They've known each other since they were four years old. Now Eric Zhu and Ian Wang think they can convince some of the world's most skeptical traders to trust algorithms over broker calls.
The two former quantitative traders quit their jobs to build Axis, an AI platform aimed squarely at institutional commodity desks. Y Combinator unveiled the startup in late February as part of its Winter 2026 batch, though "unveiled" may be generous for a two-person operation with no disclosed customers yet. Still, the pitch lands with a certain logic: if large language models can parse legal documents and write marketing copy, why can't they digest the chaos of commodities markets?
Because that's what commodities trading still is, in many ways—organized chaos. Oil tankers crossing oceans. Weather patterns threatening wheat harvests. Broker gossip over instant message. A Fed official's offhand remark that sends natural gas futures lurching. Traders at major banks and hedge funds spend hours each day sifting through this noise, trying to separate signal from static.
Zhu and Wang believe the technology has finally caught up to the problem.
When Algorithms Meet Agriculture
Zhu came out of the University of Chicago with a math degree and landed in quantitative trading, where he watched colleagues drown in information. Wang, fresh from Yale in 2025, saw the same dysfunction from a different desk. Both noticed something: while equities trading had been algorithmically strip-mined for decades, commodities—covering everything from crude oil to corn futures to physical copper trades—remained stubbornly human.
Not because traders are technophobes. The sector has plenty of sophisticated quant strategies. But commodities markets generate a particular species of unstructured intelligence that resists easy quantification. A shipping delay in the Suez Canal. An energy minister's ambiguous statement. The subtle shift in tone from a broker who usually knows which way the wind is blowing.
"Deployable models," Axis calls them, a bit of startup jargon for what amounts to configurable AI agents that monitor markets continuously. Traders can tailor these models to track specific strategies or conditions relevant to their book. The system ingests structured data—prices, volumes, the usual—but also the qualitative stuff: broker color, logistics updates, geopolitical tea leaves.
Whether it works as advertised remains to be seen. The founders haven't disclosed customer names or pricing. As of mid-March, the company still lists just the two of them.
The Incumbents Aren't Sleeping
Axis is walking into a market dominated by entrenched giants with deep pockets and decades of client relationships. Bloomberg Terminals—ubiquitous on trading floors—command premium pricing that can vary by region and subscription tier. S&P Global Commodity Insights, the rebranded Platts, provides the price assessments and benchmarks that much of the industry uses as gospel. Vortexa and Kpler have carved out lucrative niches tracking cargo flows and inventories with satellite data and algorithmic analysis.
And those incumbents have noticed the AI moment. S&P Global announced partnerships with Google Cloud in August and again in December 2025, positioning its data feeds as "AI-Ready" and promising agentic automation across its commodity intelligence offerings. Translation: they're trying to get ahead of this too.
So Axis isn't promising to replace the data providers. The founders position their platform as something different—an analytical layer that sits atop existing information flows. Not another feed, but a synthesizer. Something that can take the broker chat and the shipping manifest and the macro narrative and produce insights that would otherwise require a senior trader and three espresso shots.
It's a narrow wedge, perhaps, but possibly a smart one.
The Relationship Business Meets the Algorithm
There's an almost cinematic quality to the origin story—two kids who met in preschool, ended up in the same cutthroat corner of finance, then left to build something together. In practice, though, that backstory matters less than the fact that both spent time inside institutional trading desks and recognized the same inefficiency.
Commodities trading has always been a relationship business. Who you know. Which brokers return your calls. The veteran trader who can read a market's mood from years of pattern recognition. That's not disappearing overnight, and Axis isn't claiming it will.
But adjacent markets have automated aggressively. Equities trading floors that once employed hundreds now run on skeleton crews and server farms. If commodities desks are still parsing broker chats manually while AI models get better at natural language understanding every quarter, well, the gap becomes harder to justify.
The question—and it's a legitimate one—is whether traders who've spent careers cultivating information networks will hand over analytical duties to a startup they've never heard of. Trust is currency in this world. A blown call based on an AI model's hallucination could cost millions.
Then again, the infrastructure is being built whether individual traders like it or not. Data providers are already repackaging their feeds for algorithmic consumption. The Y Combinator stamp, thin as it may be for a brand-new company, at least signals that someone with pattern-recognition skills thought the idea had merit.
Zhu and Wang are betting that institutional desks don't just want more data—they want something that thinks like they do, processes information the way they would if they had infinite time and attention, but faster. Whether commodities traders are ready to trust that remains the gamble.
For now, it's two guys and a pitch. The markets will tell them if they're right.
