Author: Techtime

“This May Be the Fastest ASIC Development Cycle Ever Achieved”

Posted on by Techtime

By Yohai Schwiger

OpenAI and Broadcom have unveiled Jalapeño, the first custom AI accelerator the two companies have developed together. The application-specific integrated circuit (ASIC) is designed for AI inference—the stage in which a trained model generates responses to users—and marks the first step in OpenAI’s strategy to build its own AI computing infrastructure and reduce its dependence on NVIDIA accelerators.

According to the companies, Jalapeño was designed specifically for the workloads of large language models and agentic AI applications. Rather than training models, the chip is optimized for inference, where most of OpenAI’s compute demand now resides. The company says early silicon has demonstrated significant improvements in performance per watt, a critical metric given the enormous cost of operating large-scale AI models.

For OpenAI, the project is part of a broader strategy to control every layer of its AI stack—from the models themselves to the hardware that runs them. As OpenAI President Greg Brockman explained, the company has “a deep understanding of our workloads,” allowing it to build hardware tailored specifically to its own AI infrastructure instead of relying solely on general-purpose accelerators.

The new hardware platform extends well beyond the chip itself. OpenAI said the project combines its own accelerator architecture with Broadcom’s expertise in ASIC implementation, networking and connectivity technologies, alongside Celestica’s capabilities in designing boards, racks and complete data center systems. The announcement signals that OpenAI’s objective is not merely to develop a single processor, but to establish a full computing platform for future generations of AI models.

Following tape-out, Jalapeño is now undergoing silicon validation and early testing ahead of deployment in AI servers built by Celestica and eventually integrated into OpenAI’s data center infrastructure. The company describes Jalapeño as “the first step in a multi-generation compute platform,” suggesting that additional generations of custom AI accelerators are already on the roadmap.

A New Era of Accelerated Chip Development?

Yet the most significant aspect of the announcement may not be the chip itself, but how quickly it was developed.

In their official announcement, OpenAI and Broadcom stated that Jalapeño was developed “from initial design to manufacturing tape-out in just nine months,” adding that the project “represents what may be the fastest ASIC development cycle ever achieved in high-performance advanced semiconductors.”

For the semiconductor industry, that is an extraordinary claim. Developing an advanced chip typically takes between 18 and 24 months—and often longer—because of the complexity of architecture design, verification, optimization and manufacturing preparation. Completing the process in just nine months represents a dramatic reduction in development time.

The companies attribute this accelerated schedule to three key factors: close collaboration between the OpenAI and Broadcom engineering teams, Broadcom’s extensive experience in custom silicon development, and “the use of OpenAI models to accelerate parts of the design and optimization process.” While the companies have not disclosed exactly which engineering tasks were assisted by AI, this marks one of the first public acknowledgments by a leading AI developer that its own models were used to accelerate the development of advanced semiconductor hardware.

Industry analysts believe this could be the broader significance of the announcement. Until now, AI has primarily been associated with software development, content creation and data analysis. Jalapeño suggests that AI is beginning to play a role in designing the very chips on which future AI systems will run. As one post-announcement analysis argued, the real question is not whether Jalapeño is a successful chip, but whether nine-month development cycles could become the new standard for advanced semiconductor design.

If that proves to be the case, the implications could be far-reaching. Shorter development cycles would allow chipmakers to iterate more rapidly, update architectures more frequently and align hardware innovation with the relentless pace of AI model development. In other words, AI may soon accelerate not only software creation, but also the development of the semiconductor infrastructure on which it depends—potentially marking one of the most significant shifts in the chip industry in years.

Avnet ASIC to Manage Production of RAAAM Memory Technology at TSMC

Posted on by Techtime

Israeli startup RAAAM Memory Technologies has selected Avnet ASIC Israel to manage the manufacturing of its proprietary Gain-Cell Random Access Memory (GCRAM) technology, which is designed to address one of the biggest bottlenecks in modern processors: embedded SRAM memory.

Under the agreement, Avnet ASIC will serve as the Value Chain Aggregator (VCA) for production on TSMC’s 2nm process technology. The company will be responsible for adapting RAAAM’s design to TSMC’s manufacturing flow, while also providing engineering support, production management and process integration.

Avnet ASIC is an ASIC and SoC design and manufacturing center operating as a business unit of Avnet Silica, which is part of global distributor Avnet. Established about 35 years ago, the company has completed hundreds of semiconductor projects in Israel and abroad. It holds TSMC’s official VCA certification, has extensive experience with 3nm technologies and early access to the foundry’s upcoming 2nm manufacturing platform.

