Anthropic–IBM Pact Commits $190M to Enterprise AI Safety  ## Metadata - Author: [[Hard Fork AI]] - Full Title: Anthropic–IBM Pact Commits $190M to Enterprise AI Safety - Category: #podcasts - URL: https://share.snipd.com/episode/bf722e60-e2d7-4e59-84a7-338d9087aef8 ## Document Notes ## Highlights - **Anthropic Preparing For A Major IPO** - Anthropic is preparing for an IPO and has hired law firms to start the process. - The company could pursue an IPO as early as 2026 with valuations discussed near $300B. Transcript: Jaeden Schafer So the first deal that's, or the first thing that I think is really interesting is this upcoming IPO. They're preparing for an IPO. And apparently, according to the Financial Times, they just put out a report that said this could come as early as 2026 next year. They have actually already hired the law firms. They've hired Wilson, Sonisi to help kick off the process. And they're also doing a whole bunch of internal checklists to basically prepare for what could be one of the largest IPOs ever. Now, OpenAI is probably going to come out and do their IPO and then they'll top Anthropic at whatever point they do that because they have a lot more users. But this is kind of interesting and exciting and it feels like the first step. I think Anthropic, it's in their best interest to try to front run and do the first IPO. But I think OpenAI probably wants Anthropik to go first as well because they want to see how it goes and then they can decide what they would like to do next. So with all of this, they're currently looking at raising fundings right now that would value them at about $300 billion. The last round that they did was $13 billion. They did it in September. They raised $13 billion, and it gave them $183 billion valuation. So William Sinosi has been an advisor to Anthropic since 2022, so it makes a lot of sense to bring them on to prepare for the IPO. And I think all of this is really interesting because this is all coming as OpenAI currently is valued at about $500 billion. So if Anthropic is trying to do like their next raise at $300 billion, they're not that far behind OpenAI, but they're still, you know, not at the same level. And OpenAI is also reportedly testing the waters for an IPO. ([Time 0:01:25](https://share.snipd.com/snip/cf44844e-bb99-4a1a-be3c-e2403c576dad)) - **Enterprise Deals Bolster IPO Story** - Anthropic is signing large multi-year enterprise deals to strengthen its IPO position. - Announcing contracted revenue like the Snowflake deal helps showcase predictable future income. Transcript: Jaeden Schafer Whenever you see these companies go out to an IPO, the better the balance sheet, the better the deals they have in the pipe, the stronger a position, the more they can raise. So Anthropic has just signed a $200 million deal with Snowflake to use their LLM with Snowflake's customers. This is a pretty big deal. And I think as you see, you know, for a couple of things, number one, it's great for Anthropic because Snowflake is going to be, you know, using them. So it's more usage that will go through Anthropic. So it's good for their usage. It's also good for the bottom line, getting $200, $200 million. What I will say is this is not an annual deal. This is a multi-year deal. So it's kind of has a big sticker shock factor of $200 million up front, but this is not for one year. It's over time. And what's interesting is in a deal like this, it's typically not $200 million paid today. When they do these multi-year deals, it's going to be paid out. But what they'd like to do with these big announcements is to essentially, even though it's going to be paid out over time, it's a deal that's locked in. So Anthropic, when they go and do their IPO, they're like, look, we have this $200 million deal. And they can kind of point to all of the projected revenue. Look, we have, you know, a billion dollars in contracted deals out for the next year or two years, and it puts them in a much stronger position as they as the IPO. ([Time 0:03:16](https://share.snipd.com/snip/562f4947-88e9-4de5-8276-8ac188d58aa1)) - **Run Models Inside Enterprise Data Environments** - Place AI models where enterprise data already lives to reduce friction and improve security. - Use upgraded models (like Opus 4.5) to power agent-building and improve the quality of downstream tools. Transcript: Jaeden Schafer So Snowflake is talking about all this. They said that their customers are going to be able to use all of the different Claude models, including Claude Opus 4.5, the latest to run multimodal data analysis their customers Are also going to be able to use models to bring their own or to build their own custom agents which is really useful um the ceo of anthropic who is dario was talking about this as well and Said quote enterprises have