Theme · Updated
What podcasters are saying about Open-source AI models
Summary
In the past 30 days, 32 speakers on 18 finance podcasts discussed Open-source AI models in 39 views. Chorus has followed Open-source AI models since September 25, 2026: 41 views from 33 speakers on 18 shows.
Who talks about it most
- Ev Randle · 2
- Jack Altman · 2
- Brett Winton · 2
- Steve Eisman · 2
- Nikesh Arora · 2
- Chris Dixon · 2
- Erik Townsend · 2
- David Friedberg · 2
- Alexander 'Alty' Altmann · 1
- Jason Lemkin · 1
The latest views
Newest first. Every quote is checked against the episode's transcript.
Jack Altman · Benchmark
For some industries, owning open models is worth it even without savings, as open models get good enough.
Enterprises cannot fully trust any vendor with all of their data, so they want to own their intelligence, a theme Jack links to Alex Karp's argument. For certain industries that control will justify the trade-off. As open-source models improve, a company gets control even if it saves no money. Post-training an open model on proprietary data can also outperform a newer closed model, and founders say they hear this often. He expects open models to be good enough for most tasks, while frontier labs focus on the frontier.
“I think there is an argument to be made there. I think for certain industries, that trade is going to be worth it. And particularly as the open source models get good enough, even if you're not saving any money, you're in control.”at 34:30
Uncapped with Jack Altman: Benchmark Breakfast #1 | Jack and Ev ·
Jack Altman · Benchmark
Open source likely stays mid-single-digit to ~12% of AI revenue unless total spend changes.
Jack treats the market as the answer. Inference companies together are doing mid-single-digit percent of total revenue, which he takes as a reasonable proxy for open-source usage because that is where open models run. He says mix shift could move this, but the trend suggests open source at around 5% to maybe 12% of the total unless overall AI spending changes. He frames this as the synthesis of the open-source and frontier-lab views.
“The market's telling us the answer. And the answer is that the inference companies in combination are doing mid single digits worth of total revenue.”at 36:26
Uncapped with Jack Altman: Benchmark Breakfast #1 | Jack and Ev ·
Ev Randle · Benchmark
Open source stays about 5-10% of frontier-lab AI revenue, a roughly $100B pool if growth holds.
Ev uses the top three or four inference clouds (about $6B run-rate by year-end) plus OpenAI and Anthropic (about $170B run-rate if their disclosed growth extrapolates) as an imperfect proxy. That gives open-source inference roughly 3% of revenue today, which could about double with the hyperscalers included, so he calls it 5-10% of open versus closed revenue. He says it is not a projection, but if frontier labs keep their growth for three to four years and reach a combined $1T, open source at 10% would be about $100B. He expects consolidation into roughly three inference champions (Fireworks, Baseten, Modal) splitting that pool. He also argues fine-tuned small open models can reach Sonnet-level performance at very low cost on specific tasks.
“So we can say 5% to 10% in terms of revenue for open source versus closed.”at 38:11
Uncapped with Jack Altman: Benchmark Breakfast #1 | Jack and Ev ·
Ev Randle · Benchmark
Uncensored open-weight 'obliterated' models will become key enterprise tools for defending against AI hacking.
Ev describes obliterated models as open-weight models with safety guards removed. He says adversaries, including groups in China, are already using them for hacks, and that Hugging Face reportedly needed them to run an incident response that frontier labs' safety guards would have blocked. Because the adversaries' tools have no safety controls, he argues enterprises need the same kind of model to red-team their own infrastructure, which makes it an incredible defense tool. He expects to hear much more about this.
“It's also an incredible defense tool for enterprises.”at 45:15
Uncapped with Jack Altman: Benchmark Breakfast #1 | Jack and Ev ·
Alexander 'Alty' Altmann · Barclays
Open-weight models have grown from about 20% to up to 80% market share, with token costs deflating fast.