The companies said RAAAM’s technology is currently undergoing qualification and has already been integrated into a customer test chip that completed tape-out in March 2026.

RAAAM’s GCRAM is an embedded memory technology intended to replace the SRAM blocks currently integrated into processors. Because fetching data from external memory is roughly 100 times slower than transferring data within the chip, processors rely on large on-chip SRAM caches to store frequently accessed information. However, SRAM consumes a significant amount of silicon area—sometimes accounting for nearly half of the die.

According to RAAAM, conventional SRAM is becoming increasingly difficult to scale for future semiconductor nodes. The company says GCRAM can reduce silicon area by approximately 50% while lowering memory power consumption by as much as 10x, all while remaining compatible with standard CMOS manufacturing processes.

An Israeli-Swiss collaboration

RAAAM was founded in 2021 by four VLSI researchers from Bar-Ilan University and the École Polytechnique Fédérale de Lausanne (EPFL) in Switzerland. The company has received support from the European Union, the Intel Ignite startup accelerator and private investors, including NXP.

The company’s technology is based on a three-transistor gain-cell memory architecture, compared with the six-transistor cells used in conventional SRAM. The approach occupies a middle ground between SRAM, which requires no refresh, and DRAM, which relies on periodic refresh cycles. GCRAM also performs periodic refresh, but at a much lower frequency than DRAM and transparently to the surrounding circuitry.

RAAAM says its memory operates at supply voltages as low as 450mV in FinFET processes, is manufactured using a standard CMOS flow and features separate read and write ports. For modern processors running edge AI workloads and complex algorithms, this could enable twice the on-chip memory capacity within the same silicon area while significantly reducing power consumption.

The company is competing in the emerging embedded-memory market alongside firms such as Weebit Nano, whose embedded ReRAM technology also aims to provide an alternative to conventional SRAM in future semiconductor designs.

Nanox Issues Going Concern Warning as Shares Plunge Nearly 50%

Posted on by Techtime

Shares of Nano-X Imaging plunged nearly 50% in Nasdaq trading on Thursday, leaving the company with a market capitalization of approximately $50 million. The sharp decline followed the release of the company’s first-quarter 2026 financial results and earnings call, during which management delivered a series of negative updates, including the withdrawal of its full-year revenue guidance and a going concern warning tied to the need for additional financing.

Nanox, which is developing the Nanox.ARC digital X-ray imaging system based on a proprietary X-ray source technology, aims to transform the medical imaging market with lower-cost, cloud-connected imaging systems integrated with artificial intelligence. In addition to its imaging platform, the company operates businesses in teleradiology, AI software and healthcare IT. After securing regulatory approvals and beginning its initial commercial rollout, the company is now focused on expanding customer deployments and converting pilot programs and distribution agreements into recurring revenue.

According to management, however, the transition from early commercialization to meaningful revenue generation is taking longer than expected. As a result, the company withdrew its 2026 revenue guidance. Earlier this year, Nanox had projected approximately $35 million in revenue for 2026, but it now says that target is no longer achievable.

“We no longer expect to achieve the revenue target previously announced for 2026,” CEO Erez Meltzer said during the earnings call. He added that, “we do not currently intend to provide annual revenue guidance going forward.”

According to the company, the timeline between signing commercial agreements, deploying systems, activating customer sites and recognizing revenue has proven longer than expected due to factors including site readiness, infrastructure completion, customer implementation schedules, activation timing and regulatory requirements.

The more significant concern emerged in the financial section of the report. Nanox disclosed that “These factors raise substantial doubt as to the company’s ability to continue as a going concern.” At the same time, management said it is actively seeking additional funding through private equity and the capital markets, warning that any equity financing could dilute existing shareholders. The company also stated that if it fails to secure additional funding, it may be forced to delay, reduce or eliminate parts of its product development and commercialization efforts.

The balance sheet illustrates why investors reacted so strongly. At the end of the first quarter, Nanox held $44.2 million in cash, cash equivalents and deposits, down from $60 million at the end of 2025. The company also disclosed that, on a preliminary unaudited basis, its cash and cash equivalents, net of short-term bank loans, had fallen to approximately $27 million by the date of the earnings release. During the quarter, Nanox generated negative operating cash flow of $14 million.

Operationally, the company reported first-quarter revenue of $4.3 million, up from $2.8 million a year earlier, primarily driven by growth in its teleradiology and Health IT businesses. Net loss for the quarter totaled $14.3 million, compared with $13.2 million in the same period last year.

