Sam Altman disclosed on October 30th that OpenAI has a trillion dollars in capital-expenditure obligations to fund. Anthropic, Ram Ahluwalia says, has less than that but still needs to manage toward free cash flow before it goes public.
Everyone read Dario Amodei's letter about slowing AI down because of catastrophic risk as a safety statement. Ahluwalia reads it as an IPO strategy.
"We're not facing any kind of imminent catastrophic risk."
Ahluwalia runs Lumida Wealth Management as a macro hedge fund and was one of just three self-described AI bulls at a dinner last year that included Jordi Visser, the former chief risk and investment officer at Weiss Multi-Strategy.
The full segment is covered here so you can skip it. 17 minutes of audio, 7 minutes of reading.
Here are the 6 takeaways that matter.
👤 Speaker: Ram Ahluwalia, CEO of Lumida Wealth Management
📰 Published: 16 September 2026 on the Lumida Wealth YouTube channel
🔴 YouTube | ⏱️ 17 min | ✅ Time saved: 10 min
Key Takeaways
He doesn't think a Fed rate hike is a foregone conclusion, even with markets pricing 90% odds of one
He calls Kevin Warsh's AI-productivity talk "jawboning," a soft-spoken threat rather than a real signal
Anthropic and OpenAI's "slow down" letter is really about managing an upcoming IPO, not existential risk
Both firms carry huge capex obligations — Altman disclosed $1 trillion — and going public means investors will punish any stock-price drop that puts employee RSUs underwater
He isn't worried about semiconductor demand at all, slowdown letter or not
"Any chip that flies off of TSM's conveyor belt is going to get bought," with Meta ready to absorb whatever capacity is freed up
Frontier-lab insiders are privately telling him AI progress is slower than the public narrative suggests
He calls chain-of-reasoning and agentic loops clever engineering "hacks," not breakthroughs on the level of the original transformer paper
He sees the timing of the safety letter as tied to the midterms
Getting ahead of regulation with self-imposed rules, he says, beats having rules imposed after Democrats potentially retake the House
His bottom line: real capability gains are real, but the catastrophic-risk framing is overstated
He warns that overstating the danger risks undermining public adoption of what he calls a "truly transformational technology"
1. 90%-Priced Hike, Not a Lock
Ahluwalia opened with rates, ahead of that week's Federal Reserve decision.
He said the market was pricing a 90% chance of a hike, which he called "largely priced in," and framed the real swing factor as what happens if the Fed does not hike at all. "If you don't get a rate hike, markets are going to rally quite a bit."
If a hike does come, he said what actually matters is the forward guidance behind it — whether it reads as one-and-done, a single adjustment, or the start of a series of increases.
He read the committee as split: Christopher Waller dovish in recent weeks, Neel Kashkari reliably hawkish in a way Ahluwalia says doesn't track with someone who has spent his career mostly in bonds rather than stocks, and Jerome Powell not a hawk. He described Fed Chair Kevin Warsh's talk of AI productivity gains as "jawboning" — a soft-spoken threat rather than a real policy signal.
He repeated a standing view: the Fed should never have cut rates last year, when the economy already had momentum from capex and government spending. His metaphor was a massive oil tanker — you don't yank the wheel on an economy that size, you make deliberate, longer-term adjustments. He also flagged the politics: hiking this close to the midterms doesn't make sense to him, so any move now is more likely to just partially undo last year's cuts.
2. The Letter as an IPO Move
Dario Amodei's letter calling for a pause on AI development because of catastrophic risk drew quick, coordinated-looking support from Sam Altman and Elon Musk — but not from Mark Zuckerberg.
Ahluwalia said the near-simultaneous agreement from multiple lab heads signals the message was socialized beforehand — "conversations were had, messages were sent, everyone was aligned." Zuckerberg was the exception, publicly arguing Meta needs to move faster on AI, a stance he tied to comments from Trump and a call Zuckerberg made into Jensen Huang's phone on stage at the Olan Summit, where Ahluwalia estimated the combined market cap on display at roughly $7 trillion to $8 trillion.
His read on why Anthropic and OpenAI are the ones pushing "slow down": both are staring down enormous capital needs. Sam Altman disclosed on October 30th that OpenAI has $1 trillion in capex obligations to fund; Ahluwalia said Anthropic's load looks smaller, which he thinks is part of why its Claude Code product runs slower than OpenAI's Codex.
