Well that's the thing, 3rd parties touch everything. Even for you making this post you've probably used a phone from company A, that runs an operating system from Company B, using Company C's telco.
It's literally impossible to bring every task completely in house, even if it was possible you'd probably pay at least triple the current prices for everything. It's just not going to happen, it's virtually impossible.
The issue is these 3rd parties can be a bit too slow to apply patches and updates to fix vulnerabilities, that's where there needs to be more accountability and punishment especially for known issues that have patches already released that they ignored.
The same china that refused 200k H200s
The same megaspeed that Nvidia stopped selling to in _2025_
The same coreweave that's getting... A few B? Add it all up and that's definitely the 450B in revenue they made this year.
I swear to God you people failed basic math and are completely unaware of how transformative Agentic AI is. You will be left behind.
Yeah but profitable and "not good at software engineering" ain't the same. AI is at a point where it's like a human. If you give 5 software engineers the same problem you might end up with 5 different solutions that all work (but might not be extensible for the future in the same ways, in other words all could have different pros / cons). Today, what software engineers don't like about AI is that it didn't pick the same solution as the swe would have picked.
(A bit of side lifestory). When I started working at big tech the first year or two I was so impressed - this huge scale, these very interesting projects, etc... but years later I realized all the people there, even the very very high up positions don't necessarily make things any better. I've seen a team having system A. Then that team creating system B, migrating users from system A to system B. Taking years to do that, a bunch of folks get promoted during these years long projects then they switch teams. New people come to the team and they propose this years long project which is gonna bring huge benefits which actually is creating a new system C which in reality is just system A but these new folks didn't know that system A existed before. And these new folks do this huge project get a bunch of promotions and the company 5 years down the line is exactly in the same position it was before. Making your impact as objective as possible and tracking it is very important and it's the only way to get promotionsil in the beginning of your career. But once you go up high, somehow at the end of the project nobody asks whether this 3 year long project actually brought any benefits. Nobody can, because huge projects like that are agreed to be started by multiple people, so now nobody can say this was a waste of 3 years of multiple teams as every high position person would be responsible. No instead you just praise each other and all get promotions.
Uber and YouTube were not profitable for a decade (?). You gotta capture the market. It's so easy to switch AI providers, it'd take me literally 5 minutes to port all the memories, so there's no lock in. Except for my mom and by far the majority of the planet. She doesn't know that she can do that, she doesn't realize that she can ask AI itself whether she can transport all the knowledge the AI has about her to another AI. She's stuck. It's such a low effort migration but it's impossible for her. You don't know what you don't know. Think about an iphone user switching to an Android, not possible.
Opex is tomorrow Gujarati trader knows roaringkitty comeback is because shorts have to roll over synthetic gamecock shares + cover b/c their nveecock collateral is evaporating. Gujarati knows.
Pat-a-cake, pat-a-cake, baker's man,
Bake me a cake as fast as you can;
Pat it, and prick it, and mark it with a **B**,
And put it in the oven for baby and me.
The air is coming out of the bubble, there is no good news left... inflation is up, war is still going on ... AI can't finance b/c there is too much debt.