alexdoesstuff

Small nitpick: the models probably make some money on actual inference. Might not be a massive amount, but hard to see them not having a positive contribution margin purely on inference.

What's losing OpenAI money is paying for the whole of R&D, including training and staff. Microsoft doesn't pay that, so they get the money making part of AI without the associated costs.

It's unclear. That was never disclosed. It's similarly unclear what it means that they will no longer pay revenue share to OpenAI. Do they get the models for free now? How does OpenAI make money from the models hosted on Azure if not via revenue share?

It's kind of shocking, given financial transparency, that Microsoft gets away with not disclosing any details of this agreement (or the one it is replacing) to its shareholders. We know there's a cap on the revenue share from OpenAI to Microsoft, but we have no idea what that cap is (not whether it's higher, lower, or unchanged from the prior agreement).

We have no idea what it means to be the "primary cloud provider" and have the products made available "first on Azure". Does MSFT have new models exclusively for days, weeks, months, or years?

Both facts and more details from the agreement are quite frankly highly relevant to judge whether this is a net positive, negative or neutral for MSFT. It's unbelievable that the SEC doesn't force MSFT to publish at least an economic summary of the deal.

We updated our analysis of Meta's AI infrastructure spending after yesterday's Q4 call. Some specifics from the earnings call on where the money is going:

Ad models: They doubled the GPU cluster training their main ads ranking model. The architecture now reportedly "scales with similar efficiency as LLMs." They also consolidated multiple models into fewer, more capable ones (12% increase in ads quality) and moved their ads retrieval engine across NVIDIA, AMD, and custom MTIA chips, nearly tripling compute efficiency.

Internal tools: Management cited a 30% increase in output per engineer since early 2025, mostly from agentic coding tools. Power users saw 80% gains.

Ad products: Video generation tools hit a $10B combined revenue run rate. Incremental attribution drove a 24% lift in conversions.

On the financial side: Q4 ad revenue was $58.1B (+24% YoY). FY25 FCF was $43.5B, a 22% margin, down from 32% in FY24. They paused buybacks and signaled they may take on net debt. Management committed to FY26 operating income above FY25 in absolute terms against $162-169B in guided expenses.

For context, Meta's 2025 capex alone roughly equals the total lifetime investment in Reality Labs over 10+ years.

The article covers the full AI model stack (GEM, Lattice, Andromeda), capital intensity breakdown, and peer comparison.

This is primarily a story of a failure to supervise the creation of the report, rather than anything related to AI.

The role of the outsourced consultancy in such a project is to make sure the findings withstand public scrutiny. They clearly failed on this. It's quite shocking that the only consequence is a partial refund rather than a review of any current and future engagements with the consultancy due to poor performance.

There shouldn't be a meaningful difference if the error in the report is minor or consequential for the finding, or if it is introduced by poorly used AI or a caffeinated up consultant in a late-night session.

To expand on the overlooked point: it gives you a DB and a programming environment (however challenged) that you can use without needing sign-off from IT. In any moderately sizeable organization, getting approval to use anything but standard software is slow and painful.

Nobody wants to explain to IT that they need to install Python on their machine, or drivers for sqlite, or - god forbid - get a proper database. Because that requires sign-off from several people, a proper justification, and so on.

The $13bn investment in 2023 was so clearly structured to skirt antitrust concerns that it's unsurprising that that avenue is discussed.

Since then, MSFT has made other regulatory-aggressive investments, and the recent Meta / Scale AI is similarly aggressively designed.

Full agree!

Being close to the edge of AI usage, it's important to realize that most AI use cases are not "fully autonomous AI software engineer" or "deep research into a niche topic" but way more innocuous: Improve my blog post, what's the capital of France, what are some nice tourist sites to see around my next vacation destination.

For those non-edge use cases, costs are an issue, but so are inertia and switching costs. A big reason OpenAI and ChatGPT are so huge is that it's still their go-to model for all of these non-edge use cases as it's well known, well adopted, and quite frankly very efficiently priced.

Reading through the source [1] they basically get to that huuuuge number by including AI-enabled devices such as phones that have some AI functionality even if not core to their value proposition. That's basically reclassifying a big chunk of smartphones, TVs, and other consumer tech as GenAI spending.

Of the "real" categories, they expect: Service 27bn (+162% y/y) Software 37bn (+93% y/y) Servers 180bn (+33% y/y) for a total of $245bn (+58% y/y)

That's not shabby numbers, but way more reasonable. Hyperscaler total capex [2] is expected to be around $330bn in 2025 (up +32% y/y) so that'll most likely include a good chunk of the server spend.

[1] https://www.gartner.com/en/newsroom/press-releases/2025-03-3...

[2] https://www.marvin-labs.com/blog/deepseek-impact-of-high-qua...

Author here

I mostly agree on the first point. Even prior to the price race to the bottom, no AI Lab managed to make any money above marginal cost on inference, let alone recoup investment in infrastructure or model training. Clearly, investment in infrastructure and model training have been largely subsidized by VCs. It's a bit unclear how much of a subsidy inference costs had. The fact that AWS runs hosted inference at roughly similar cost than AI Labs suggests to me that there's at least not a massive subsidy going on at the moment.

