Intel hosted technology deep dive sessions in Israel, sharing details on its investment priorities, fabrication practices, and ongoing culture changes. We left the event persuaded that its expanded software portfolio is its smartest bet.
During the week of September 12, Omdia analysts attended Intel’s technology deep dive sessions in Israel. Intel provided in-depth insight into the design of its new Raptor Lake CPU and shared details on its investment priorities, software portfolio, fabrication practices, and ongoing culture changes. By the end of the event, we were persuaded that Intel’s expanded software portfolio is the company’s smartest bet.
Intel Israel is a change agent for the company
Intel began investing in Israel three decades ago and discovered that a large pool of engineering graduates, coupled with an open and collaborative work culture and strong venture capital (VC) presence, makes for an ideal home away from home. It currently employs 14,000 people in Israel across four sites, including “Fab 28,” where its current generation of desktop and notebook processors are being manufactured on its 10nm fabrication process.
Significantly, when Pat Gelsinger, Intel’s recently appointed CEO, looked for a dependable leadership team in his efforts to turn Intel’s fortunes around, he filled two key positions with local talent:
- Shlomit Weiss: Senior vice president and general manager of the Design Engineering Group
- Daniel Benatar: Vice president and general manager of Worldwide Semiconductor Manufacturing
What was clear from Omdia’s interaction with Weiss and Benatar is that they share a very strong focus on delivery and people. Their track record in enabling some of the company’s biggest successes suggests they are well-positioned to succeed in their new appointments. Focus areas that both highlighted were roadmap predictability, cadence, and manufacturing quality.
And is powering Intel’s budding software business
In May, Omdia said that Intel is more of a software company than you think. If anything, the deep dive sessions in Israel solidified our belief that Intel’s next billion-dollar business is likely to be a software one. This is not accidental. Since 2019, the architect of Microsoft’s startup incubation program, Tzahi “Zack” Weisfeld, has been setting up Intel’s own startup program, dubbed Intel Ignite. Ignite provides mentorship, networking, and information to early-stage startups.
One of the first businesses to come out of Intel Ignite is Granulate, the workload performance optimization software it acquired earlier this year. Omdia briefly summarized this acquisition in our May analyst opinion, “Intel is more of a software and services company than you think.”
Intel might have found a unicorn in Granulate
Granulate’s software agent sits between the OS/runtime and the application, passively learning data flows, processing patterns, and all-around resource consumption. The learning process takes one to two weeks and results in a visual map of the software environment, including CPU and memory use. Using the information, Granulate can optimize the runtime and OS for throughput, CPU utilization, or response time. In a live demonstration, the company showed a marked reduction in latency for one of its clients. Using Granulate, some companies can optimize for low enough latency using existing resources (either in the cloud or on-premises) and avoid deploying more compute at the edge. Clearly, Granulate can also save money by reducing the number of physical or virtual servers a company needs.
What is particularly exciting is the business model. Granulate gets paid a share of the savings a company achieves, enabling a win-win scenario. This explains why this very young company (four years ago, it had five employees) has such an impressive partner and adopter ecosystem. Granulate has signed co-selling agreements with the three largest infrastructure as a service providers, Amazon, Microsoft, and Google. Its client count is now in the dozens and includes Walmart and Snapchat. By Omdia’s internal estimates, Granulate has already surpassed $10m in annualized revenue and is getting a growth boost since being acquired by Intel, driven by new client acquisition. Omdia would classify most of Granulate’s existing clients as SMBs and sees the Intel sales team opening the doors to the larger enterprises with which it has long-standing direct relationships.
cnvrg.io (Converge IO) is set to expand beyond MLOps
cnvrg.io, the Israeli-based software maker that Intel acquired in 2020, was identified as a very strong competitor in the MLOps market in our recent report, Omdia Universe: Selecting an Enterprise MLOps Platform. The Omdia team noted deep hardware integration, straightforward and advantageous pricing, solid data and model versioning, and an emerging community ecosystem as strong points for cnvrg.io.