spent years building secure trusted data environments and now they want ai that can work within those environments without compromise. This partnership brings Claude directly into Snowflake where that data already lives. It's a meaningful step towards making Frontier AI genuinely useful for business. Okay, what I will say about this deal, first of all, phenomenal, and it's actually coming at a really good time. So with the latest Sonnet 4.5 and Opus 4.5 that have come out, Claude has made a massive upgrade. And I think specifically with Opus 4.5, I've noticed for AI Box, my company, we have this like Vibe tool builder where you essentially can say, hey, I want to build a tool that does all Of these different things. We use Opus 4.5 to run that. ([Time 0:05:15](https://share.snipd.com/snip/d86b2eda-eb9a-4d82-be92-1222b009fbfa)) - **Model Upgrades Improve Tool Quality** - Upgrading the underlying model can noticeably improve the quality of tools built on top of it. - Jaeden saw a ~20% jump in tool quality after moving from Opus 4 to Opus 4.5 at AI Box. Transcript: Jaeden Schafer And I think specifically with Opus 4.5, I've noticed for AI Box, my company, we have this like Vibe tool builder where you essentially can say, hey, I want to build a tool that does all Of these different things. We use Opus 4.5 to run that. And what we found that was really interesting is when we upgraded from Opus 4 to Opus 4.5, and I bring this up because I think Snowflake is doing some similar things. But when we upgraded, literally the quality of our tools, so like we use it to build tools, the tools and the quality of the tools increased like 20%, I would say, which is kind of interesting Because it's not just like, oh, you know, the model gives me a 20% better response. When you're using these models to become kind of like the thinking engine and the intelligence, which is what Snowflake is doing with Snowflake Intelligence, when you're using it As kind of the engine to power building things and building agents, right? They're trying to build this like AI agent tool, this AI agent for enterprise customers. When you use this upgraded tool to actually power those, all of a sudden the tools that are coming out are actually built better and they work better and the prompts that are going into Them are better. So it's kind of interesting how the upgrade of a model can all of a sudden make tools get better, not just because of how, you know, they're responses, but because how they're built. They're built more intelligently. They're built smarter. And I've seen this with AI Box when we upgraded. ([Time 0:06:09](https://share.snipd.com/snip/80e0af1c-5793-4307-a0b2-b4ad0069910e)) - **Anthropic's Enterprise-First Product Strategy** - Anthropic focuses on enterprise and developer customers rather than maximizing consumer user counts. - That niche strategy yields strong adoption inside organizations and developer preference for Claude tools. Transcript: Jaeden Schafer Anthropic has a bunch of large enterprise deals right now that they've made in the last few months. Right now, I think they're trying to prioritize selling to businesses instead of individual users. You'll see OpenAI tout numbers like 800 million weekly active users, which is crazy. And Anthropik does not have those deals or that kind of like usage from individuals. But what they do have is a lot of businesses that use them. And I'll even say like for just purely coding, you know, using Claude Code and Anthropik's Claude Code is what we use exclusively to build AI Box and it is miles ahead of everyone else. It seems to be the number one tool that developers want to use when building and coding because it can look at your whole code base. They have a whole bunch of really cool things they've built into there. So Anthropic really seems to be focusing on more of that target audience. I think this is definitely a big contrast from what OpenAI is doing, which is just trying to get like max users anthropic in october made a deal with deloitte to use claude in for the consulting Giants um employee base they have over 500 000 staff and so claude is going to be kind of the default model there they made that deal in october um i think the same week they struck a partnership With ibm bring a bunch of their LLMs into IBM's software products. And so I think with all of that right now, Anthropics Enterprise success is not very surprising. Their models are very popular. I think there was a Menlo Venture survey in July that saw that employees preferred Anthropics AI products over models by other AI companies like Google and OpenAI. So we're seeing like inside of organizations, this is a very popular model for developers. This is a very popular model. ([Time 0:07:36](https://share.snipd.com/snip/c4d42cbc-f315-4d87-ac8a-4121e081002e))