During the earnings call, management emphasized that commercial deployment activity is improving and highlighted newly signed distribution agreements that could eventually result in hundreds of system sales over the coming years. However, when analysts asked about second-quarter performance and the expected pace of growth, the company declined to provide numerical guidance. Meltzer said only that “Q2 will be better than Q1,” while suggesting that a more meaningful acceleration is expected in the third and fourth quarters, following the onboarding and training of distribution partners. Those assurances failed to ease investor concerns, sending the stock to one of its steepest single-day declines since the company went public.

Dream Raised $260M for “Sovereign AI”

Posted on by Techtime

photo above: Dream co-founders Shalev Hulio (left) and Sebastian Kurz. CREDIT: Dream/Eclipse Media

The Austrian-Israeli AI and cyber defense company Dream, has reached a $3 billion valuation during a $260 million funding round co-led by Bicycle Capital and Group 11. The financing follows nearly $300 million in total contract value secured since Dream began commercial operations in late 2024. The new capital will accelerate deployment of Dream’s sovereign AI and national cyber defense platforms across Europe, the Middle East, Asia, and the Americas.

Dream provides tools to face a growing problem: While artificial intelligence is becoming the next critical infrastructure (following roads, power grids, communications networks, and defense systems) –  most governments do not control the AI systems they increasingly depend on. They rely on models built by foreign companies. Infrastructure they do not own. Technology that can be restricted, interrupted, or withdrawn. At the same time, the data that powers governments remains fragmented across ministries, agencies, and critical infrastructure.

The next super nations

The company helps nations secure their most sensitive information, transform fragmented data into actionable knowledge, and deploy advanced AI entirely under sovereign control. Founded in 2023 by Shalev Hulio, former Austrian Chancellor Sebastian Kurz, and Gil Dolev, Dream serves governments and critical infrastructure organizations across Europe, the Middle East, and Southeast Asia. The company employs approximately 350 people across Tel Aviv, Abu Dhabi, and Vienna. With this financing, Dream has raised $412 million.

“Land created empires. Industry created nations. Artificial intelligence will create the next super nations,” said Shalev Hulio, Co-Founder and CEO of Dream. “Every nation has data., but few can protect it and Fewer can use it. Sovereign AI is the key. The future of a nation should never depend on technology it does not control.”

Sebastian Kurz: “Nations that want to control their future need the ability to operate advanced AI under their own authority, on infrastructure they govern, and in alignment with their own interests. Sovereign AI is becoming a foundational layer of national resilience, competitiveness, and security.”

To face this callenge the company developed three platforms: Sphere helps governments and critical infrastructure operators defend against nation-state cyber threats. Hero is an autonomous AI security researcher that discovers vulnerabilities, identifies attack paths, and reasons like an adversary at machine speed to prevent the most sophisticated cyber threats, and Atlas is Dream’s sovereign AI platform. It enables governments to connect fragmented national data, transform information into structured knowledge, deploy mission-specific AI agents and models, and generate actionable insights entirely within secure government-controlled environments.

Mobileye to Launch Robotaxi Service, Taking on Waymo and Tesla

Posted on by Techtime

Above: Illustration of Mobileye’s planned robotaxi service. Credit: Mobileye

Mobileye unveiled a major strategic initiative that could significantly expand its role in the autonomous mobility industry. The company announced plans to launch and operate its own robotaxi service in the United States, taking responsibility for fleet operations, user experience, service management, and day-to-day transportation activities.

Under the plan, Mobileye intends to deploy an initial fleet of approximately 100 autonomous vehicles in a major U.S. metropolitan area in 2027. Following the initial rollout, the company aims to scale the service aggressively, targeting roughly 17,000 vehicles within five years.

The service will be built around Mobileye Drive, the company’s autonomous driving platform, and Moovit, the mobility platform acquired by Mobileye in 2020. Moovit is expected to provide the consumer-facing layer, including ride booking, route planning, fleet management, multimodal transportation integration, and operational interfaces.

“The robotaxi revolution is only beginning,” said Prof. Amnon Shashua, Mobileye’s founder and CEO. “Today we are taking the next step by combining our technology with direct operational capabilities in order to build a robotaxi business designed for financial and geographic scale.”

From Technology Provider to Service Operator

The announcement marks a significant evolution in Mobileye’s business model. For more than two decades, the company has established itself as one of the world’s leading suppliers of ADAS and autonomous driving technologies. Its technology is currently deployed in more than 230 million vehicles worldwide, but its business has largely focused on supplying chips, software, and autonomous driving systems to automakers and mobility operators.

The new initiative allows Mobileye to participate not only in the technology layer, but also in the revenue generated by transportation services themselves. Instead of selling only the “brain” of an autonomous vehicle, the company now seeks to capture part of the value created by every ride.

