OpenAI CFO Friar on New AI Assistant, Capital Raise

Author: Bloomberg Television · Watch on YouTube

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: OpenAI CFO Friar on New AI Assistant, Capital Raise · Channel: Bloomberg Television · 04:17 · Watch on YouTube

Topics: AI agents, Enterprise AI applications, AI infrastructure and hardware, Markets and economy

Summary

OpenAI CFO Sarah Friar says the company is initially positioning its new proactive assistant for prosumers and enterprise customers, where she sees immediate business value. She describes a unified platform that could serve both businesses and consumers, and says enterprise customers can benefit from features that catch missed tasks and questions. Friar discusses consumer and enterprise economics, citing free users, a new advertising business, and opportunities to sell and expand enterprise products. On capital, she says OpenAI is well capitalized but continues to invest ahead of demand in compute, which she says the business could use more of.

Key points

The assistant is launching first in the prosumer-to-enterprise arena, where Friar says OpenAI can bring value to business customers now.
Friar says OpenAI has 2.5 million businesses on the platform and that its business segment grew 100% quarter over quarter.
She describes the assistant as proactively flagging missed tasks, such as an unanswered customer question or a changed detail before an interview.
Friar says the platform is unified, with underlying intelligence intended to benefit users in different parts of their lives.
She says about 90% of consumer users are free users, and that a first-party chip could lower inference costs across the platform.
Friar describes OpenAI’s advertising business as the fastest to reach a $1 billion run rate, taking less than 200 days.
She says enterprise products can include coding, cybersecurity, and offerings tailored to specific industries, with opportunities to upsell customers.
Friar says OpenAI is well capitalized and that investments in compute must be made ahead of time so the business does not need to slow down.

Chapters

Start	Chapter	Summary
00:00	Initial market for the assistant	Friar says OpenAI is starting with prosumers and enterprise customers, citing immediate business value and growth in the business segment.
01:03	Proactive features and a unified platform	She explains how the assistant might catch missed business tasks and says the platform can serve both enterprise and consumer users.
02:07	Consumer and enterprise business models	Friar discusses compute costs, free consumer users, advertising, and ways to sell and expand enterprise offerings.
03:08	Capital and compute investment	Asked about a reported capital raise, Friar says OpenAI is well capitalized while emphasizing the need to invest in compute ahead of demand.

Quotes

We’re starting in more of the prosumer to enterprise Arena — 00:00

My dot is now proactively tapping me on the shoulder — 01:03

It took us less than 200 days. — 02:37

we have made the investment in compute so that we never need to slow the business down. — 03:43

we could use more compute if we had it today. — 04:17

Concepts and entities

OpenAI, Sarah Friar, ChatGPT, AGI, inference, IPO

Full transcript (original language)

00:00 What’s interesting with darts, right? The question straight away with Sam was is it a consumer prosumer or enterprise product at first? Yeah. Where do you stand on that? We’re starting in more of the prosumer to enterprise Arena, and it’s really for a couple of reasons. Number one, I think there’s a lot of value we can bring today to our enterprise customers. Right. We have 2.5 million businesses on the planet. The business segment just grew 100% quarter over quarter. So it’s on fire at the present day through today. Yeah, from just the beginning of July.

00:30 And what’s special about that is we can now show them this whole other proactive side of I think Sam said it well, in the keynote, he said, this is always our dream of what I or AGI could be. And I think we’re getting very close to that. We have some options, right? In terms of true agents instinct, Messi’s muse grumble, the prophecy and okay, well make the case. I mean, the pricing on docs was the interesting part to me. Yeah, in the marketplace. So on dots today we’re starting an insider pro skew and then moving it up into our enterprise offerings. And we think that there’s tremendous

01:03 value to be had there for anyone in the enterprise. right? My dot is now proactively tapping me on the shoulder and noticing that I might have missed something that needed to go out last night right ahead of Dev Day, or reminding me that something’s changed before I sat down in this interview. I think that has tremendous ROI for a business, right? Because you miss those little things like a customer question in your inbox, that means you can’t close a deal at the end of the quarter. That is real ROI, and that’s why we started there. We will go much bigger. We have 1.2 billion users weekly using our product, right?

01:34 That was a new stat today to at that day it’s up 20%. Yes, in about six weeks. So clearly we know there’s tons of consumer interest here too. Um sometimes we go business down to consumer or sometimes we go consumer up into business. I think the point is it’s a single unified platform where the raw intelligence that sits behind it will benefit you in all walks of your life. Sam said 1.2 billion people are using ChatGPT every week. One of the questions I get for you a lot, and it’s from quite a broad audience, is the margin profile of a consumer user versus enterprises.

02:07 I know that that is not a straightforward calculation, but what are the differences for you? That’s how you manage them, right? To compute costs come into account. So I think there’s a couple of things there. So first of all, we do build a unified platform. And that’s going to benefit from things like investments in the compute stack. Right. When we do our own first party chip, we’re able to offer inference at a lower cost that benefits everyone. When you think about our consumer business today, 90% ish are still free users. That’s very mission aligned. AGI for the benefit of humanity, not

02:37 just the people who can pay, but we’ve kicked off an amazing ad business now the fastest advertising business to $1 billion run rate ever. It took us less than 200 days. So now I can see that kind of more of a roadmap towards how we improve the margins on the consumer side. On the enterprise side, like enterprise businesses are wonderful businesses. They tend to be quite sticky today. When we walk into the enterprise, we can sell coding, we can tell cyber, we can go deep into their particular vertical. Um, we’re seeing just more and more ways to get into those enterprises and then

03:08 upsell them. And so that business is very, very strong. That’s why it doubled quarter over quarter in Q3 capital management. So we’re at Dev Day. Bloomberg reported this morning that, uh, these are early talks with with private market investors to raise capital 30 billion at $1.4 trillion valuation. What can you say about that? But I think Sam’s been pretty clear today about an IPO happens in the future. You have a period to manage through. So first of all we’re really well capitalized. We did on $122 billion raised in Q1.

03:43 And that’s literally a balance sheet that sits on the balance sheet. That was that was record breaking. Um, that said, we’re in a very fast growing business, and my job is always to make sure that, number one, we have made the investment in compute so that we never need to slow the business down. And those are investments that need to get made well ahead of time. In addition that every time we make investments that we’re looking at it with the right ROI lens. Um, if anything, I would say, you know, a year ago we were taking down a lot of compute folks like you were giving us a little bit of a hard time for it. Um, and I think right now, we still sit

04:17 in the same place where we could use more compute if we had it today. We could use more compute in 2020.