rewrite this title and make it good for SEOExclusive: Element Labs raises funds at $4b valuation

Exclusive: Early-stage Israeli chip startup Element Labs, which operates in complete secrecy under the founding team of Habana Labs, which was previously sold to Intel, continues to grow in stealth. “Globes” has learned that existing investors, including insurance giant Fidelity, have invested another $300-400 million in the company at a valuation exceeding $4 billion.

The company’s founder serial entrepreneur Avigdor Willenz became famous for a series of chip companies that he founded and sold for billions of dollars over almost three decades. But market executives describe Element Labs, which is based in Caesarea, as “his greatest promise, and the project that is dearest to his heart at this time.”

Earlier this month, another of his portfolio companies, Xsight Labs of Kiryat Gat, raised $300 million at a valuation of $3 billion, led by his longtime partners – Gavin Baker’s Atreides Fund and Fidelity. Xsight develops communications processors for data centers for processing AI.

But unlike Xsight Labs and the other companies founded by Willenz, Element Labs operates in complete secrecy. The company has no website, it does not have a page on the social networking site LinkedIn, and its executives do not give media interviews. Hiring of employees is conducted mainly through friends referring friends and not through conventional channels such as advertising on job sites and placement agencies. The assumption is that the company not only operates in a very competitive field but already has business relations with several US cloud giants that require confidentiality. A processor for fast, critical calculations

Element Labs is tackling the most difficult but rewarding challenge in the AI market these days: the need to provide cheap and efficient processors for running AI agents and running processes for models that have already been trained by Nvidia processors. This AI processing is gradually becoming significant in the activities of AI processing centers and imposes heavy costs on the cloud and AI giants.

Nvidia’s graphics processor, which is considered very effective in training models, is today seen as expensive and has insufficient performance for the challenges of AI agents and running language models that already reach trillions of parameters – the basic processing units of AI. Chip companies like Cerebras Systems, SambaNova, which is currently supported by Intel, and Grok, which was acquired by Nvidia itself, are trying to offer a new type of processor, but one that is mainly suitable for fast calculations for critical and very specific uses and is priced accordingly.

Element Labs is trying to lower the costs of AI processing by offering a new structure of servers, and fundamentally different AI processing and communication chips, which it is targeting the largest market in AI: the cloud giants – such as Amazon, Microsoft, Meta, language model makers, such as Anthropic and OpenAI, and neocloud companies (companies that rent data center space for AI processing) including Crusoe, Nebius and CoreWeave, which are looking for an alternative to Nvidia’s expensive processor in the early stages. In these stages, trained AI models are typically used to provide answers to smart chat engines, recognize images, and even operate AI agents – a kind of intelligent software capable of performing human tasks – such as writing software code, compiling research, and producing a video.

RELATED ARTICLES

Former partner appointed Intel CEO

The computing phase requires cloud giants to adopt a more efficient operational approach, in which the emphasis is on the number of tokens calculated per kilowatt of electrical power, and not on memory bandwidth or computational power, as in the operation of training a language model. Element Labs has tried to solve this in several ways: developing a special chip for computing workloads – considered cheaper and more efficient than the expensive and cumbersome GPU processor; communication infrastructures designed for dense AI clusters and multiple server racks; and software that manages the communication network and AI processing and is reminiscent of Mellanox’s activities in Israel.

Element Labs is not the only company in this field and competes with Broadcom and Marvell in the market for AI chips tailored to the specific requirements of cloud giants. Prior to the current financing round, the company had raised, according to PitchBook, only about $130 million, with the most recent valuation given to it in 2025 being $1.1 billion. Estimates are that the company currently has 350 employees in Caesarea and Tel Aviv, and several hundred more outsourced contract workers.

Element Labs was founded by a team that is highly sought after in the investment world – the entrepreneur and serial investor, chip billionaire Avigdor Willenz, and the founders of Habana Labs: David Dahan and Ran Halutz. Willenz was also a partner in the investment in Habana Labs, and the new startup brought in one of his former colleagues, Manuel Alba-Marquez, his former partner at Galileo, as an early investor. Willenz’s senior position and the founders of Habana Labs open doors for Element Labs at any large chip factory and any electronics company. Willenz’s former partner in Israel investments, Lip-Bu Tan, was only appointed CEO of Intel in March 2025.

Wilenz is an entrepreneur and serial investor in the chip sector who has sold Habana Labs to Intel for $2 billion; Annapurna Labs to Amazon for $380 million; and Galileo to Marvell Technologies for $2 billion in 2000. Willenz told Globes in an interview two years ago that he had moved to Switzerland and significantly reduced his activities in Israel, due to the government’s conduct and the judicial reform. Now, it seems that Willenz has returned to being more significantly involved in an Israeli company.

No response to this report has been forthcoming from Element Labs.

Published by Globes, Israel business news – en.globes.co.il – on June 29, 2026.

© Copyright of Globes Publisher Itonut (1983) Ltd., 2026.