Beyond revenue opportunities, operating its own robotaxi network will provide Mobileye with hands-on experience in fleet management, maintenance, charging infrastructure, customer support, pricing, and remote operations—areas that have become key competitive differentiators for leading robotaxi operators.

The company emphasized that the move does not replace its existing partnerships with automakers and mobility providers. Rather, it represents an additional route to market, with customer-led robotaxi programs and Mobileye-operated services expected to progress in parallel.

Mobileye Drive at the Core

At the heart of the initiative is Mobileye Drive, the company’s full autonomous driving system designed for robotaxis, autonomous shuttles, and driverless transportation services.

The platform combines multiple sensing and software technologies developed by Mobileye over the years, including cameras, radar, LiDAR, REM mapping, and EyeQ processors. Unlike Tesla’s vision-centric approach, which relies primarily on cameras and AI, Mobileye has adopted a multi-sensor architecture in which independent sensing systems operate in parallel and continuously validate one another.

A key component of the platform is the LiDAR technology supplied by Israeli company Innoviz Technologies. Mobileye selected Innoviz’s LiDAR sensors for the Mobileye Drive platform in 2021, and the technology is now integrated into several major robotaxi projects based on the system.

For Innoviz, Mobileye’s expansion could represent a significant commercial opportunity. If Mobileye succeeds in scaling its fleet to thousands of vehicles, demand for the LiDAR systems embedded in Mobileye Drive could increase accordingly.

The ID. Buzz Robotaxi Program

One of the most advanced deployments of Mobileye Drive today is the collaboration between Mobileye, Volkswagen, and mobility provider MOIA.

The project centers on the autonomous ID. Buzz AD, a driverless version of Volkswagen’s electric van designed specifically for robotaxi services. The vehicle integrates Mobileye Drive with a sensor suite that includes cameras, radar, and Innoviz LiDAR units.

More than 100 autonomous ID. Buzz vehicles are already participating in pilot programs across the United States and Europe, with Orlando selected as the first U.S. city slated for commercial deployment. The program is widely viewed as one of the most advanced robotaxi initiatives built on Mobileye technology and offers a glimpse into what Mobileye’s own future service could look like.

If the competition among Mobileye, Tesla, and Waymo has until now focused primarily on whose autonomous driving technology is superior, today’s announcement signals a new phase in the industry’s evolution. The challenge is no longer only teaching a vehicle to drive itself—it is building a scalable transportation business around that capability. For Mobileye, the move may represent the most significant opportunity in the company’s history.

After Years of Legal Disputes, IQE and Tower Join Forces in AI Photonics

Posted on by Techtime

UK-based materials company IQE and Israeli chipmaker Tower Semiconductor have announced a long-term supply agreement covering indium phosphide (InP) epitaxial wafers, a key technology used to manufacture photonic components for optical networking, data centers, and artificial intelligence applications.

Under the agreement, IQE will supply Tower with InP-based epitaxial wafers used in the production of lasers, photodetectors, and other advanced optical devices. Tower plans to integrate these materials into its photonics manufacturing processes, which serve customers in optical communications and AI-related markets.

The announcement is particularly noteworthy given the companies’ contentious legal history.

In 2022, IQE filed a federal lawsuit in the United States accusing Tower of misappropriating trade secrets and proprietary know-how related to porous silicon technology, which enables advanced integration of electronic and photonic components on silicon substrates.

According to the complaint, IQE alleged that it had shared sensitive technical information with Tower during discussions about a potential collaboration under confidentiality agreements. The company claimed that Tower later filed patent applications that incorporated knowledge disclosed during those discussions. IQE sought damages and asked the court to review the inventorship of certain patents involved in the dispute.

Tower consistently denied the allegations, arguing that its patents had been developed independently and that the claims lacked merit. Legal proceedings continued for several years, but no final court ruling was issued publicly determining the validity of either side’s claims.

Against that backdrop, the new partnership suggests that both companies have moved beyond the dispute and are now focusing on the commercial opportunities emerging in the rapidly growing photonics market.

Indium phosphide has become one of the semiconductor industry’s most strategically important materials in recent years. It is widely used in high-performance lasers and optical interconnects that enable high-speed data transmission between AI processors inside modern data centers. As AI workloads continue to expand, demand for optical networking infrastructure has surged, driving increased investment across the photonics supply chain.

For IQE, the agreement secures a strategic long-term customer and a stable source of revenue in a fast-growing market. For Tower, it provides reliable access to advanced photonic materials needed to expand its position in optical communications and AI-related applications.

