: enterprise AI monetization, compute cost and time-to-power, Starship economics, and broad Starlink TAM adoption. Near-term catalysts include Starship Flight 13 and Flight 14, first operational Starship payloads, additional neocloud deals, Starlink broadband and mobile updates, government contracts, and new enterprise connectivity agreements.

Below: Each SpaceX segment is interconnected

5. SpaceX or Spacecake.

“We forecast no FCF-positive year before 2035 and average external capital needs of roughly $84bn per year from 2027 to 2034. If debt markets cannot absorb this financing need, SpaceX may need to issue equity, reduce growth investment, or slow deployment.”

Below: China is the biggest threat to replicate a Starlink-like constellation. Similar to Huawei/Xiaomi vs. Apple in consumer electronics, Chinese constellations may be less of a competitive threat in the US or Western Europe, but can be serious competitors in the developing and rest-of-world which represents a much larger overall TAM. A large % of projected Starlink growth is in developing markets (South America, Africa, Eastern Europe, etc.) which are less likely to reject China-based satellite internet services for geopolitical reasons.

Not a subscriber yet?