Daily Alpha · X

Chamath: falling token costs push value to the application layer and pressure AI-native service firms

Chamath argues the marginal cost of tokens is approaching zero, so selling tokens is not where value accrues; value is moving up the stack to the application layer. He expects AI-native companies that spend heavily on labs in legal, sales and customer service to be hit first, followed by workflow 'connective tissue' apps.

The view shifts AI exposure from model providers toward application owners and profit pools.

Watch Token pricing and spending by AI-native service firms in legal, sales and customer service.

A lot of people pushed back on my take, and I appreciated their feedback.

The most reasonable explanation I heard was that Lumentum’s CEO was never really equipped to take the company to the next level. He previously spent just 1.5 years at FNSR before selling the company. His career has essentially been about growing companies to a certain point and then selling them. By that logic, having driven an 18-fold increase in Lumentum’s share price in just 20 months and now sold the company, Hurlston has done his job. Still, it’s incredibly disappointing to see Lumentum give up on growing independently and choose to sell itself. It feels cowardly. Quoting @jukan05: The only way to interpret Lumentum’s management accepting this deal is that they have no faith in the current AI cycle. They think the market will collapse in two to three years.

$LITE standalone would likely be up on earnings after their new $5B CPO/NPO announcement (separate from $NVDA).

The margins on that are likely 60%+… which shows how profitable laser players would be at scale. (Greensboro alone would be $5B revenue/year) In terms of the $MRVL deal: - Extremely positive for Marvell strategically. Also makes remaining laser players very valuable - $LITE shareholders can cry though about the 1+ year overhang until deal goes through in 2028. There’s less independent price discovery as well since a lot of it will be pegged around 4.5 Marvell shares : 1 - I think Lumentum shareholders probably would be happier if Marvell shares didn’t crash before the deal was announced.. since it was originally valuing $LITE at $115B with a 23.9% premium. - Marvell dropping does show the downside of the deal straight away though, if it rose then $LITE shareholders might be a tad happier given valuation. Having a fixed dollar MC value floor probably would have been better than the…

TheValueist: Anthropic's amended S-1 shows $16.96B of quarterly revenue and $895.4B of commitments

The amended filing shows Q3 revenue of $16.956B, up 46.7% sequentially, with adjusted operating margin falling to 14.5% from 19.3% as short-term compute costs rose. It quantifies about $895.4B of aggregate infrastructure commitments, including $69.8B due in the next 12 months.

The disclosure turns Anthropic's IPO into a test of whether revenue scale can outrun fixed compute obligations and frames the fixed-cost burden investors must underwrite.

Watch IPO roadshow demand, Q4 revenue, and the pace of commitment drawdown.

CoreWeave bonds back below 80 cents as neocloud financing stress builds

Zero Hedge notes CoreWeave bonds are trading back below 80 cents on the dollar, implying yields above 15%.

A sub-80 bond price implies a double-digit yield and signals that bondholders are demanding a large risk premium for CoreWeave's debt.

Watch CoreWeave bond price and yield, and any recovery toward par.