Going public forces financial discipline that private markets don't demand. Private venture capital rewards storytelling and momentum; public markets punish it, and investors are already watching how AI-adjacent names trade after their IPOs.
He connected this directly to talent retention: OpenAI has already lost people to Anthropic, including Amodei himself, and a post-IPO stock-price drop would leave employee RSUs underwater, risking another wave of departures.
3. Not Worried on Chips
Whatever a "slowdown" in AI development might mean, Ahluwalia said it doesn't change his view on semiconductor demand.
"Any chip that flies off of TSM's conveyor belt is going to get bought." He said the same is true of Micron's and SK Hynix's output, and that this is what ultimately drives the forward path of chipmaker earnings, regardless of which lab is talking about slowing down.
He argued Meta alone would absorb whatever spare compute capacity opens up, and that a multi-year "slowdown" would still leave years for new entrants, open-weight models, and sovereign buyers — he named the US government and Saudi Arabia — to keep demand growing.
He estimated US small-business penetration of these AI models at around 20%, calling it "a long ways to go," and pointed to humanoid robots as a demand source that hasn't even shown up in the numbers yet. "I'm not worried about this semi compute story at all."
4. Progress May Be Slower
Ahluwalia said contacts at frontier labs have privately told him AI development is progressing slower than the public narrative suggests — and used that to separate real breakthroughs from engineering workarounds.
He cited a debate over how to define AGI, referencing a claim from Jordi Visser that today's AI operates around a 150 to 160 IQ. Ahluwalia's own practical test is whether AI can do real work — yes, he said — versus whether it can fully substitute for human labor, which he said it still can't, though agents doing so for narrow roles like UX design could arrive as soon as next year.
He called the two big recent innovations — chain-of-reasoning and agentic loops — clever engineering "hacks" rather than genuine breakthroughs, unlike the original transformer paper. "It's like duct tape. It's practical. It unlocks value. It's not an incredible breakthrough."
He credited Meta's AI head — brought in to replace Yann LeCun — with shipping Muse, a fast-response AI platform built using looped AI pairs, where one model evaluates another's output repeatedly until it improves. He noted Meta claimed these setups produce output equivalent to 100 engineers.
He put reinforcement learning in the same bucket: a well-established machine-learning concept rather than a novel discovery, alongside scaling laws — more data, more training, more compute — which he said will keep improving models regardless.
5. The Politics of the Letter
Ahluwalia tied the timing of the safety letter directly to the midterm elections.
He said Amodei, Altman and Musk are trying to "take the venom out of the sting" by embracing self-regulation before regulation gets imposed on them — a move he called smart, especially with betting markets suggesting Democrats could retake the House. He thought the messaging itself was badly handled, saying "there's still way too much drama around this stuff."
He said he saw only clips rather than the full discussion, but that David Sacks's take on the controversy struck him as generally accurate.
6. No Catastrophic Risk
Ahluwalia closed by separating what he thinks is real from what he thinks is overstated in the safety debate.
His closing line was that the capability gains are real, but the emergency framing isn't. These models are already great at generating code and will unlock enormous opportunity, he said, without the kind of imminent catastrophic risk the letter implies.
He warned that overcorrecting the public narrative has a cost: family members had asked him over the weekend whether AI posed the dangers described in the letter, and he worries that tone undermines adoption of what he called "a truly transformational technology that will raise living standards for everybody."
Bonus Insights
Ahluwalia said he was one of just three self-described AI bulls — alongside Jordi Visser and an investor he referred to only as "Pomp" — at a dinner organized by a friend last year, and reads a two-sided market like that as a healthy sign rather than a warning.
Products, Companies & Tools Mentioned
Anthropic (The lab behind the "slow down" letter, which Ahluwalia reads as partly a pre-IPO capital-discipline move)
OpenAI (Disclosed $1 trillion in capex obligations; competes with Anthropic for capital and talent)
Meta (The lab that declined to sign onto the slowdown message and shipped the Muse platform)
Nvidia, TSMC and Micron (The chip supply chain he says will keep selling out regardless of any AI "slowdown")
SK Hynix (Named alongside Micron as a memory-chip maker whose output he expects to keep selling)
Watch the full episode:
If this was worth your time, send it to someone who has to have a view on this.
Get the latest market chatter as it happens: