Max Mergenthaler Canseco has a timing problem—ironic, given his company forecasts the future for a living.
On February 5, his San Francisco-based startup Nixtla closed a $16 million Series A led by Energize Capital, with True Ventures and GreatPoint Ventures joining the round. The capital injection comes at a precarious moment: Google, Amazon, and Salesforce have all begun folding time-series forecasting into their cloud platforms, threatening to commoditize the very market Nixtla spent five years cultivating.
Yet Mergenthaler Canseco isn't backing down. His pitch centers on something the hyperscalers can't easily replicate—a hybrid model that pairs proprietary AI with radical deployment flexibility. Nixtla's customer list suggests the strategy has legs. Microsoft distributes one of the startup's foundation models through its Azure AI catalog. Zalando and Decathlon use its forecasting tools. Even Lyft has signed on, claiming an 85% drop in false positives after switching to Nixtla's platform.
Still, in enterprise software, traction is one thing. Survival against trillion-dollar competitors is another.
The Open-Source Gambit
Nixtla didn't start by chasing Fortune 500 contracts. The company spent years building credibility from the bottom up, releasing a suite of open-source forecasting libraries—collectively dubbed the "Nixtlaverse"—that data scientists could download and tinker with for free.
The approach worked, at least by one metric. Those libraries have been pulled down 41 million times, according to the company, accumulating 15,000 GitHub stars along the way. That's the kind of grassroots adoption that opens doors when sales reps eventually come calling.
At the core sits TimeGPT, a 500-million-parameter model pretrained on what Nixtla claims is 100 billion tokens of temporal data. The company says the model delivers forecasts up to 42% more accurate than legacy methods, with inference speeds ten times faster. Whether those benchmarks hold in messy real-world deployments—think supply chains with spotty data or retail inventories subject to Black Friday chaos—is harder to verify. Nixtla provided case studies but not independent audits.
What's clearer is that Nixtla found itself in a suddenly crowded field. Last July, Google embedded its own TimesFM model directly into BigQuery ML, letting existing cloud customers forecast without leaving their analytics dashboards. Amazon followed suit, shipping its Chronos model through SageMaker JumpStart. Salesforce then released Moirai, a time-series suite with what it calls "agentic capabilities"—automation features meant to eliminate manual tuning.
All three offerings share a key advantage: they're already bundled into platforms enterprises use daily. Nixtla, by contrast, requires a separate procurement conversation.
Flexibility as Differentiation
This is where Nixtla is making its stand. The startup's Enterprise 2.0 product functions less like a single model and more like a "model zoo," offering customers a menu that includes not just TimeGPT but also Amazon's Chronos and Google's TimesFM—accessible through one API.
More importantly for regulated industries, Nixtla supports self-hosted and on-premises deployments. That matters in healthcare, where HIPAA compliance often means data can't touch third-party clouds, or in European enterprises navigating GDPR's strict data residency rules. Hyperscalers can offer compliance certifications, but migrating sensitive datasets to their infrastructure is a harder sell than keeping everything internal.
Mergenthaler Canseco framed this in the funding announcement as essential for "production-ready systems." Translation: enterprises need forecasting that plugs into existing workflows, not science experiments that require data engineering overhauls.
The company has also built an automation layer atop its forecasting models—what it calls an "MCP-based agentic" system that strings together data ingestion, model selection, and output delivery without manual intervention. In theory, this reduces the back-and-forth between data scientists and business users. In practice, the jury's still out on whether these automation promises deliver or simply shift complexity elsewhere.
The Capital Question
Juan Muldoon, the Energize Capital partner who led the round and joined Nixtla's board, comes from a particular vantage point. His firm manages over $1.8 billion focused on climate and industrial software—sectors where energy forecasting isn't a nice-to-have, it's operational bedrock.
"We see time-series as foundational to energy transition," Muldoon said in a statement, highlighting use cases in grid optimization and renewable generation forecasting. That lens may explain why Energize committed while others hesitated. The firm isn't betting on Nixtla outrunning Google; it's betting the startup can carve out defensible niches in verticals where specialized deployment trumps general-purpose tools.
Nixtla plans to funnel the Series A into engineering hires, product expansion, and a sales buildout. The company also pledged continued investment in its open-source libraries—a promise that doubles as customer acquisition and a hedge against irrelevance. If the commercial product struggles, at least the Nixtlaverse keeps developers engaged.
Founded in 2021, Nixtla previously raised a seed round from True Ventures in April 2023 and secured a grant from Google's Latino Founders Fund in 2022. Fast Company named the startup to its 2024 "Next Big Things in Tech" list.
A Footnote Worth Noting
One detail doesn't quite add up. CB Insights, the venture tracking database, lists a "Series B" round on December 16, 2025—roughly seven weeks before the Series A Nixtla just announced. The earlier round supposedly included the same lead investors: Energize and True Ventures.
Nixtla hasn't commented on the discrepancy. It's possible CB Insights misclassified the round, or perhaps the December date reflects a convertible note that later rolled into the Series A. Or maybe something more complicated transpired behind closed doors.
For now, the company is sticking with its Series A narrative. Whether that precision matters depends on who's asking—and whether Nixtla can convince enterprises its forecasting edge is worth the switching costs before the hyperscalers close the gap entirely.