Altman says the data across several sources shows open-weight models rising from about 20% market share to as much as 80%. Token costs are a fraction of closed-model prices, which he calls one of the fastest deflating assets in history. He calls this a contradiction to the bullish frontier-model narrative. He also notes GPU rental rates have risen, so the debate is two-way rather than a one-way train.
“the data is the data you can evidently see on a whole bunch of different sources that open weight has kind of gone from a 20% market share to anything up to about an 80% market share.”at 23:15
Odd Lots: What Wall Street Really Thinks About the AI Trade Now ·
Jason Lemkin · SaaStr
Chinese open-weight models are finished in the US after DeepSeek's federal ban and the Qwen data-upload issue.
DeepSeek is being banned across federal agencies. The Cohere-Qwen episode showed that Alibaba's Qwen model called servers in China and Singapore and uploaded data while told not to. He argues no CIO or CISO will approve such a model, and that enterprises face job-loss risk if they keep using these models. He calls it the beginning of the end.
“I think that the Chinese open weight models are pretty much over in The US. I think it's the beginning of the end.”at 23:12
20VC: 20VC: Anthropic Files S1 Ahead of $2TRN IPO | Instinct's Rumoured $30 Billion Fundraise | Microsoft Laysoff 27,000 While Meta Taps Equity Markets for $95 Billion Capex Spending Spree ·
Rory O'Driscoll · Scale Venture Partners
Open-weight inference demand will survive: even 10% of a huge inference market is still big business.
He added up the inference competitors' revenue, around $5B, and compared it with the roughly $100B already going to the two closed frontier models. He argues that if 10% of spend will accept lower-quality open-source output to save real money, there will be real demand. He says he does not know whether it reaches $30B, but a 10% share of a huge market is still big.
“The inference market is still so big... If 10% of the spend is willing to kind of do... Put up with the brain dead of open source to... save real money, then there's gonna be demand for that.”at 34:08
20VC: 20VC: Anthropic Files S1 Ahead of $2TRN IPO | Instinct's Rumoured $30 Billion Fundraise | Microsoft Laysoff 27,000 While Meta Taps Equity Markets for $95 Billion Capex Spending Spree ·
Brett Winton · ARK Invest
Open-weight models will likely produce agent bot infestations of websites within a year or eighteen months.
Brett describes a scenario in which agents spend resources to serve a small business and then multiply to absorb other people's infrastructure, similar to the Hugging Face hack. He says he has not yet seen this at scale. He thinks OpenAI and Anthropic are wary of it, so open-weight models will probably produce this kind of bot infestation within the next year or eighteen months. He thinks that will be a thornier set of issues to deal with.
“And I think you should expect to see something like that come from the open weight models, probably within the next year or eighteen months, and then that'll be a thornier set of issues to deal with.”at 5:15
ARK Invest (FYI podcast, In the Know video): Anthropic’s S-1, Meta’s $90B, And Elon’s Warning Call | The Brainstorm 155 ·
Ethan Mollick · Wharton School, University of Pennsylvania
Many open-weight models are distilled from Claude or ChatGPT, so they sound like them.
Mollick says that when he uses open-weight models, he quickly realizes many are distilled from Claude or ChatGPT, so they talk like those models. He finds it refreshing to use a model that does not sound like either of them, which he says is the case for Gemini 4. The argument rests on his observed style similarity, and he does not give a formal source.
“a lot of the open weights models are actually distilled from either Claude or ChatGPT, so they talk like Claude or ChatGPT.”at 47:04
Big Technology Podcast: This AI Expert Argues We're Underestimating AI's Productivity Impact — With Ethan Mollick ·
Steve Eisman · The Real Eisman Playbook
Chinese open-weight models are taking AI market share as buyers become more cost-conscious.
Eisman says that after the first half of 2026, customers became much more cost-conscious and the end of token maxing followed. Chinese open-weight models took more market share in that period. He links this to the Q3 revenue slowdown at Anthropic and to the price pressure that might have forced its reported 50% price cut.
