Inflation data disappoints, notes on Crunchbase's AI strategy, and the latest OpenAI math
Welcome to Cautious Optimism, a newsletter on tech, business, and power.
📈 Trending Up: China meddling with Taiwanese nationals … “persistent specific supply chain issues” … Stripe + Stablecoins … spine … Relyance … women’s sports … Rhode Island billionaire drama … angst in the WordPress community …
DHH, on said drama: “That's the deal. That's open source. I give you a gift of code, you accept the terms of the license. There cannot be a second set of shadow obligations that might suddenly apply, if you strike it rich using the software.”
📉 Trending Down: Internet Archive … honesty in crypto … cybersecurity … rural living … cross-platform angst … declines in manufacturing jobs …
September’s inflation data
The BLS dropped new inflation data this morning. The Consumer Price Index (CPI) rose 0.2% in September for a trailing twelve-month change of 2.4%.
That’s progress: “The all items index rose 2.4 percent for the 12 months ending September, the smallest 12-month increase
since February 2021.”
But the market had expected a 0.1% gain, not 0.2%. Concerns that inflation might still be a sticky will now fill financial headlines.
Stocks are falling in response to the inflation miss. Initial jobless claims also ticked higher, indicating some weakness in the labor market.
Pretend you are the Fed: What would you do at your next confab:
Crunchbase’s AI Strategy
🗂️ Bias Alert, Crunchbase edition: CO uses Crunchbase, PitchBook, YCharts, and Yahoo Finance daily. In the case of Crunchbase, I’m conflicted since I worked there and retain some stock from that time. That said, I keep in touch with private-market data providers because I can’t do my job without their data. So, when I got a chance to rap with Crunchbase’s chief product officer Megh Gautam, I dove in.
I was curious: AI moats are built on proprietary data today. So, shouldn’t Crunchbase be doing something pretty cool, given its data set? (The company recently introduced natural-language, AI-powered search builder, and a “Predictions and Insights,” accessible via the Crunchbase API, for reference.)
Why was he hired? “I came in with a data-related mandate.”
A fun question to ask ourselves at this juncture is if there are any CPOs left in a few years that do not have a data-related mandate. No data edge, no AI edge. No AI edge, no future?
What people want from Crunchbase: “There's still a huge market for prioritization. Like, ‘I have two amazing sales reps, give me 20 accounts to go after,’ or ‘I have two amazing associates, give me the next 10 things that they should go deep into.’ And we talked to a cross section of people, right? We talked to VCs, PE, we talked to sales, we talked to RevOps, we talked to compliance — and they're all coming back with the same thing: Just help us be smarter, because you have a lot of information, [and] you don't make it very easy to get that information.”
I wonder what fraction of the AI market is just this: Help me be smarter, you have all the information and I want access to it. Now. Half? More? I suppose that’s a bullish perspective for Google.
On data scraping: “We want to aggressively block scrapers and crawlers. I don't think [allowing them to scrape] makes a ton of sense. We've been talking to all the players you could think of [and can share more shortly] because we are trying to figure out what’s the [Crunchbase] posture.”
Recall that OpenAI offered Medium $1 million yearly for access to its data. Crunchbase also has a lot of data. Presumably it would want at least that much from each model that would want to license its data. But would that be enough to balance lost potential revenue? I wonder if Crunchbase does more than a single, exclusive deal in time.
OpenAI is going to burn a mountain of cash — does it matter?
Over on TWiST, Jason and I have discussed OpenAI’s business model, business results, and valuation several times. Here’s the most recent, timestamped to start at 1:12, in case you need to scrub ahead:
I’m a bit more bullish than he is, but thanks to new data reported by our friends over at The Information, we can do a bit more today on the question of is OpenAI incredibly overvalued?
As a reminder, CO previously argued that OpenAI’s revenue growth is quick enough today to make its valuation pencil-out. Or as we put it a few days ago, the OpenAI math maths.
Let’s start at a high level. AI models quickly cannibalize their predecessors, and rapid improvement in AI models has led to the time in which a model can be considered state-of-the-art — put another way: commercially viable — very short.
The concern is that companies like OpenAI and its rivals will burn through mountains of cash in the process, leading to a simply unsustainable capital incineration. Unless a veritable mountain of revenue follows, AI model companies could become corporate examples of the concept of Pyrrhic victories.
The Information’s latest reporting on OpenAI doesn’t limit those concerns. Cory Weinberg writes that OpenAI won’t become profitable until 2029, when it will manage the feat against $100 billion worth of top line. But, with losses scaling to $14 billion in 2026 per OpenAI projections, total losses between 2023 and 2028 worth $44 billion, and training expenses of $9.5 billion per year by 2026, you might wonder if it will get there at all.
Those are all very big numbers, but I am curious: What did people think OpenAI was doing with all the money it raised? Sitting on it? No, you only raise $6.6 billion after raising billions while also tacking on $4 billion worth of debt if you are going to spend.
Can any private company spend as much as OpenAI expects to, and survive? You might hear the faint peel of a bell in the background that has ‘WeWork’ scribbled on its lip. Don’t worry. WeWork lost a lot less money than OpenAI intends to.
Jokes aside, there are two good arguments why OpenAI’s investment in AI models, and its planned cash outlays, are not overly-ambitious corporate suicide:
The short window to monetize new AI models is less risky to companies that have a huge base of revenue — recurring (subscription) and otherwise (API calls, etc), because they can be certain of a good chunk of top line against their spend.
OpenAI’s ChatGPT business is somewhere between $2 billion and $4 billion per year, today.
OpenAI has raised more than $20 billion worth of equity and debt since the start of 2023, implying that it has already raised about half of the $44 billion it expects to need through 2028, the last year it expects to not be profitable.
You can imagine another $5 billion private-market investment to get OpenAI ready to list, and then a $10 billion IPO haul and, presto, the company is nearly entirely financed to profitability.
There’s huge amounts of risk in the above:
OpenAI could struggle to maintain its market leadership in AI model advancements, limiting revenue while not conserving cash.
OpenAI’s revenue growth could undershoot targets, leading to greater capital needs, sooner.
Private-market backers may balk at putting more cash into the company before it goes public.
The public markets could prove unwelcoming to OpenAI if its losses are too painful compared to its revenue growth.
But that’s investing, baby! You gotta risk it for the biscuit, or whatever phrasing chestnut you prefer.
If I had a few billion to put to work in the AI era, backing OpenAI during a period of revenue growth this sharp might seem a lot more reasonable than backing a host of smaller AI-focused startups with $50 million apiece, yeah?