AI reasoning track takes off: Inferact raises $150 million in funding, led by a16z

NullWhisperer · 2026-01-22T15:48:14+00:00

The AI inference track has recently secured significant funding, with Inferact completing a $150 million seed round, valuing the company at $800 million. The company focuses on optimizing AI model inference, aiming to reduce costs and improve stability, reflecting a trend in AI competition shifting from training to inference, attracting the attention of several top venture capital firms.

NullWhisperer

2026-01-22 15:48:14

Abstract generation in progress

【Blockchain Rhythm】Recently, the AI inference track has seen significant funding. The core team behind the open-source vLLM project founded the AI company Inferact, which just completed $150 million in seed funding, valuing the company at $800 million. This round of financing was led jointly by a16z and Lightspeed, with top venture capital firms like Sequoia, Altimeter, and Redpoint participating, along with ZhenFund.

Why pay attention to this company? Inferact focuses on AI model inference optimization — a very practical pain point. Currently, the inference costs of large models are exorbitant, and stability is also insufficient. Their approach is to start with bottom-layer optimization to make inference cheaper and more stable. The project originated at Berkeley and is now overseen by the PyTorch Foundation, with a strong technical background.

The signals from this round of funding are clear: the competition in the second half of AI is shifting from training to inference, and investment in infrastructure layers is heating up. For Web3 investors, this is also worth watching — because the logic of efficiency optimization is equally applicable in the blockchain space.

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.

12 Likes

Reward
12
6
Repost
Share

Comment

0/400

GasSavingMaster

· 01-25 15:36

The cost of reasoning is indeed the biggest part; the areas that no one pays attention to are actually the most profitable... With a16z investing so much money, it seems they've also discovered this gold mine.

View OriginalReply0

CryptoPhoenix

· 01-25 07:05

Seeing this news, I knew another bottom range has emerged. The optimization of reasoning is truly severely underestimated; the a16z folks wouldn't throw money around randomly. Those entering now are all preparing for the next rally. People who have gone through cycles understand that opportunities are hidden when others are anxious. This time, the signals are very clear. Hurry up and build your position. $150 million. If this funding had happened two years ago, it might have made the headlines, but now few people are paying attention. This is the real opportunity for rebirth. Don't wait any longer; those who wait patiently will be the last to laugh. I've finally understood that when training is fully saturated, reasoning is the real gold mine. The problem is, how many people are truly serious about doing this? You can still increase your position; don't think about bottom-fishing—just go for it. History will ultimately prove that those who act early earn the most.

View OriginalReply0

liquidation_watcher

· 01-22 16:18

With such high reasoning costs, someone should have taken care of this long ago. It looks like Inferact is serious this time.

View OriginalReply0

HappyToBeDumped

· 01-22 16:15

Inference costs are indeed a big issue. Deploying a model now is extremely expensive... This guy went from vLLM to a valuation of 800 million, and a16z invested 150 million, which is impressive. Artificial Intelligence Funding AI Inferact

View OriginalReply0

DegenRecoveryGroup

· 01-22 16:05

Reasoning is really a gold mine, and finally someone is taking it seriously. The vLLM team started their own venture, and a16z directly invested 150 million, indicating that the big players have seen through it all — the era of burning money to train models is over, and now the bottleneck is the inference cost.

View OriginalReply0

ZenChainWalker

· 01-22 15:56

The cost of reasoning is indeed a big issue, but speaking of which, a16z has been quite aggressive lately, and the valuation is increasing too rapidly.

View OriginalReply0