While neither company has presented the agreement as a formal legal settlement, the joint announcement marks a clear shift in their relationship. After four years of public litigation, IQE and Tower are now collaborating in one of the semiconductor industry’s fastest-growing sectors, turning a dispute over intellectual property into a partnership centered on the AI-driven photonics boom.

Intel Israel and Microsoft Remove a Key Limitation in Secured Cloud

Posted on by Techtime

[Photo: The Intel Israel engineering team behind the development of the new capability. Photo credit: Aviv Harel]

By Yohai Schwiger

At its Build conference in Seattle, Microsoft unveiled a new capability called Confidential Live Migration, which allows encrypted virtual machines (VMs) to be moved between servers in Azure data centers without interruption. The technology was developed by a team of approximately 20 engineers from Intel Israel’s security group and is built on Intel’s TDX platform, which is integrated into the company’s Xeon server processors.

The new capability addresses one of the most significant limitations in the field of confidential computing—a rapidly growing area designed to enable organizations to run sensitive workloads in the cloud while protecting data even from the cloud provider itself. While conventional virtual machines have long supported live migration for maintenance, load balancing, and infrastructure upgrades, encrypted VMs have traditionally required downtime whenever they needed to be moved to another server.

The challenge stems from the security mechanisms that make confidential computing possible. Technologies such as Intel TDX encrypt and isolate a virtual machine’s memory at the hardware level, preventing access not only by the cloud provider’s administrators but also by the host operating system and cloud management software. This additional layer of protection is particularly valuable for financial institutions, healthcare organizations, government agencies, and other enterprises handling highly sensitive information.

“This feature is entirely an Israeli development,” said Boaz Tamir, Senior Director at Intel. “We can move an encrypted virtual machine from one server to another while it continues running, without the customer noticing anything. For customers, this means higher service availability. For cloud providers, it means greater flexibility in managing infrastructure without compromising the security model.”

According to Tamir, the architecture, development, and validation of the new capability were all carried out by Intel’s security team in Haifa. The engineering challenge involved securely transferring the VM’s execution state, encrypted memory, encryption keys, and trust mechanisms between two separate servers, while cryptographically verifying the target environment and maintaining uninterrupted service.

Bringing Secure Hardware Trust to the Cloud

At the heart of the development is Intel TDX (Trust Domain Extensions), one of Intel’s flagship security technologies for its Xeon processor family. TDX enables the creation of isolated and encrypted computing environments known as Trust Domains, where sensitive workloads can run without exposure to the host operating system, cloud management software, or even data center operators.

In essence, Intel is extending the hardware-based trust model traditionally associated with secure enclaves into the world of cloud infrastructure and hyperscale data centers.

TDX is part of the broader confidential computing movement, which has emerged as one of the fastest-growing segments of cloud infrastructure. While traditional security approaches focus on encrypting data at rest or in transit, confidential computing aims to protect data while it is actively being processed. This means that information remains protected even when loaded into server memory and being analyzed by applications or AI models.

Demand for such capabilities continues to rise as organizations move increasingly sensitive workloads to public cloud environments. Banks, insurance companies, healthcare providers, government agencies, and technology firms are seeking to leverage cloud scalability while reducing the need to place complete trust in infrastructure providers. At the same time, the rapid adoption of AI is driving new requirements for secure processing of medical, financial, industrial, and defense-related data.

Eliminating a Major Operational Barrier

For Microsoft, the new capability removes one of the primary operational constraints that have historically affected confidential computing services. Routine maintenance tasks—including server upgrades, hardware replacements, and load balancing—have often required temporary shutdowns of encrypted virtual machines.

With Confidential Live Migration, those operations can now be performed transparently, much like they are for standard virtual machines, while preserving the security guarantees of confidential computing.

The project also highlights Intel Israel’s role in developing core technologies for the global TDX platform. Unlike customer-specific software projects, TDX is a foundational element of Intel’s server security architecture. As a result, the expertise and technologies developed by the Israeli team become part of a broader platform used by cloud providers and secure applications worldwide.

As Microsoft rolls out the new capability across Azure’s DCesv6, ECesv6, DCedsv6, and ECedsv6 server families, Intel is already working on the next generation of confidential computing technologies. The company says its Israeli TDX development team is also building a new capability called TDX Connect, designed to further expand the possibilities of secure computing within data centers.

Over the past decade, competition among cloud providers has largely focused on performance, cost efficiency, and AI capabilities. In the years ahead, however, one of the industry’s most important battlegrounds may be the ability to guarantee that even the cloud provider itself cannot access customer data.

The joint Microsoft–Intel development offers a glimpse of that future: a cloud environment where security, privacy, and high availability coexist without compromise.