Back in October 2024, Poolside was an early AI star. Cofounded by former Github CTO Jason Warner, the startup had raised $500 million on a $3 billion valuation to build coding agents for governments and large companies. But over the next 18 months, Poolside largely disappeared from view, while OpenAI and Anthropic ballooned to nearly trillion-dollar valuations with a crop of Chinese labs building open source models nipping at their heels.
Now Poolside is back with a new model called Laguna that on public benchmarks beats its American and Chinese open source competition — with the very notable exception of Chinese lab Moonshot’s latest AI model, Kimi K3.
“As an American company building for the West, we’ll be the most capable open model in the West,” Warner, Poolside’s co-CEO and cofounder, tells Forbes. “Globally, in this weight class of the 118 billion parameter model, we are the leader.”
Warner claims that the startup spent those 18 months where it went quiet building the infrastructure for a “model building factory” that could pump out Laguna and continue with new and more powerful iterations every five weeks. “The classic notion of model building is an artisanal process…We’ve built an industrial model-building process,” he says.
He concedes that Laguna has a long way to go to match the capabilities of the top closed models from OpenAI and Anthropic, and the new version of Kimi that was dropped by Beijing, China-based lab Moonshot on July 16. But its models are significantly smaller. OpenAI and Anthropic have not disclosed the size of their latest models but Moonshot said that Kimi K3 was a 2.8 trillion parameter model— a whopping 20 times the size of Poolside’s new release. Warner says that Poolside plans to build bigger models that could in time compete with frontier labs.
The startup also launched a smaller series of models earlier this year without much fanfare. Warner said it had shared model weights from the get go with its small, undisclosed number of government and corporate clients but since April began sharing them publicly.
Poolside says it’s based in San Francisco, but it has more of an international flavor. Cofounders Warner and Eiso Kant live in Canada and Portugal respectively, with the majority of its team working remotely. The company in the past was reported to be headquartered out of Paris, France, where its coders would assemble every few months to regroup.
Silicon Valley is in a rush to build homegrown alternatives to Chinese AI models. These are often billed as open source but Chinese startups like Deepseek don’t publish their code or training data. Instead, they release what are called model weights, which allow models to be downloaded, run locally and fine-tuned by developers. That’s proven irresistible to startups and increasingly large companies that have chafed at the control OpenAI and Anthropic hold over their own closed models, and the mounting cost of using AI tools like ChatGPT and Claude.
Last week, former OpenAI CTO Mira Murati’s Thinking Machines Lab also released its first AI model, an open-weight model called Inkling. In a refreshing departure from the AI industry’s relentless hype machine, the lab said the model is “not the strongest overall model available today, open or closed,” but that it was meant to be flexible and easily customize-able.
OpenAI, Google and even Nvidia are also in the open-weight game (Meta once had one of the leading open models with Llama, but switched its focus to a new closed model called Spark last year). Few have come close to matching the capabilities of the top Chinese models, the most dominant of which is now from Moonshot. Kimi K3 outranked OpenAI’s top model and Anthropic’s Fable 5 at some coding tasks, according to AI leaderboard Arena. It’s still behind on some measures, but charges around two thirds less than Anthropic, and a third less than OpenAI, for the same task. It is so capable and cheap that its release on Friday last week spooked U.S. markets.
Over the last month, LLM marketplace OpenRouter reported that models from Chinese corporations like Xiaomi, Tencent and startups like Z-AI were the five most popular. For now, Kimi K3 is only available to run on Moonshot’s servers (located in China, making that a tougher call for U.S. companies). But it said that it will publish the model weights on July 27 which would allow companies to download and run it on their own systems.
There’s big money at stake for the small number of American challengers. Thinking Machines raised $2 billion on a $12 billion valuation in 2024. Last year, The Information reported that Poolside had tried to raise $2 billion, also at a $12 billion valuation. Reflection, another coding agent startup that pivoted to building open source models, raised $2.5 billion at a $25 billion valuation in March. It recently signed a $1 billion compute deal with neocloud Nebius as it builds its model. Reflection still has not released any products. “Our aim is to ensure that there is a competitive frontier open lab in the Western world,” cofounder and CTO Ioannis Antonoglou told Forbes last week. “We want to have a flourishing open ecosystem, and we want to contribute to it.”
They might look across the pond for inspiration. France’s Mistral made over $200 million last year from open-weight models targeted at Western companies and governments, with its primary selling point that it’s not Chinese (or American, for that matter).
Warner, Murati and other American open source builders could get a huge boost from the Trump Administration. Axios reported that the White House and Commerce Department had debated measures that would effectively ban the use of Chinese open source models by American companies.
“We are at a critical inflection point in AI policy. The leading closed labs, already a duopoly in terms of AI model revenue, want the government to eliminate their open-source competition,” wrote venture capitalist and White House AI adviser David Sacks on X.
U.S. open source is still an open field, especially since training models is such an costly proposition. Warner tells Forbes that Poolside hasn’t raised since 2024 (he declined to share the startup’s current revenue). The company also found itself involved in a complicated, and potentially expensive side project to build a 2 gigawatt facility in West Texas with AI data center giant CoreWeave dubbed Project Horizon. The Financial Times reported in April 2026 that the deal was scrapped after deadlines were missed (the original announcement blog post has since been deleted). Warner says that the project lives on as a standalone business called PIC (Poolside Infrastructure Company). This had been run by Poolside cofounder Kant, but he’s now handed the reins to former construction executive Robert Bonar.
Poolside might not need to raise again even with this pricey distraction, and a self-imposed deadline to build new and more powerful AI models. That’s thanks to a quiet and previously unreported investment that Poolside made in another startup, Fluidstack. The British neocloud got its start as an “Airbnb for GPUs” but has since started building its own data centers, landing deals with both Google and Anthropic.
It’s not clear when Poolside invested in Fluidstack, but British corporate filings show that the London-based company raised a $24.7 million SAFE round at some point in 2024. Later filings show that by March 2026, a Luxembourg-based company called Argo Participation I would become one of the largest shareholders in Fluidstack with a 15% stake. Warner and Kant are listed as directors of Argo Participation I in corporate filings in Luxembourg. A filing from November 2024 lists Poolside as the sole shareholder of Argo Participation I.
Fluidstack announced on Monday that it had raised $830 million in a round led by Leopold Aschenbrenner’s hedge fund Situational Awareness at a $7.5 billion valuation. That means Poolside’s Fluidstack investment is worth $960 million. Warner said he could not comment on Poolside’s investment in Fluidstack. Bloomberg reported in April that Fluidstack was in talks with investors to raise $1 billion at a $18 billion valuation.
