Why is the $PLTR price action lower? The market wants more than a strong beat from Palantir and this one seems pretty standard for them. Their own growth outlook is also slower for next quarter. This could be lower expectations so the next earnings look better. Palantir’s weak spot is new deals. They grew 6% YoY while sales grew 96%. Their backlog of signed contracts also didn’t grow from last quarter. This means growth will only keep up if they can increase these new deals. Sales are still strong and almost all of their sales growth is coming from the US. For the first time they make more money from US companies than from the US government.

On the call $PLTR said customers are signing shorter contracts, which seems to be their reason why the new deals look small. Palantir expects these customers to sign longer deals over time as they see more results. So this could explain why the backlog didn’t grow. The backlog doesn’t count deals shorter than a year. If you include those shorter deals their backlog (which also includes contract options) grew 2%. One thing is on the past 2 calls they gave out the length adjusted deal number. I don’t see it here and that would answer whether the new deals look small because demand is slowing or because the contracts got shorter.

$PLTR is growing very fast. For the first time they make more money from US companies than from the US government. Most of their growth is turning into profit because their costs are growing slower than their sales. They expect strong sales to continue especially in the US. https://t.co/5CN2cU2Yee

$ANT Anthropic’s margins are going to get worse based on the amended S-1 released today. The Q3 operating loss which wasn't in the original release comes from expensive short term compute and a donation matching program. To keep growing fast they need to pay extra to keep up with demand. They are paying more to get compute while tokens get cheaper. If we assume demand is unlimited they and many AI related companies are still limited by power, data centers, chips and memory. But they have billions in commitments they need to pay. These fixed payments mean they need to keep growing, and to grow they need all this compute. Even if we ignore the AI bubble talk I am struggling to see how they can even meet all this demand.

$ANT Anthropic has to pay hundreds of billions for compute whether they grow or not. To cover the compute they have already committed to, they need to keep growing fast. To keep growing they need that compute to actually come online. That compute means chips, the chips need data centers, and the data centers need a ton more power. But there isn’t enough grid power where it’s needed and new connections can take years. The data centers also need money to get built. Lenders are getting more selective and charging some borrowers more, which could slow how fast new data centers get built. If power and financing delay those data centers, Anthropic ends up short of the compute they need and that could slow their growth. But they are locked into paying for that compute for years. If growth slows, those commitments are still there so there’s less revenue than they were counting on and more of it goes to paying for compute. And from my reading of the S-1 those costs are expected to keep going up as more compute comes online. Would you buy this IPO?

S&P changed their outlook on $META to negative today, which means they could cut Meta’s credit rating. A big part of the reason is the rent from the data center leases I wrote about here. Free cash flow doesn’t show that rent yet but S&P counts it as debt. They say Meta’s debt could be too high for their rating as early as 2027. Meta’s free cash flow was less negative than Amazon, Alphabet and Oracle in Q3 and S&P is still worried about how much they are taking on. Microsoft has signed about the same amount in these leases but they can rent that capacity out to Azure customers. Meta is still working on renting theirs out, so their ads are paying for most of it.

Q3 is the first quarter that $AMZN, $GOOGL, $META and $ORCL all had negative free cash flow at the same time. In 2024 they mostly paid for the AI buildout with their own cash. In 2025 they added debt. This year Alphabet, Meta and Oracle started selling stock. And free cash flow doesn’t show everything they have agreed to spend. Many of the data centers they plan to use in the coming years aren’t being bought at all. Other companies are borrowing to build them and will rent them out, and that rent doesn’t show up in free cash flow until it starts being paid. Depending on how the lease is set up some of that rent never shows up there at all. $MSFT is the only one still positive, so they can keep paying for AI with the cash their business makes. But they rent data centers the same way, so that positive FCF doesn’t show all the rent they have signed for either. Which one would you rather own while they pay for all this?

Kevin Warsh is treating more expensive borrowing as evidence his policy is working without explaining how much it’s doing about the inflation from AI spending. He held because he thinks the September hike is already making borrowing more expensive. Most of that is coming from AI and AI investment is what’s speeding up the growth he says is strong. So higher borrowing costs should slow AI spending. Two weeks ago he said that spending was pushing inflation up. Slowing that spending should bring some of that inflation down. But how much are higher borrowing costs actually doing to bring it down? Productivity improving for a couple of years doesn’t answer that.

The $AMZN cash burn isn’t ending next year. Amazon plans to spend even more in 2027 because demand for AWS is still bigger than what they can build. AWS has $578B in contracts lined up and they expect AWS to grow over 40% in both 2027 and 2028. What they didn’t say is how much they will spend in 2027 or when the cash flow turns positive.

Look at the $AMZN cash burn! Amazon spent $34B more than they brought in this quarter. The estimate was for about $10B. It is the usual story of building data centers and buying chips for AI. Amazon is adding capacity so AWS can serve more customers running AI. They still say they will spend about $220B this year, so spending should come down in Q4 after this big jump in Q3. Any hope on this cash flow coming back?