During our meeting with them in Petah Tikva, the company demonstrated a practical understanding of the day-to-day challenges that data scientists face. cnvrg.io estimates that a majority portion of data scientists’ time is spent configuring hardware and software platforms, from selection and experimentation to resource orchestration, scheduling, and management. This leaves little time for actual data science modeling and insights production. This issue inspired the cnvrg.io founders, working data scientists at the time, to construct a solution.
Today, cnvrg.io has 60 employees and is nearing 100 enterprise clients—an impressive result for a young company. Importantly, its library of predefined machine learning (ML) models and application optimization tools has grown. The Intel acquisition looks to have also inspired experimentation, and the MLOps vendor shared with Omdia that it intends to launch an as-a-service offering and expand into further fields like database and analytics. This would significantly increase cnvrg.io’s opportunities.
Investments that could turn into full-blown acquisitions
Beyond the two acquisitions described above, Intel is actively investing time and efforts in helping quite a few software startups in Israel grow:
- ClassiQ automates the design of quantum circuits by enabling the abstraction of hardware and software. It delivers an optimized quantum program based on parameters defined by the designer. It has already received $63m in funding.
- Cady, the earliest-stage startup Omdia met in Tel Aviv, automates the detection of errors in electrical schematics design. It has raised $4m in funding so far and is about to launch its first commercial product at the end of the year. As would be expected, Intel’s processor designers are already using Cady’s software.
- Orca AI is a ship navigation software delivered as software as a service (SaaS). Already installed on hundreds of ships, it enables navigation and fleet management. In February 2022, a ship completed a fully autonomous journey using the Orca AI software.
- Hi Auto is a virtual assistant for drive-through restaurants that has already been deployed in 250 stores (in about a year). Also delivered as SaaS, it can help with turnover, which stands at 120% for takeout restaurants, according to Hi Auto. Omdia’s analysts tested it in multiple ways and accents to surprisingly positive results.
While not strictly software, two robotics companies, Indoor Robotics and Verobotics, both part of the Intel Ignite program, demonstrated fascinating solutions during live demos in Tel Aviv. The former enables unmanned security and monitoring through indoor drones. Indoor Robotics demonstrated the application of the technology in data centers, where it can detect intruders and hotspots. Verobotics solves an increasingly prevalent real-world problem—skyscraper window cleaning and inspection. Its autonomous cleaning robot can match or beat the dangerous human-managed suspended window cleaning process. Importantly, Verobotics’ solution is priced at one-tenth of the current average window cleaning price and is billed in a pay-as-you-go model.
Other software efforts are making Intel’s processors more appealing for AI
Since before it acquired Israel-based Habana Labs, Intel has tried to challenge NVIDIA’s grip on the AI training market. In Israel, Omdia saw a compelling updated pitch that end users should match their AI training applications to processors. Real-world use cases and new benchmark results demonstrated how traditional ML methods run well on Intel’s latest CPUs while deep learning is best computed on the massively parallel Habana Gaudi2. More than that, an expanded software library and in-house developed proof-of-concept applications convinced Omdia that Intel is approaching this space with renewed vigor—most likely to compensate for CPU market share loss.
Intel reminded us that AWS has given the thumbs up on Habana Gaudi and showed impressive gen-on-gen performance improvements. So long as TSMC can quickly manufacture enough of the co-processors, Intel’s fortunes in the deep learning acceleration market might change for the better.
In addition to the above, there were many more client computing-oriented software demonstrations and under-embargo announcements at the technology deep dive sessions Intel hosted in Israel. Differentiating through software is a good strategy at a time of processor performance parity or disadvantage. More importantly, Intel’s new commercial software and SaaS businesses are carving a new path for the company. People from Granulate, cnvrg.io, and the other startups Omdia met in Israel are solving real-world problems and have a unique perspective to offer. It is not a coincidence that Israel is known as “Startup Nation,” home to the highest number of tech companies per capita.
Vlad Galabov, Director, Cloud and Data Center Research