Pomegra
Herzliya's Axionic Labs, founded by ex-Qualcomm VP Assaf Touboul, has $50M of an up-to-$80M pre-seed to build an AI inference chip aimed at Nvidia and Cerebras.
Key Takeaways
- Axionic Labs has raised $50M and is opening its pre-seed to as much as $80M.
- Amplify Partners, Wing VC and Qualcomm Ventures lead; Pitango, Viola and 10D also invest.
- Valuation is undisclosed, and the company has about 20 employees.
Lead
Axionic Labs, a chip startup based in Herzliya Pituah, Israel, is raising up to $80M in a pre-seed round and has already secured $50M, according to reporting published on October 14, 2026. Former Qualcomm vice president of technology Assaf Touboul founded the company in June. It is building a chip for the inference stage of AI, the phase in which trained models and agents answer queries. The company says the chip will cost less and use less energy than Nvidia GPUs.
Who Is Backing Axionic Labs?
The round is led by Amplify Partners, Wing VC and Qualcomm Ventures, with the Israeli funds Pitango, Viola and 10D participating. Touboul opened the round to more investors after demand exceeded his original plan, which lifted the ceiling to $80M. The $30M gap between what is in and what is targeted is still open.
Valuation has not been disclosed. For a company with roughly 20 staff, 10 of them PhDs, a $50M first check is large by pre-seed norms. It prices the team and Touboul's résumé more than any product, since no silicon has been publicly shown.
Who Is Assaf Touboul?
Touboul is one of the most senior Israelis to have worked at Qualcomm, where he held the vice president of technology role. He oversaw cellular communication chip development and ran the company's innovation and AI labs. His work there included ultra-low-power analog AI chips, language model compression and training algorithms, and he holds dozens of patents.
That background matters because the inference market rewards power efficiency. Analog and low-power designs are one route to it, though Axionic has not said which architecture its chip will use.
What Is Axionic Building, and Whom Does It Compete With?
Axionic is building an inference chip for AI models and agents that it says will undercut Nvidia GPUs on cost and energy use. It names Cerebras as a direct rival, and the field also includes Groq and SambaNova, which have pursued their own inference-focused designs.
The company has a US parent, R&D in Herzliya and a UK-registered entity. It has disclosed no customers, no tape-out date and no performance figures. Every efficiency claim is therefore a target, not a measurement.
Why Are Investors Funding Inference Chips Now?
Inference demand is growing as AI agents run continuously, and operators are looking for cheaper alternatives to Nvidia hardware. Training happens once per model, while inference repeats with every user request. That makes per-query cost and power draw the metrics buyers watch most closely.
The round also reflects a split in investor appetite. Chip suppliers keep earning revenue from AI spending, while many AI software companies still face questions about durable margins. Backers are placing money on the hardware layer as a result.
Skeptics have grounds. Nvidia's software ecosystem and manufacturing relationships are hard to match, and earlier challengers have needed several funding rounds and hundreds of millions of dollars to reach volume production. A pre-seed of this size buys a design team and perhaps a first test chip, not a product line.
What Comes Next for Axionic?
The nearest milestone is closing the remaining $30M of the round, which would complete one of the larger pre-seeds out of Israel's chip sector. After that, the company has to show working silicon and benchmark results against GPUs and Cerebras systems on standard inference workloads.
The presence of Qualcomm Ventures gives Axionic a possible route to partners and manufacturing contacts, though no commercial arrangement has been announced. A follow-on round will probably hinge on test-chip data, since a founder's pedigree can carry a first raise but not a second.
Outlook
Axionic Labs has $50M committed, as much as $80M in sight and a founder with senior Qualcomm experience. It still has to turn a cost and efficiency claim into measured results. Valuation, architecture and customers remain undisclosed, and those details will determine whether this round reads as early conviction or as an expensive bet on a résumé.
