Let's just value OpenAI at infinite money and be done with it
Also: Should we start a bribe-o-meter?
Welcome to Cautious Optimism, a newsletter on tech, business, and power.
📈 Trending Up: Consumer interest in DeepSeek … domestic AI models … it’s o3 day from OpenAI, reportedly … leaks at Meta …
Trending Flat: Inflation data for December, with consumer prices (and consumer prices ex-food and energy prices) meeting expectations. PCE Price Index rose 0.3% and 2.6% YoY, while Core PCE rose 0.2% and 2.8% YoY. That’s still more inflation than the Fed wants, but we’re far from being in the middle of the woods.
📉 Trending Down: North Korean troops in Ukraine? … cybersecurity …
Masa, you roguish dog
During the Project Stargate news cycle it was clear that SoftBank’s Masa Son was in a dealmaking mood. After a second — second! — trip to a podium with President Trump in recent weeks, Masa was stoked to announce plans to put tens of billions of dollars to work on OpenAI’s behalf.
Little did we know that he had more up his sleeve. The former WeWork financier — and backer of Alibaba, it’s worth recalling — is now considering far more cash for the AI concern. Here’s the WSJ:
In what would be the largest-ever investment in a startup, Son is preparing to put as much as $43 billion toward Altman’s OpenAI in a pair of transactions. SoftBank is in talks to invest between $15 billion and $25 billion in the ChatGPT maker as part of a blockbuster $40 billion funding round, The Wall Street Journal reported Thursday. The round would value OpenAI at up to $300 billion—nearly double its valuation in October and an undeniable sign of Son’s confidence in its prospects.
Those figures are not inclusive of the funds that SoftBank has pledged towards Project Stargate itself.
What’s notable about Son’s potential wager is not that it’s huge, it’s that he’s willing to ratchet up OpenAI’s valuation risk. The company’s prior fundraising at a $157 billion price tag was considered by some folks to be a bit aggressive, given that there’s doubt amongst investors that all present-day AI revenue will prove to be sticky.
Masa isn’t worried about that pedestrian shit. No. He’s ready to get to work.
With $3.7 billion worth of 2024 top line, a $300 billion price tag would put OpenAI on a 81x trailing revenue run rate. Far lower if we took its January revenue and turned that figure into a run rate, but that’s a very, very bullish number for a company that was recently jolted by DeepSeek.
Intel, Apple, Atlassian earnings
We’re buried in earnings at present — lucky us!. You can read our notes on Microsoft, Meta, and Tesla earnings published yesterday here.
Intel earnings brief: “Intel reported earnings per share (EPS) of $0.13 on revenue of $14.3 billion. Analysts were anticipating EPS of $0.12 on revenue of $13.8 billion. The company saw EPS of $0.54 and $15.4 billion in revenue in the same quarter last year.”
You can find Intel’s investor deck here.
Revenue and gross margins were projected under street expectations for the current quarter. The company will continue to shrink through at least the current quarter, as will gross margins.
Intel is now hunting for a CEO. The company’s earnings were good enough to earn back a few bips of share price, but the story remains the same.
Here’s an amazing moment from the Intel earnings call:
“Yes, OK. So, the co-CEOs are trying to figure out who should answer OK, go ahead, Michelle.”
Apple earnings brief: Apple narrowly beat revenue expectations ($124.3 billion, +4% YoY compared to an anticipated $124.1 billion). Mac and iPad revenue beat expectations per CNBC’s accounting, while iPhone sales came in light.
Apple returned $30 billion to shareholders during the quarter. That figure included “$3.9 billion in dividends and equivalents and $23.3 billion through open market repurchases of 100 million Apple shares,” per an earnings call transcript.
Apple’s China business was off 11% in the quarter. Shares of Apple are up around 4% in pre-market trading.
Atlassian earnings brief: Revenue of $1.29 billion (+21% YoY), +30% subscription revenue in the quarter, and $342.6 million worth of free cash flow (27% margin) were more than the street expected.
The company also forecasted around $1.35 billion worth of top line in the current quarter, and 18.5%-19.0% growth in its current fiscal year.
The above financial reporting was further buoyed by Atlassian reporting that it now has “over 1 million monthly active users (MAU) of AI across our platform, and the number of AI interactions has increased over 25x year-over-year.”
The company’s Rovo family of AI products includes agents. That came up on its earnings call. Here’s how Atlassian’s CEO Michael Cannon defines agents (transcript source):
“They have to have goals. They have to be aiming at outcomes. They generally have some sort of personality and character. They have sets of knowledge that they can do and sets of actions that they can take and some sort of control parameters in terms of commissioning and everything else, which ends up making them look extremely like a virtual teammate and represented in the software as such.”
What about DeepSeek? More from Cannon (emphasis added):
“Any agent is going to be qualified by the quality of models for sure. So, what is underlying these agents is a series of different AI models. We have a very comprehensive multi-model strategy. So, we believe that models will continue to get faster and cheaper and quicker to deploy and more capable. Therefore, our Atlassian Intelligence needs to be able to keep adapting modern models as fast as possible. Again, we're running more than 30 models from more than seven different vendors today. We continue to evaluate new models. Obviously, we had a lot of movement in that space in the last week or two.
Should we start a bribe-o-meter?
A recent choice by Meta to settle a frivolous suit that the President had filed against it for $25 million is brass-necked ass-kissing.
The massive surge in donations to the President’s inauguration slush fund was also pretty gross.
And now Amazon is bowing to X to help keep its CEO happy enough to potentially avoid bad-mouthing the company to the President? In case you were curious if rising authoritarianism — see attacks on the separation of powers, above — would engender corporate spine, here’s your answer.
Tariffs
Tariffs are coming tomorrow, per the President. Thus far it appears that we’re looking at a 25% tax on imports from Canada and Mexico. China is reportedly on the list, too, but it’s less clear what the timeline of new tariffs on the nation will prove to be or at what level.
There had been some hope amongst business-types that Trump would not impose blanket tariffs on major trading partners. Those hopes now appear dashed, so long as Trump follows-through with his promise.
Tariffs are shit, in case you needed a reminder.
Antitrust
Another major business hope was that the Trump administration would radically reshape federal policy on mergers. More, and with little to no oversight was the expectation.
Instead, the Department of Justice has sued “to block Hewlett Packard Enterprise Co.’s (HPE) proposed $14 billion acquisition of rival wireless local area network (WLAN) technology provider Juniper Networks Inc. (Juniper).”
Here’s the argument (excerpted):
“Juniper has been a disruptive force that has grown rapidly from a minor player to among the three largest enterprise-grade WLAN suppliers in the U.S […] This competitive pressure has forced HPE to discount its offerings and invest in its own innovation […] Now, HPE seeks to acquire its smaller, innovative rival. The proposed transaction between HPE and Juniper, if allowed to proceed, would further consolidate an already highly concentrated market.”
That sounds like a pretty standard antitrust perspective on a corporate combination — that, while friendly to the two businesses in question, would lessen overall market competition. But what business wanted was the ability to do deals of just this sort.
One data point is not a trend. But if you were hoping for no oversight on antitrust, you might not get what you thought you voted for.