Brad Gerstner: Does OpenAI Need a Bailout? Mamdani Wins, Socialism Rising, Filibuster Nuclear Option

The episode opens with the hosts joking that Brad Gerstner, a guest on this episode, had "crashed the stock market" and popped the AI bubble. The jokes refer to an interview with Sam Altman on the BG2 podcast the previous Friday. One host says portfolios "rolled down 15% this week". Another asks whether OpenAI could put a moratorium on public statements for a couple of months. The tone is teasing, but the group agrees the episode is worth discussing seriously.

The hosts describe the question as mundane, even a "softball" or "an underhanded pitch," and one says it was asked "in a very fair way." They then play the clip. The question is about the gap between OpenAI's reported revenue and its spending commitments:

how can a company with 13,000,000,000 in revenues make 1,400,000,000,000.0 of spend commitments?

The reply in the clip begins by saying OpenAI is doing "well more revenue than that." It then offers to find a buyer for the questioner's shares and says many people who express "breathless concern" about OpenAI's compute plans would be glad to buy shares. On substance, the reply says the company expects revenue to keep rising and is making "a forward bet" that it will. It also says one of the rare appealing things about being a public company would be letting critics short the stock and watching them "get burned."

We do plan for revenue to grow steeply. Revenue is growing steeply.

The discussion notes that the internet took the clip as a sign that Altman was angry and hostile. The person who asked the question gives a softer reading. They say the two "had a good laugh after," that the reply was "feisty" but meant partly as a joke, and that Altman knows the questioner wants to buy more shares, not sell. In this telling, the clip went viral because the question is important: people are nervous that the market is walking into an AI bubble. The questioner adds that they were "a little disappointed" the feistiness "got in the way of the answer."

The rest of the explanation tries to reconstruct the answer from the numbers. Altman is described as saying several times on that podcast that OpenAI expects revenue above $100 billion within the next couple of years. A chart circulated to the team showed The Information's forecasts for OpenAI and Anthropic revenue. When another host asks whether those are leaked figures or guesses, the reply is that they are reportedly based on leaked internal numbers, "according to the information."

The arithmetic offered goes as follows. The $1.4 trillion is spread over five or six years. The speaker estimates that about half will be paid by partners, which leaves about $700 billion. Spread over that period, the later years would carry perhaps $150 billion of CapEx for OpenAI. If revenue reaches around $150 billion, the speaker says, the plan "begins to pencil out a little bit more." These are the speaker's own estimates and are labeled as such.

The point treated as most important is a conditional one. According to this account, Altman said that if the revenue doesn't arrive, OpenAI must cut spending to match it. The speaker expects the company would "extend, recut the deals" to keep the expenses manageable.

and if we don't have those revenues, we've got to match our revenues to our expenses.

Another host flags the main limitation. Nobody outside has seen the actual contracts, so it isn't known whether they include conditions, exit clauses, deferral rights or cancellation terms. That may only come out in public filings. The host also notes that the market reacted badly regardless. Partners named as close to OpenAI's deals, Microsoft, NVIDIA, Oracle, Broadcom and CoreWeave, were "all down six to 20%," which the host calls a significant correction in the AI boom.

One participant argues that the interview and the market move are "totally and completely uncorrelated." They acknowledge that everyone has bad days on camera. They suspect that, given a second chance, Altman would have said it differently, or at least delivered the joke better. But they locate the cause of the sell-off elsewhere.

Their account has two parts. First, the market is still digesting the CapEx already spent. Analysts are building models to predict the return on that spending and its effect on future earnings. This applies less to OpenAI than to the large "mag seven" companies. The speaker cites mixed results: Google's earnings and AI numbers were "blazing hot," Facebook's were "terrible." They also say Apple seems likely to cede its AI business to Google and pay Google billions a year, the reverse of the search deal under which Google pays Apple. Second, they describe year-end behavior: investors book gains, harvest tax losses and rebalance. In their view this now starts in mid-November instead of mid-December because traders try to get ahead of the usual December price moves.

So I wouldn't pin this on Brad and Sam. I just think this is natural market machinations. But to be clear, we are very much getting into a phase of risk off.

The host says the anxiety got worse on Wednesday. OpenAI's CFO, Sarah Friar, told the Wall Street Journal she hoped the US government would "backstop" financing for the company's $1.4 trillion of data centers. The host says she argued federal guarantees would "really drop the cost of financing." That would let OpenAI borrow more, at lower rates, from a larger pool of lenders. The remark went viral and fed a story that OpenAI might be insolvent. The host calls that story "a little bit ridiculous." They note that people compared the moment to the dot-com bust and the 2008 financial crisis.

The same night, Friar walked the comment back. In the statement read on the show, she says the word "backstop" "muddied the point." She says her actual argument was that American strength in technology will come from building real industrial capacity, with both the private sector and government playing a part. The host adds that her statement put OpenAI on pace for $13 billion. That is the figure Altman had disputed, so the host leaves the question open.

OpenAI is not seeking a government backstop for our infrastructure commitments.

The host says David Sacks, introduced as the group's "czar of AI," had tweeted that there would be no federal bailout. The host's reasoning is that if OpenAI failed, five major frontier model companies could take its place. The follow-up explains how the story came together online: some people linked the prickly interview answer with the "backstop" remark and concluded that OpenAI was insolvent and asking for a rescue. The answer given is categorical: there will be no federal bailout for AI. If one frontier company fails, others will replace it. The speaker calls AI perhaps one of the most competitive sectors of the American economy and says that if a company overextends and goes bankrupt, "the chips are gonna fall where they may."

The same speaker gives OpenAI the benefit of the doubt. They don't believe anyone there was asking for a bailout. In their reading, Friar was "clearly searching for the right word" and landed on one she now regrets. They call the episode "a tempest in a teapot." They then argue that the real policy question is making infrastructure easier to build, through faster permitting and easier power generation, without raising household electricity rates. They warn that fear of higher bills is starting a NIMBY backlash against data centers. Their proposed fix is regulatory reform that lets AI companies generate their own power "behind the meter," which they say the president has called for.

Build out, not bailout.