prediction markets had their biggest month ever.

gm.

bitcoin kept sliding. zcash got hit by a security bug. crypto headlines were mostly negative.

but prediction markets just had their biggest month ever. and microsoft is now coming directly for openai.

bitcoin in the low $60s. it's drifted lower for several weeks. ETFs bleeding hard, $4.4 billion out over 13 straight days.

the fed meets next week (june 16-17). warsh's first as chair.

oil eased to around $80. iran talks still ongoing.

microsoft launched its own frontier AI models. that's a big shift.

1. prediction markets just had their biggest month ever.

kalshi and polymarket together did $28.4 billion in trading volume in may. that's a record. june 2 alone was an all-time daily high across both platforms.

the timing isn't a coincidence. that same day, crypto liquidations spiked across exchanges. traders rotated to binary contracts, where you can't lose more than what you put in.

quick explainer if you haven't tried these: prediction markets let you bet on real-world outcomes (elections, fed decisions, sports, even "will the odyssey crack 90% on rotten tomatoes"). a $0.40 share pays $1 if the outcome happens, $0 if not. no leverage, no liquidation risk.

kalshi is the bigger of the two now, doing about 60% of total volume. it's federally regulated by the CFTC. polymarket is the more crypto-native option.

2. zcash had its moment. then a bug.

zcash, a privacy-focused crypto, has had a multi-fold run over the past several months. it traded under $50 last summer and has peaked above $600 multiple times since. last week, developers disclosed a critical security vulnerability in their Orchard protocol (the part that handles their private transactions).

the price dropped almost 40% in a few days. emergency patches went out. it's clawed back to around $430 since.quick context: privacy coins let users transact without revealing wallet contents (think encrypted vs. transparent transactions). the rally before this bug was driven by grayscale filing for a US privacy coin ETF and the SEC dropping its investigation in january.

the broader takeaway: even crypto's bright spots run on infrastructure that can break. it doesn't kill the privacy coin thesis. but it's a reminder to be measured about any single coin's bull case.

3. microsoft just declared war on openai.

at microsoft build, the company announced its first in-house AI models for coding and reasoning, an agent system called Scout, a Mac mini competitor, and an android-based operating system designed to run agents instead of apps.

microsoft's AI chief mustafa suleyman, in an interview during the conference:

"the goal is to prove we can become one of the top four labs in the world. there's three labs that matter: google deepmind, openai, and anthropic. we are not one of them at the moment, and that's always been my intention."

context: microsoft was openai's biggest backer for years. after renegotiating that partnership in april, they're no longer content being a distribution channel. they want to be a frontier lab.

the AI race just went from 3 players (openai, anthropic, google) to 4. expect more capital, faster releases, and more pressure on margins as competition heats up.

AI's biggest bottleneck isn't smarter models. it's the physical infrastructure that runs them.

only about a third of the data center capacity planned for 2026 is actually being built right now. the rest is waiting on transformers and electrical equipment with 2-3 year lead times. vertiv and eaton (the picks-and-shovels of power infrastructure) have multi-billion dollar order backlogs.

a layer deeper: AI chips need to talk to each other constantly, and copper wires can't keep up. the industry is switching to fiber and laser-based connections. companies like lumentum, coherent, applied optoelectronics, and corning are riding this wave. lumentum's revenue is up about 90% year over year, and nvidia put $2 billion directly into them in march to lock up laser supply. meta signed a $6 billion fiber deal with corning for its AI data centers.

and a layer deeper still: someone has to physically manufacture all of this. fabrinet (think of them as the foxconn of optical modules) is growing close to 40% year over year. they manufacture for basically every major optical company.

while everyone watches the AI labs, a lot of money is quietly flowing to companies most people have never heard of.

speaking of the AI race: we built goals (our other app) to put it to work on something practical. it links every to do to the goal it supports, captures ideas by voice, and lets AI agents own the follow-up. iphone and mac. weekly briefings included.

see you next week.

not financial advice. aggregated community trends and commentary.