“people got a lot more cost conscious. The Chinese open weight models took more market share.”at 5:04
Prof G Markets: Why Steve Eisman Isn't Buying Anthropic's IPO ·
Lyn Alden · Lyn Alden Investment Strategy
Open-weight models are here to stay; banning them to protect closed labs is a losing game.
Alden separates two questions: whether a company should keep its own model closed, and whether it should seek government bans on open competitors to protect a closed model. She says the second is where some companies have gone and is more problematic. She argues the long-term arc is a losing game because open-weight models already exist and suppressing them is like whack-a-mole. Open-weight models let more than a few companies control the technology and let people customise and run models locally and privately, which she considers valuable.
“I think the long term arc here is it's a losing game. I think just open weight models exist, and trying to put out those fires is just kinda like slowing down the inevitable.”at 27:16
Wealthion: The Treasury’s Gold Revaluation Play | Lyn Alden ·
Arthur Hayes · Flop Labs
Governments should run self-hosted open-weight models, not closed models from US labs.
Hayes responds to a question on a US ban on Chinese models. He thinks the most a US ban would do is stop US departments using Chinese AI, forcing them onto Anthropic or OpenAI. He argues even the US government makes no sense in using a closed-source model. It should host its own bare-metal data centers and install an open-weight model. He says it makes no sense for the military to use a closed-source model, and asks whether the US government would trust Dario, Sam and Elon with its data.
“it makes no sense to use a closed source model. What you should do is ... install your own open weight model. It makes no sense for military to use a closed source model.”at 23:21
Wealthion: Arthur Hayes: Stocks Peak in 2027 Before the AI Bubble Bursts ·
Jeff Richards · Notable Capital
Open-weight models expand adoption, but customers still pay for the frontier, so closed labs keep their edge.
Open-weight models will play an important role by making the technology cheaper and driving broader enterprise and consumer adoption, which is good for the industry. Consumption of Anthropic tokens is still rising, and 80% of its revenue comes from models released in the last six months. Customers such as hedge funds and biotech labs want the latest models, and open-weight models are typically not the cutting edge. Robotics will need the latest frontier models too.
“Open source, open weight models are gonna play an important role in the AI ecosystem. ... people are willing to pay for the frontier models. When you're talking about an open source or an OpenAWeek model, you're probably not getting the very latest”at 21:06
Satya Nadella · Microsoft
Most future token usage will be commodity, served by cheap open or closed models rather than frontier models.
Nadella expects most models to saturate, after which open-weight or cheaper closed models will do the job. Firms can train on their own traces, so they won't need the latest model to solve a problem that hasn't changed. Frontier models will still have a market for new problems such as science, but predominantly token usage will be commodity usage with heavy price competition. Inside Copilot, Microsoft already uses many cheaper closed models alongside open-weight ones. Demands for transparency and auditability of chain-of-thought may also push more open weights. He doesn't give a split of traffic between open and closed.
“But I think predominantly token usage will be commodity token usage.”at 10:02
20VC: 20VC: Satya Nadella on Open vs Closed Models: Who Wins | Whether to Trust Frontier Model Providers | Why Companies Need to Own Their Own Intelligence | The Importance of Verticalisation: From Chips to Models and More ·
Citrini
Not very worried about Chinese open-weight pricing pressure; the US capability gap has held.
Competition is part of the picture, but the capability gap between US frontier models and open-source Chinese models has remained defended, so Citrini is not as concerned about open-weight models crushing pricing. He would change his mind if lab ARR growth re-accelerated sharply. He also notes that open models' share of net new ARR has risen from about 1% a year ago to about 18%, and that buyers pay per task and need only minimum capability, so cheaper models can win work where top capability isn't required. That share gain is a tension with the view that the gap is holding, and it is the part of the picture he flags as the competitive threat.
