هذا التقرير متاح أيضًا بـ العربية
US technology companies are racing to acquire Israeli chipmaking expertise as competition intensifies to cut the costs of artificial intelligence and strengthen US dominance in the face of China.
The latest development came on 26 October 2026, when US company “Unconventional AI” announced an agreement to acquire Israel’s Neureality for between $150 million and $200 million. Unconventional AI is valued at $4.5 billion.
The deal comes amid an expanding technology partnership between Washington and Tel Aviv, and a US push to build alliances that secure its edge in chips and data centres. Acquisitions of Israeli companies form part of an economic and strategic competition that extends beyond the US market.
Who is behind the Neureality deal?
Neureality was founded in 2019 by three Israeli engineers, Moshe Tanach, Zvika Shmueli and Yossi Kasus, who established its headquarters in Caesarea, between Haifa and Tel Aviv, after years working at chip companies including Intel, Marvell, Mellanox and Habana Labs.
The founders focused on developing technologies to help data centres run AI applications at lower cost, in a market preparing for a wave of massive investment.
A year later, Naveen Rao, a US engineer and entrepreneur specialising in AI, became one of the company’s early investors when it was valued at about $12 million, according to Israeli business newspaper Globes.
In February 2021, Neureality announced that it had raised $8 million and that Rao, Intel’s former head of AI products, had joined its board.
Rao had founded Nervana Systems, which specialised in AI processors. Intel acquired it in 2016 for more than $400 million.
He then co-founded MosaicML to develop AI models. US data analytics company Databricks acquired it in 2023 for about $1.3 billion. In 2025, he founded Unconventional AI with the aim of developing processors that consume less power.
His new venture attracted exceptional funding within months of its founding. In December 2025, the company announced that it had raised $475 million in a round led by US technology investment funds Andreessen Horowitz and Lightspeed. Sequoia Capital, Amazon founder Jeff Bezos and other investors also participated, bringing its valuation to $4.5 billion. Rao invested $10 million of his own money.
Neureality’s backers, meanwhile, include global companies such as South Korea’s Samsung and SK Hynix, alongside US and Israeli funds, most notably Varana Capital, its largest shareholder.
The fund is led by US investor Ezra Gardner, a Neureality board member. Documents filed with the US Securities and Exchange Commission in 2025 show that Varana focuses its investments on companies linked to Israel and participates in their management and advises on mergers and acquisitions.
Neureality has raised about $105 million since its founding, according to the Israeli business press. “Globes” estimated the current acquisition price at between $150 million and $200 million, compared with a valuation of $220 million in 2023, citing US financial data platform PitchBook. The company may therefore be sold for less than one of its previous valuations, despite the years and money invested in developing its products.
The agreement includes a mix of cash and shares in the US company, giving shareholders a chance to benefit from any future rise in its value. Their returns will vary according to their stakes and the terms of the sale, particularly as the deal allows them to move into a US venture with massive funding and ambitions to enter a new processor market.
The most striking irony, however, is that despite the money it has raised, the US company needed the engineering expertise Neureality had accumulated over seven years. Rao explained on 26 October that his venture had expected to spend about three years on research and development and five years to reach a commercial product. But it progressed faster than expected using AI tools and reached the point where it needed engineers capable of turning designs into chips that could be manufactured and marketed.
In the acquisition announcement, Rao said that assembling a team capable of producing chips and building the systems around them was one of the industry’s hardest tasks. He argued that bringing in the Neureality team would significantly accelerate his company’s plans.
About 40 employees from the Israeli company are expected to join the US group, including the founders and members of its development teams, while an engineering presence will remain in Israel. The deal is expected to close in the fourth quarter of 2026, once the closing conditions have been met.
Why are US companies racing for chips?
The deal’s significance begins with the cost of running AI. Every prompt a user sends to a model such as “ChatGPT” requires computations inside data centres packed with advanced processors. As users and applications multiply, so do the costs of buying those processors and providing the electricity, cooling and networks needed to run them.
Figures from the International Energy Agency show the scale of the problem. On 16 April 2026, it said data centres had consumed about 485 terawatt-hours of electricity globally in 2025, with consumption expected to rise to 950 terawatt-hours by 2030. Capital spending by five major technology companies also exceeded $400 billion in 2025 and was expected to increase by roughly 75 per cent in 2026.
Faced with these costs, companies are looking for ways to get as much work as possible from the processors they buy. Some capacity remains unused while processors wait for data or tasks are distributed across different systems, even as electricity and cooling continue to be consumed. This is where Neureality’s technologies stand out: it developed chips and software to organise data movement and help processors handle more requests with the same resources.
In June 2024, the company announced test results showing that its systems, integrated with processors from US company Qualcomm, could reduce the cost of running some AI applications by an average across tests of up to 90 per cent compared with systems using processors from Nvidia, one of the world’s largest chipmakers.
The tests used different configurations under Neureality’s supervision. The results therefore depend on the applications and system components used, and broader independent testing is needed to establish the scale of savings in commercial use.
As it developed its products, Neureality worked with global companies including US computing firm IBM, UK chip designer Arm and US chipmaker AMD. Taiwan Semiconductor Manufacturing Co., or TSMC, produced its first processor, allowing the Israeli company to move from design to testing systems that could be used inside data centres.
Investment in the technology’s development also came from Brussels. Neureality received a €2.5 million grant from the European Innovation Council, part of the European Union, under a project that ran from July 2023 to June 2025 to complete development of its processor and test it before commercial launch. The council’s fund later participated in a $20 million investment round in the company in 2024.
