Lot of people using @anthropic news re: Fable to rekindle AI FUD... just read something intellectually shallow about $CRWV bonds somehow being the new 2008 CMBS hiding in plain sight... $CRWV is not some new subprime. Three things gave rise to the conditions in 2008 and none of them are here. 1/ Side bets. The GFC was about the trillions in side bets / synthetic leverage that created a notional exposure many times larger than the underlying mortgages themselves. $CRWV has none of that. No side bets, no bets on the bets. The most anyone loses is what was actually lent. 2/ Who actually pays. Subprime ran on loans to people who could never pay them back. The "safe" $CRWV loan is bolted to $META and the riskier loans are stamped JUNK... that is public information and they are priced like JUNK. The weak credit here is not hidden inside something safe. It is labeled for the whole world to see... you either believe it's a risky loan or you don't. 3/ The ratings trick. In 2008 the credit rating agencies said: spread the mortgages across enough cities and they cannot all go bad at once, so it must be AAA. That assumption in MBS-land broke everything. The one FAIR worry IMO ... AI is a single bet. If demand cracks, the chips lose value, the customers stop paying .. blah blah everyone crammed in the AI trade is fucked. That correlation is real. But with no side bets and no fake AAA on top, that is just a risky bet on one industry, priced IN THE OPEN. That is not subprime. Hate the stock, short it, fine. Just stop pretending the 2008 wiring is somehow in $CRWV bonds. It's not... it's just a concentrated bet. Fear mongering is lazy. Don't be lazy.

This is the first time the USG has forcibly pulled a deployed frontier commercial model. That's a precedent with direct implications for every frontier lab and it lands days before @AnthropicAI was to kick off public-market preparations.

$LITE CEO Hurlston

Anthropic just poached the guy who was heading OpenAi’s custom chip with $AVGO

Added to existing $ALAB $BRUN $NOK $MRVL

Sarah Friar mentions the new chip that OpenAI is spinning up with $AVGO Speaking at @theallinpod conference youtube.com/watch?v=TjrShuj_…

Sold my remaining $SIVE shares this am. IMO the trade was ride the exuberance and sell into earnings. Institutions currently using US retail as exit liquidity.

What happens to space-levered names if this is true? Whitney Tilson: "I think SpaceX will fall 70% from its opening-day close. It's shaping up to be one of the most overhyped, overvalued large-cap IPOs of all time. The numbers are absurd... SpaceX looks set to trade near 100x revenue, with 15% top-line growth last quarter and accelerating losses. When things are this obvious, making a "big call" like this is easy..."

I wrote up my full thoughts on $SIVE / $SIVEF The TLDR: > I think their gross margin target is great > I think their AI Data Center Photonics pipeline opportunity is very low, and has to increase meaningfully > I think they need to be very timely in their CPO readiness > I think they are at the right place at the right time t and more I also lay out what I think is fair value today, what price I would be interested in, and what I can underwrite for price over the next 12 to 24 months you can read it here : open.substack.com/pub/cruxca…

$SIVE earnings were a miss. Net sales fell 22% to SEK 61.9m, adjusted EBITDA was −13.8m and loss widened from −6.0m. That’s why the stock is down. Management blamed the US government shutdown, defense budget delays, and FX, and leaned on a 77% pipeline jump to $799M with 2027 ramps “on track.” Translation: zero current-quarter validation. The bull case is now entirely forward-looking TAM talk and future margins with no timeline.

Ayar Labs and Wiwynn are partnering to bring CPO to rack-scale AI systems that support next-generation hyperscale AI workloads. By combining Ayar Labs’ CPO solution with Wiwynn’s rack-level system design and manufacturing capabilities, the two companies are enabling a new class of rack-scale AI infrastructure.

Trimming $SIVE into earnings

Strong prints on $COHR and $LITE got sold two weeks ago. Today $NVDA guides $91B for Q2 and confirms Rubin shipping Q3. The optical buildout is real and accelerating. Thesis just got stronger while the stocks got cheaper. I expect both names trade higher.

Retail is just asking for it 🤷

At Sivers Semiconductors, we develop both #photonics and #wireless technologies for connectivity and sensing across #AI data centers, #5G, and #SATCOM applications, with a focus on performance, and efficiency. Explore the Sivers portfolio 👉 sivers-semiconductors.com

$BOT is not like $DRAM. This is publicly-listed venture fund with an annual expense of 3.22% / 5x what $DRAM cost (.65%). Much of the portfolio is held via third party SPVs not direct. Those SPVs likely charge a variety of management / carried interest fees that will impact the value of $BOT returns. There is also a massive resale overhang. There are almost 20m shares registered for resale by insiders who bought at $10... they can dump without notice.

BREAKING: US data center construction spending jumped 34% YoY in March, to a record $50 billion annualized rate. Spending on data centers is up 437% since the beginning of 2021, when the annualized rate stood at ~$9 billion. This is also up 688% since the start of 2018, when the annualized rate was just ~$6 billion. Meanwhile, office building construction spending fell -9% YoY in March, to $46 billion, the lowest since 2015. This means that spending on data centers now exceeds office building construction by $4 billion, or 9%. To put this into perspective, office construction spending exceeded data center spending by $65 billion, or 650%, in 2020. AI is fundamentally transforming the US economy.