obsidian capital

prefshares Verified account @PrefShares · Nov 2

The $ANT S-1/A gives us some insights into a restructured $AMZN deal, and the potential headwind that posed to AWS growth this past quarter.

It's possible that AWS & Anthropic restructured a deal this past quarter and that it partially explains the Q3 AWS miss (39% vs. the ~42% the buyside was looking for)... and the Anthropic S-1/A actually allows us to size this.

The filing says S&M as a % of revenue fell in Q3 partly due to "changes in the terms of our channel partner distribution arrangements." So something changed.

Anthropic books CSP marketplace sales gross and runs the partner's cut through S&M (that cut is AWS/Google/MSFT revenue).

Piecing it together:

(i) Share of revenue sold through CSP marketplaces was 48% in 1H26 and 44% for 9M26, which implies ~40% in Q3, or ~$6.8B.

(ii) S&M was $2.4B in Q3. Partner fees were ~$351M in 2025 ($43M of 2024 S&M less $33M of advertising, plus a reported $341M increase in partner fees), and the filing says they rose $4.8B Y/Y over 9M 2026. That leaves ~$530M of non-partner marketing over 9M 2026, or ~$207M in Q3 with a modest step-up through the year. This implies Q3 partner fees were ~$2.15B.

(iii) That puts the partners' cut at ~35% in Q2 vs ~32% in Q3. ~3.7 pts on ~$6.8B = ~$250M of fees partners didn't collect.

The $250M is across all CSPs, and the filing doesn't say which partner's terms changed (the $110B AWS compute agreement isn't described as amended; Google's & MSFT's are).

If AWS absorbed ~75-100% of it, that's ~55-75 bps of Y/Y growth.

So something was restructured, but the headwind looks quite manageable. I think it explains a piece of the miss, not the whole thing (as shown here, I think capacity timing is the larger culprit: x.com/PrefShares/sta…).

Table attached.

View quoted post on X

Blueprintsmb Verified account @blueprintsmb22 · Oct 31

Ah fak the podboi acceleration/deceleration crap! I'm a real fundamental investor! Okay, you can hate the podboi brain rot, but it works as accel/decel controls narratives. You could have traded GOOGL well based on acceleration/deceleration solely based on comps.

GOOGL is going to be in the penalty box for the foreseeable future into tough comps after seeing Search growth deceleration from high teens to low teens yet the 2 year stacks have held constant. Are the search bears going to be around if GOOGL is able to get back to high teens growth in H2 2027 while free cash flow inflects positive? Who knows. If Search can't re-accelerate meaningfully in H2 2027, GOOGL will certainly be in trouble

View quoted post on X