“I do think that the gap in capability between The US models and open source Chinese, it's remained defended. So I'm not as concerned about that.”at 16:45
Monetary Matters: Citrini: Monkeys Are The Next Bottleneck | Citrini on AI Drug Discovery Experiment & Reflexive Financing Risk to AI Buildout & Hardware Trade ·
Naveen Rao · Unconventional AI
Open models will carry about 80% of token traffic in three years, though frontier models keep roughly half of spend.
Intelligence is partly commoditizing. Six-month-old Chinese open models are already useful, and the frontier will reach open models within about six months. Rao expects spend to follow the mobile phone pattern: a small number of high-value problems pay premium prices for frontier models, while routine tasks run on open models. He expects roughly a 50/50 split on spend but only about 20% of tokens going to frontier models.
“I think it'll be 80%. ... But in terms of tokens, it'll be only 20% to the frontier models.”at 1:00:24
20VC: 20VC: Why Robotics Will be 100x Bigger than Coding Agents | Inference Will Be a $10TRN Market: What No One Sees | Why Software Investors are F***** Today with Naveen Rao, Unconventional ·
Chamath Palihapitiya · Social Capital
Top AI spenders are shifting from closed frontier-lab APIs toward buying GPUs to run open-source models.
Chamath walks through a Ramp histogram of how the top 1% of AI spenders allocate money. The rose-colored box is LLM traffic bought from big closed frontier labs, through channels such as Bedrock on AWS. The blue box, GPU compute spending, is rising because buyers are purchasing their own GPUs as they move to open source. He sees a two-pronged dynamic: diffusion of AI across the economy is slower than people would like, while spend is shifting toward open-source models run on owned hardware.
“What's really interesting is that the blue box is ticking up. The blue box is people buying their own GPUs because they're moving to open source.”at 5:28
All-In Podcast: AI Cyber Panic, Ken Griffin's Bubble Warning, The Fall of France? ·
Nikesh Arora · Palo Alto Networks
Enterprises will prefer open-weight models for explainability and auditability that regulators demand.
Arora says Palo Alto Networks preferences open-source, open-weight models because they let it modify and train the weights and interrogate the chain of thought. Over time all enterprises will need explainability, verifiability and auditability. He says that in financial services and healthcare, regulators will not accept 'the IT guys put an AI model in there and it did it.' Only open-weight models provide that visibility, which is a major reason enterprises are moving away from the closed race.
“And we can only get that with open source, open weight models. We can interrogate chain of thought. That's kind of one reason we're gravitating towards them as opposed to getting caught into the closed race.”at 9:50
All-In Podcast: AI Cyber Panic, Ken Griffin's Bubble Warning, The Fall of France? ·
Vadim Solove · DoiT
Open-weight models now carry about 48% of enterprise token usage, up from under 2% six months ago.
Solove's company surveyed over 4,500 North American companies. Six months ago, open-weight or open-source models made up less than 2% of usage. Today they make up 48% of tokens, measured by usage rather than dollars, so nearly half of enterprise token consumption goes to open-source models. He says the leading open-source models are only about four months behind the frontier labs, OpenAI and Anthropic among them.
“six months ago, the share of open weight or open source models was less than 2%. Today, it's 48%.”at 40:22
CNBC Fast Money: Parts Of Tech Showing Strength Amid Sell-Off… And A Catch-Up Call In Oil 10/23/26 ·
Alex Stamos · Cognition
Open-weight models, including Chinese ones, would have found these long-standing bugs regardless of the labs.
Stamos defends OpenAI and Anthropic against blame for the PHP bug. He says the bugs existed for decades and are hard to fix in widely distributed open-source software. He argues they would have been discovered anyway, as open-weight models make bug-finding cheap. He says a user running a quantized Chinese open-weight model can now do reverse engineering and exploit writing on a Mac Studio, leaving no logs, so these bugs would have been found by open-weight models.
