Bryan Sim
LightNews LightNews
notquitepianoman.bsky.social
Bryan Sim
@notquitepianoman.bsky.social
96 followers 46 following 32 posts
Posts Replies Media Videos
notquitepianoman.bsky.social
Pink Phony Club
May 14, 2025 at 1:19 PM
Reposted by Bryan Sim
edzitron.com
if it turns out that, say, LLMs do not have any real business returns and haven’t this whole time and the whole thing was a huge overblown bubble that large chunks of the media and markets fell for, that would be a huge problem! If that’s the case I’m going to be insufferable
April 18, 2025 at 3:43 AM
Reposted by Bryan Sim
edzitron.com
I'm sorry this is my favourite story in tech for a while. Mark Zuckerberg went to Microsoft, Amazon and other big tech companies asking them to give him money to train their LLM without offering anything of any kind in return. They're absolutely cooked. Nobody believes in this shit anymore.
In its discussions with other companies, Meta primarily asked for money. It also sought servers or other resources that would offset the cost of training its models, according to two of the people briefed on the discussions. In return for their assistance, Meta discussed offering other companies promotion of their services alongside Llama—for example, a Meta executive might appear at a conference hosted by a consortium partner—or providing more insight into the training process for the model, one of those people said. It also offered assistance with creating a version of the open-source model for a company’s particular usage. But Meta didn’t want to offer ad credits or anything that would constitute a financial exchange, the person said. One problem Meta faces is that many of the companies it has pitched are already investing heavily in AI. That’s particularly the case for Microsoft, which has already committed more than $13 billion to ChatGPT creator OpenAI and Amazon, which has invested $8 billion in Anthropic and is developing its own AI models. That likely reduces their incentive to back Llama. Meta’s proposal spotlights the challenges the social network giant faces in developing Llama. Cloud firms can make money from their own AI model through sales of AI-powered cloud services. Microsoft and Google also charge for consumer access to their AI chatbots. The near-term costs are daunting for all tech companies. Meta, for instance, expects to spend between $60 billion and $65 billion in capital expenditures this year, up about 60% from 2024, as it develops more AI data centers. That’s about one third of the revenue it is expected to generate this year, according to analyst estimates tracked by S&P Global Market Intelligence. While Microsoft, Google and Amazon are each spending similar amounts or more on capex, largely because of their adoption of AI, all three companies are bigger in revenue terms than Meta. To be sure, Meta has deeper pockets than most businesses. It had $49 billion in cash on its balance sheet as of Dec. 31, after taking into account debt. Last year it generated $91 billion in cash from its operations, money that goes toward capex, stock buybacks and dividends. Still, Meta’s ability to keep paying for AI while also meeting shareholder expectations for buybacks and dividends is likely to grow increasingly difficult in the coming years. There have been other signs the company has been looking for alternative ways to fund its AI development. Bloomberg recently reported that Apollo was in talks to lead a $35 billion debt financing by Meta for data centers.
April 18, 2025 at 3:09 AM