gm.
midweek catch-up. five things that moved since monday.
the 5
1. meta wants to sell you AI computing power.
bloomberg reported yesterday that meta is quietly building a cloud business to sell AI infrastructure. the initiative is called "meta compute."
what they'd sell: (a) access to their own AI models (called muse spark), and (b) raw GPU capacity to companies that need to train or run their own AI. that's a direct shot at amazon web services and google cloud.
meta hasn't confirmed, but zuckerberg hinted at it on the may earnings call. the play makes sense: meta has spent hundreds of billions on chips and data centers for their own AI ambitions and hasn't fully monetized it yet. selling the excess would help recoup the spend.
market reaction was sharp. meta stock jumped 11%. specialized cloud computing companies (coreweave, nebius) sold off double-digits, presumably worried about meta undercutting them.
2. anthropic's fable 5 is back.
three weeks after the US government forced anthropic to disable fable 5 (their most powerful consumer AI model), it's back globally as of yesterday.
quick recap of the arc:
june 9: anthropic launches fable 5 and its cybersecurity twin, mythos 5
june 12: government forces anthropic to disable both. reason: amazon researchers showed fable 5 could be tricked into finding software security bugs
june 26: mythos 5 gets approved for about 100 US companies operating critical infrastructure
june 30: export controls fully lifted
july 1: fable 5 back for everyone
what changed: anthropic added a new safety filter that blocks the specific technique amazon exposed. blocked requests get automatically routed to a slightly less powerful model. commerce department signed off. anthropic also announced a joint safety framework with amazon, microsoft, and google.
paid claude users get fable 5 back at full capacity through july 7, then it moves to a credit system.
3. rocket lab bought iridium for $8 billion.
on monday, rocket lab agreed to acquire iridium at $54 per share (a 24% premium to friday's close). cash-and-stock deal, closes mid-2027.
quick primer on what these companies do:
rocket lab is a US space company. they build small rockets (electron) and satellite components. they're the second-most-launched US rocket company after spacex. they're also developing a bigger reusable rocket called neutron.
iridium operates the world's only truly global satellite phone network. 66 satellites in low earth orbit that provide communications for ships, planes, mining sites, the US military, and hikers- anywhere cell service doesn't reach.
the strategic angle: rocket lab wants to build a legitimate alternative to spacex's starlink. combining a launch business (rocket lab) with an operational satellite constellation (iridium) gives them the pieces to compete.
both stocks jumped on the news. RKLB up 16% and IRDM up 25% on monday.
Privacy-first email. Built for real protection.
End-to-end encrypted, ad-free, and open-source. Proton Mail protects your inbox with zero data tracking.
4. robinhood launched its own blockchain.
on wednesday, robinhood launched robinhood chain at an event in london.
what it is (in plain english): a blockchain is a shared digital ledger where transactions can be recorded and verified without a central authority. robinhood built their own on top of ethereum so they can offer:
tokenized US stocks and ETFs. users in 120+ countries can trade digital versions of apple, tesla, and other US stocks 24/7 (not limited to US market hours).
decentralized finance (defi). lending, borrowing, and other financial services running on-chain.
they also launched robinhood earn, a product that lets users lend their dollar stablecoin (USDG) and earn about 7% annually.
the bigger picture: robinhood is expanding from "crypto trading app" to building actual crypto infrastructure. HOOD stock rose about 5% on the news.
5. meta almost bought kalshi before building arena.
NPR reported tuesday that mark zuckerberg personally met with kalshi CEO tarek mansour in 2025 to explore acquiring the prediction markets company. the talks were exploratory, not a formal offer.
this is a direct follow-up to monday's story about meta building their own prediction markets app (arena). so meta considered buying kalshi first, then decided to build a competitor instead.
two theories on why the talks fell apart:
mansour wasn't a willing seller
meta decided kalshi's regulatory battles were too messy to inherit
context: kalshi is now valued at $22 billion (up from $2 billion a year ago). prediction markets did about $220 billion in monthly volume this june, up from $28 billion in june 2025. huge growth.
meta and kalshi did announce a threads partnership earlier this year. now meta is building their own competitor.
see you monday.
not financial advice. aggregated community trends and commentary.