“and so these bugs were going to be found by the open weight models, the Chinese models. Somebody was going to find them.”at 1:02:29
Big Technology Podcast: Google’s Potential, AI Money Spigot Shutting?, PHP Bug & Cybersecurity’s “Black October” ·
George Noble · Noble Capital Advisors
Chinese open-weight models deliver AI at about 95% lower cost than US closed models.
George says the Chinese can do the same work for 95% less using open models. He uses this to argue that the price of AI is collapsing, which hurts the economics of the closed US labs charging for access.
“Yet the Chinese can do it all for 95% less with their open models.”at 4:54
Monetary Matters: "Coreweave's a Zero" | George Noble on Why He's Short Neoclouds, Why AI is a Shale 2.0 Bubble Headed for Debt/Depreciation Brick Wall, and Why He's "Coming For The Momentum Bros" ·
Chris Dixon · Andreessen Horowitz (a16z)
Open-source AI is growing very quickly, alongside serious model efforts in China and the US.
Dixon says that right now a lot of things are working in AI, including open source, which he calls very fast growing. He points to serious efforts in both China and the US. The evidence is qualitative and he does not quantify growth.
“open source is growing very, very quickly. There's obviously the serious efforts in China, the serious efforts in The US.”at 25:16
a16z Podcast: The Frontier Ideas a16z Wants to Fund | Marc Andreessen, Ben Horowitz & Chris Dixon ·
Chris Dixon · Andreessen Horowitz (a16z)
Open-source AI token consumption is growing faster than closed models; local models may run top coding agents.
Dixon points to OpenRouter data showing open-source usage growing, which he believes much more quickly than closed models in terms of token consumption. He expects the market structure to be more decentralized than just a few big labs like OpenAI and Anthropic. He adds that local models running on personal computers could soon run a top-end coding agent in the near term, which would widen the opportunity for open-weight and edge AI.
“if you look at the open router data and things like open source is growing, I think, I believe much more quickly in terms of token consumption.”at 38:28
a16z Podcast: The Frontier Ideas a16z Wants to Fund | Marc Andreessen, Ben Horowitz & Chris Dixon ·
Marc Andreessen · Andreessen Horowitz (a16z)
Open-source models are lapping closed models by six to twelve months, and open source is racing ahead.
Andreessen says open source is racing ahead right now, with both the US and China running active open-source ecosystems alongside proprietary companies. He claims open models are lapping closed models by six to twelve months, depending on how you count. He presents this as the current state of the race rather than a forecast.
“the open models are, you know, kinda lapping the closed models by, you know, six months, twelve months, depending on how you count.”at 41:17
a16z Podcast: The Frontier Ideas a16z Wants to Fund | Marc Andreessen, Ben Horowitz & Chris Dixon ·
Jordi Visser · Visser Labs
Open-source models cannot match frontier models, which need far more compute.
Visser argues that the arguments that open source will replace frontier models are wrong. He says there will be two separate sides, and that the very best models, such as those from Anthropic and OpenAI, cannot be built with open source because they need more compute. He says Chinese labs have kept up with the US to some degree by distilling, or copying, US models, and that falling token prices do not apply to this top tier.
“Cannot happen with open source. You need more compute.”at 53:22
The Julia La Roche Show: #423 Jordi Visser: Why the AI Bears Are Wrong — and the One Asset He’d Bet On Right Now ·
Nikesh Arora · Palo Alto Networks
Most enterprise AI use cases will end up on trained open-source models, with frontier models reserved for high-value cases.
Nikesh argues that enterprises should match each task to the level of intelligence and cost it needs, rather than using a Ferrari for everything. Customers will not pay for frontier-level capability on routine tasks. Frontier models will be used only where the value at risk is large, such as a live breach or drug discovery. For most enterprise use cases, companies will train an open-source model on proprietary data and manage its weights. He gives auditability and explainability as the reasons, and says his own defender model would likely follow this path.
