VCs continue to pile into AI inference chip startups

Rebellions, a South Korean chipmaker, is at least the ninth company specializing in AI inference to announce new funding so far in 2026 as VCs continue to pile into the category.

Investors’ interest is largely driven by a simple thesis: As AI moves from experimentation into real-world deployment, the infrastructure layer that serves output—inference—is shaping up to be a recurring-revenue business.

“Inference represents this transition from experimental AI to the delivery of real-world value at a massive scale,” said Mo Jomaa, Partner at CapitalG, which has invested in Baseten, an inference infrastructure startup. “The focus is rapidly shifting from training models to making them operational and accessible in order to power the next generation of AI-native applications.”

The appeal to investors is structural. Unlike training AI models, which lasts a finite amount of time, inference is the meter running: every API call, agent action and enterprise query. And as companies and customers are increasingly demanding more from AI applications, the need for inference has only grown.

“Once a product is ready and customers start using it, when it starts to scale, that’s when inference is needed,” said Assaf Araki, investment director at Intel Capital, an investor in SambaNova, an AI inference hardware development startup. “It’s way more stable from a revenue perspective than development.”

Inference startups are catching tailwinds from other hot companies like Cursor, which develops a code editor, or legaltech startup Legora, gaining traction with enterprise clients. In early January, AI inference infrastructure startup Inferact announced it had raised $150 million in seed funding. Weeks later, in early February, Baseten unveiled its own $300 million round.

And as more companies selling AI services gain a foothold, the need for inference will continue to grow.

“Inference is the core of every interaction with an AI product or model,” said Baseten CEO and co-founder Tuhin Srivastava. “Investing in inference is the same as investing in the growth and adoption of AI as a whole.”

For certain markets, the race to provide inference-focused chips has created a new opportunity. South Korea, for example, in December established the National Growth Fund, a KRW 150 trillion (roughly $100 billion) effort to invest in companies, funds and other channels to help fuel the country’s semiconductor industry. It participated in Rebellions’ new funding round.