Project documents published by the European Commission show that Neureality owns 24 patent families. When the project ended, patents had been granted in six families, while applications in the other 18 remained pending. These relate to improving task distribution and data transfer between processors. The documents also recorded tests showing performance gains of more than 60 per cent for one large language model in a specific comparison.
Rao’s statements illustrate the scale of the economic bet on these technologies. On 24 October 2026, he warned that the rapid expansion of data centres could exhaust the surplus electricity capacity available in the US within a few years. He estimated that the market for running AI models after training could reach $10 trillion over the next decade.
His company is seeking to develop processors aimed at improving operating efficiency by a factor of 1,000 compared with conventional systems. This research goal faces challenges in manufacturing and commercial production.
In December 2025, Andreessen Horowitz, one of the project’s major backers, highlighted both the significant potential of these technologies to reduce energy consumption and the difficulties of scaling up their use. This gives Neureality’s chip production expertise particular value for the US venture.
What do the US and Israel gain?
The Neureality deal follows multibillion-dollar US acquisitions of Israeli companies specialising in chipmaking. The most prominent include Nvidia’s purchase of Mellanox in 2020 for about $7 billion to obtain the high-speed networking technologies that connect thousands of processors inside data centres, and Intel’s acquisition of Israel’s Habana Labs, which specialises in AI processors, for about $2 billion in 2019.
In 2024, Nvidia acquired Israeli company Deci, which develops technologies to improve the performance of AI models, in a deal that Israeli business newspaper “Calcalist” estimated at about $300 million. It then acquired Run:ai, which specialises in managing processor resources, for about $700 million.
In September 2026, US semiconductor company Microchip Technology completed its purchase of Israel’s Hailo, which develops AI processors used in devices and industrial systems.
These deals target complementary areas of the AI industry, from processors themselves to the networks and software that determine how efficiently they operate. US groups often retain Israeli development centres and engineering teams.
A joint report published by the Israel Innovation Authority and European platform Dealroom in September 2025 showed that there were about 1,500 deep-tech companies in Israel, which raised $28.6 billion in investment between 2019 and 2025.
These economic interests intersect with US policy towards the AI industry. On 23 July 2025, President Donald Trump issued an executive order to expand exports of the US technology stack to allied countries, including chips, servers, software and data centres. The aim was to strengthen US dominance and reduce global markets’ reliance on rival states’ technologies.
China is central to this competition. Washington uses export controls on advanced processors to limit Chinese companies’ access to some computing capabilities while seeking to expand the use of US technology among allied countries.
Control over chips and the systems around them gives Washington economic and security advantages, particularly as Chinese companies and other countries rely on technologies and components subject to US licensing.
In a study published by Israel’s Institute for National Security Studies on 29 July 2026, researcher Hadas Lorber argued that AI superiority had become tied to control over its production chain, from chip design and manufacturing to energy and data centres. Lorber called for strengthening Israel’s position in the US-led technology system and using the partnership to develop industrial capabilities that serve the economy and meet security and military needs.
Cooperation between Washington and the occupation government took formal shape in the months before the deal. In December 2025, Israel joined the US Pax Silica initiative, which seeks to secure supply chains for chips, minerals and energy among partner countries.
On 16 January 2026, the two governments signed a joint declaration in occupied Jerusalem launching a strategic partnership in AI, semiconductors, and research and development. Israeli Foreign Minister Gideon Sa’ar described Israel as an important asset to US national security interests.
The initiative extended to the Gulf when Qatar and the UAE joined in January 2026. Its participants also include Israel and other industrial countries such as Japan, South Korea and Britain.
This comes alongside Israeli proposals to divide roles in the AI industry between US technology, Israeli expertise, and Gulf capital and energy, allowing an expanded Israeli technological presence in regional projects despite differences in the bilateral relations between participating countries.
Washington’s ability to influence these projects is evident in its chip export licences. On 19 November 2025, the US Commerce Department announced approval for exports equivalent to a combined total of 35,000 advanced processors to G42, the UAE-based AI and cloud computing company, and Humain, the Saudi company owned by the Public Investment Fund. The licences were subject to security and oversight conditions.
This policy gives the US leverage over Gulf AI projects, which need US processors to run new data centres, amid its competition with China over markets and technology.
In Gaza, the most dangerous military uses of Israeli AI are on display, after the occupation army used systems to analyse Palestinians’ data, classify them and compile target lists during its assault on the Strip.
In a report published in September 2024, international rights group Human Rights Watch documented the use of digital tools including “Lavender” to classify people as potential targets, and “Habsora” to compile lists of buildings to be targeted. It warned of the dangers of relying on data and estimates that could lead to civilians being targeted and to violations of international humanitarian law.
The use of these systems reveals technology’s growing role in the occupation army’s surveillance and targeting operations, and the responsibilities this imposes on companies and investors tied to the sector.
This environment raises questions about responsibility for technology’s end use, particularly as investment expands in a sector where civilian applications overlap with the development of the occupation’s military and intelligence capabilities.
Ultimately, the Neureality deal reveals a trend in which US companies’ interests intersect with Israel’s economic and technological goals. The companies gain engineering expertise developed over years to accelerate chip production, while Israel preserves its place in global development chains. This comes as Washington works to expand its influence by controlling the processors, energy and computing infrastructure that have become the foundation of competition over AI.