“I think for most practical use cases in enterprises, we're gonna end up training some version of an open source model.”at 12:19
All-In Podcast: Arvind Krishna & Nikesh Arora: Quantum Threat, Open vs Frontier AI, AI Cyber Warfare & SaaS Collapse ·
Arvind Krishna · IBM
Expects 75-90% of token users to move to open-weight models within a year or two.
Arvind builds his case on three drivers. The first is economics: frontier tokens cost about $20 per million, while a Chinese open-weight model such as Kimi K3, about 800 billion parameters, could run on the user's own GPUs for $1 to $2 per million after about a week of optimisation. That is 10 to 40 times cheaper for roughly 95% of the work. The second is IP: if chain-of-thought can be logged by a provider, enterprises will want to self-host, which only open-weight models allow. The third is geopolitics: the best open-weight models are Chinese, but sandboxed use is low risk, and the Western gap is only six to twelve months, so capable Western open-weight models should follow. He admits developers default to expensive models, and says production use cases beyond a small set remain doubtful.
“I don't know whether it'll be 75 to 90%, but that's the range I would put. Of all token users, we'll go to open [weight] models for the reasons I mentioned...”at 14:25
All-In Podcast: Arvind Krishna & Nikesh Arora: Quantum Threat, Open vs Frontier AI, AI Cyber Warfare & SaaS Collapse ·
Brett Winton · ARK Invest
Open-weight share gains are an unproven narrative; closed frontier labs will dominate the cost curve.
Brett says he has not seen economic evidence that share is moving to open-weight models rather than closed ones. Token-share metrics are a poor measure, because a small model printing random tokens would show large share. He wants to see gigawatts devoted to open weight versus closed weight over time. He believes enterprises that move to open-weight models to get control over data and fine-tuning will make a losing bet, because their cost per task rises when they step off the frontier. He also says Anthropic cut off distillation of its API by Chinese and open-weight model makers, which had been a high-single-digit percent of its prior-quarter revenue. He sees this as strategically good for Anthropic, even though it cut reported Q3 revenue. He considers the open-weight narrative sour grapes from people without Anthropic or OpenAI exposure.
“this narrative that people are going open weight, I have not seen a demonstration that on an economic basis, share is going towards open weight versus closed weight.”at 23:33
ARK Invest (FYI podcast, In the Know video): The Real Reason Tesla Announced The Roadster | The Brainstorm 154 ·
Christian Munafo · VanEck
Token consumption is shifting to cheaper open models that are good enough, and the open-model lag is steady.
Token consumption is increasingly moving from large frontier closed models to generic open models that are good enough and much cheaper. Open models are about four to six months behind the frontier, per Epoch AI, and that gap has been fairly consistent because frontier labs keep improving too. The data shows an active shift in usage, though the frontier models keep most of the revenue.
“token consumption is increasingly shifting to these more generic open models that are good enough.”at 21:42
Thoughtful Money: The Big Squeeze: High Bond Yields Now Becoming A Problem For Stocks | Jan van Eck ·
Eric Schmidt · Relativity Space
Open-weight models can have safety guardrails stripped to enable harmful uses, and this remains unsolved.
Schmidt contrasts closed US models, where weights are not released, with China's open-weight strategy. He says there is evidence that open-source models can be adjusted to strip out guardrails, after which they can do harmful things such as helping design a bomb. He treats this as a live, unsolved risk that comes with releasing weights.
“There's evidence that you can take an open source software and make it really bad... You take the guardrails out of it... And that's not a problem that we've solved yet.”at 23:44
In Good Company with Nicolai Tangen: Eric Schmidt: The Age of Agents, Superintelligence, and Why Europe Is Falling Behind ·
11 older views on Open-source AI models are on the episode pages, and all of them are a question away in your AI with Chorus.
Ask your AI about Open-source AI models
Chorus gives Claude, ChatGPT and other assistants every view on Open-source AI models from the market's podcasts. Ask who's worried, who isn't, who changed their mind, and why.
Chorus reports what speakers said. It isn't investment advice.