I don't subscribe to the narrative that nation states (i.e. China) massively support DeepSeek. Thus, while their core business as a hedge fund is clearly profitable, they have considerably less deep pockets and willingness to front losses than the investors in VC supported AI Labs. Consequently, I expect their inference cost to at least cover their marginal costs (i.e. energy) and maybe some infrastructure investment.

All that suggests that they've managed to lower cost (and with that presumable resource and energy requirements) of inference considerable, which to me is a clear game changer.

Fully agree on the premise: there are X different ways to do anything on the web. But - prior to this - the solution seemed to be: everyone starts from scratch with some ad-hoc Regex, and plays a game of whackamole to cover the first n of the x different ways to do things.

Best of my knowledge there isn't anything more modern than Mozilla's readability and that's essentially a tool from the early 2010s.

Feels surprising that there isn't a modern best-in-class non-LLM alternative for this task. Even in the post, they described that they used a hodgepodge of headless Chrome, readability, lots of regex to create content-only HTML.

Best I can tell, everyone is doing something similar, only differing in the amount of custom situation regex being used.

Have you ever seen the research coming out of some of the outsourcing shops that the OP discusses in the post? They are hard not living up to that standard. It's important to realize that this is input for the analyst at a fund or investment bank to do some more digging on the companies and in the process potentially discover more. This isn't going straight to the CEO to form the basis of an investment decision.

I thought the role of "second best tool for most things" belonged to Excel

Fair point on the publishing! Thanks for sharing the library.

Do you mind sharing the code for this, or point me to the audio fingerprinting solution you use? I'd love to replicate this.

Brother DCP-L2550DW here. One of the cheapest b/w multifunction devices with automatic document feeder and reasonable print and scan performance. Works like a charm on Linux, Windows, Android, and IOS.

I am using it with [NAPS2](https://www.naps2.com/), which is brilliantly simple, multi-platform, free, and open-source.

Actually true in a technical sense. At least for Starbucks, their gift cards don't expire anywhere globally. In accounting terms though, the company still reduces the value of its gift card liability every year by about 10-15% [1], and claims that this is based on historical data.

Now clearly, there are circumstances in which banks do something similar and close accounts of account holders that are unknown. However, if that occurred at even one hundredth of the Starbucks breakage rate, all regulatory hell would break loose on the bank.

[1] Starbucks reports breakage of around $212.7m in FY22 ($181.1m in FY21). Their liabilities to Stored Value Cards are $1,641m and $1,596m respectively, coming out to a breakage ratio of 13.0% and 11.4% respectively.

Starbucks gets prepayments for goods and services and just has a neat bit of accounting and marketing gimmick around it. Besides being a sensationalist take, it's also misunderstanding what makes a company into a "bank".

You can't withdraw your balance in cash, as you can with a bank account. You can't transfer your balance to someone else like you can with a bank account. And, unlike a bank, your Starbucks gift card balance expires after some period of time. Can you imagine your bank telling you that all the money in your account is theirs because you haven't used the account in a few months?

Does anyone know if Azure's OpenAI Studio is down as well? For everyone using ChatGPT APIs in production, this needs to be the most straightforward failover mechanism.

Using this to plug our open-source tool https://github.com/Marvin-Labs/lbgpt which allows ChatGPT consumers to quickly load balance and failover between OpenAi and Azure models.

They don't have to be board members to fight. The 51% shareholders are clearly wronged here in my view if MSFT just hires all of OpenAI staff AND takes the existing models and trade secrets.

I feel people tend to forget that MSFT owns only 49% of OpenAI. There are people owning the rest, and they presumably have some interest in protecting their share of the company from MSFT's aggressive grab. Even if all new development goes to hell, GPT-4 and co are still worth something.

Maybe MSFT really got the deal of a lifetime where they (a) tricked the FTC to not consider anything they are doing to touch antitrust laws due to "only buying 49% of the business" and (b) have a way to "extract" the core assets of OpenAI to the detriment of the remaining shareholders. That's not impossible given the quaint structure OpenAI has and the unusual structuring of the deal with MSFT, but would be quite surprising.

the people owning the remaining 51% of OpenAI?

True. Being a founder myself, I've experienced the back and forth of investment negotiations.

We are though taking about OpenAI, by that time probably one of the five most valuable startups in the world.

This is quite literally the CTO of Microsoft saying: "Know that if needed, you have a role at Microsoft that matches your compensation and advances our collective mission." I don't think there is a reasonable argument where this is not poaching.

I have all the sympathy with OpenAI employees signing the letter and looking for new jobs.

However, MSFT is basically doing a cheap de-facto acquisition of OpenAI, a company they have invested in. This sets a terrible precedent for companies taking investment from strategic investors where they may stage an opportunistic de-facto acquisition at the sign of any trouble.

Presumably, this violates MSFT's investment agreement with OpenAI. Any reasonably competent counsel would add "no-poach" protection for a strategic investor investing in a startup, and this is as clear a case of poaching as there is.